How-To Tutorials

This article is written by Steven Lott, author of the book Functional Python Programming. You can find more about him at http://slott-softwarearchitect.blogspot.com. (For more resources related to this topic, see here.) What's This About? We're going to look at various ways that Python 3 lets us define things which behave like functions. The proper term here is Callable – we're looking at objects that can be called like a function. We'll look at the following Python constructs: Function definitions Higher-order functions Function wrappers (around methods) Lambdas Callable objects Generator functions and the yield parameter And yes, we're aware that the list above has six items on it. That's because higher-order functions in Python aren't really all that complex or different. In some languages, functions that take functions are arguments involving special syntax. In Python, it's simple and common and barely worth mentioning as a separate topic. We'll look at when it's appropriate and inappropriate to use one or the other of these various functional forms. Some background Let's take a quick peek at a basic bit of mathematical formalism. We'll look at a function as an abstract formalism. We often annotate it like this: This shows us that f() is a function. It has one argument, x, and will map this to a single value, y. Some mathematical functions are written in front, for example, y=sin x. Some are written in other places around the argument, for example, y=|x|. In Python, the syntax is more consistent, for example, we use a function like this: >>> abs(-5)5 We've applied the abs() function to an argument value of -5. The argument value was mapped to a value of 5. Terminology Consider the following function: In this definition, the argument is a pair of values, (a,b). This is called the domain. We can summarize it as the domain of values for which the function is defined. Outside this domain, the function is not defined. In Python, we get a TypeError exception if we provide one value or three values as the argument. The function maps the domain pair to a pair of values, (q,r). This is the range of the function. We can call this the range of values that could be returned by the function. Mathematical function features As we look at the abstract mathematical definition of functions, we note that functions are generally assumed to have no hysteresis; they have no history or memory of prior use. This is sometimes called the property of being idempotent: the results are always the same for a given argument value. We see this in Python as a common feature. But it's not universally true. We'll look at a number of exceptions to the rule of idempotence. Here's an example of the usual situation: >>> int("10f1", 16)4337 The value returned from the evaluation of int("10f1", 16) never changes. There are, however, some common examples of non-idempotent functions in Python. Examples of hysteresis Here are three common situations where a function has hysteresis. In some cases, results vary based on history. In other cases, results vary based on events in some external environment, such as follows: Random number generators. We don't want them to produce the same value over and over again. The Python random.randrange() function, is not obviously idempotent. OS functions depend on the state of the machine as a whole. The os.listdir() function returns values that depend on the use of functions such as os.unlink(), os.rename(), and open() (among several others).While the rules are generally simple, it requires a stateful object outside the narrow world of the code itself. These are examples of Python functions that don't completely fit the formal mathematical definition; they lack idempotence, and their values depend on history, other functions, or both. Function Definitions Python has two statements that are essential features of function definition. The def statement specifies the domain and the return statement(s) specify the range. A simplified gloss of the syntax is as follows: def name(params):   body   return expression In effect, the function's domain is defined by the parameters provided in the def statement. This list of parameter names is not all the information on the domain, however. Even if we use one of the Python extensions to add type annotations, that's still not all the information. There may be if statements in the body of the function that impose additional explicit restrictions. There may be other functions that impose their own kind of implicit restrictions. If, for example, the body included math.sqrt() then there would be an implicit restriction on some values being non-negative. The return statements provide the function's range. An empty return statement means a range of simply None values. When there are multiple return statements, the range is the union of the ranges on all the return statements. This mapping between Python syntax and mathematical concepts isn't very complete. We need more information about a function. Example definition Here's an example of function definition: def odd(n):   """odd(n) -> boolean, true if n is odd."""   return n % 2 == 1 What do does this definition tell us? Several things such as: Domain: We know that this function accepts n, a single object. Range: Boolean value, True if n is an odd number. This is the most likely interpretation. It's also remotely possible that the class of n has repurposed __mod__() or __rmod__() methods, in which case the semantics can be pretty obscure. Because of the inherent ambiguity in Python, this function has provided a triple-quoted """Docstring""" parameter with a summary of the function. This is a best practice, and should be followed universally except in articles like this where it gets too long-winded to include a docstring parameter everywhere. In this case, the doctoring parameter doesn't state unambiguously that n is intended to be a number. There are two ways to handle this gap, they are as follows: Actually include words like n is a number in the docstring parameter Include the docstring parameter test cases that show the required behavior Either is acceptable. Both are preferable. Using a function To complete this example, here's how we'd use this odd little function named odd(): >>> odd(3)True>>> odd(4)False This kind of example text can be included into the docstring parameter to create two test cases that offer insight into what the function really means. The lack of declarations More verbose type declarations—as used in many popular programming languages—aren't actually enough information to fully specify a function's domain and range. To be rigorously complete, we need type definitions that include optional predicates. Take a look at the following command: isinstance(n,int) and n >= 0 The assert statement is a good place for this kind of additional argument domain checking. This isn't the perfect solution because assert statements can be disabled very easily. It can help during design and testing and it can help people to read your code. The fussy formal declarations of data type used in other languages are not really needed in Python. Python replaces an up-front claim about required types with a runtime search for appropriate class methods. This works because each Python object has all the type information bound into it. Static compile-time type information is redundant, since the runtime type information is complete. A Python function definition is pretty spare. In includes the minimal amount of information about the function. There are no formal declaration of parameter types or return type. This odd little function will work with any object that implements the % operator: Generally, this means any object that implements __mod__() or __rmod__(). This means most subclasses of numbers.Number. It also means instances of any class that happen to provide these methods. That could become very weird, but still possible. We hesitate to think about non-numeric objects that work with the number-like % operator. Some Python features In Python, functions we declare are proper first-class objects. This means that they have attributes that can be assigned to variables and placed into collections. Quite a few clever things can be done with function objects. One of the most elegant things is to use a function as an argument or a return value from another function. The ability to do this means that we can easily create and use higher-order functions in Python. For folks who know languages such as C (and C++), functions aren't proper first-class objects. A pointer to a function, however, is a first class object in C. But the function itself is a block of code that can't easily be manipulated. We'll look at a number of simple ways in which we can write—and use—higher-order functions in Python. Functions are objects Consider the following command example: >>> not_even = odd>>> not_even(3)True We've assigned the odd little function object to a new variable, not_even. This creates an alias for a function. While this isn't always the best idea, there are times when we might want to provide an alternate name for a function as part of maintaining reverse compatibility with a previous release of a library. Using functions Consider the following function definition: def some_test(function, value):   print(function, value)   return function(value) This function's domain includes arguments named function and value. We can see that it prints the arguments, then applies the function argument to the given value. When we use the preceding function, it looks like this: >>> some_test(odd, 3)<function odd at 0x613978> 3True The some_test() function accepted a function as an argument. When we printed the function, we got a summary, <function odd at 0x613978>, that shows us some information about the object. We also show a summary of the argument value, 3. When we applied the function to a value, we got the expected result. We can—of course—extend this concept. In particular, we can apply a single function to many values. Higher-order Functions Higher-order functions become particularly useful when we apply them to collections of objects. The built-in map() function applies a simple function to each value in an argument sequence. Here's an example: >>> list(map(odd, [1,2,3,4]))[True, False, True, False] We've used the map() function to apply the odd() function to each value in the sequence. This is a lot like evaluating: >>> [odd(x) for x in [1,2,3,4]] We've created a list comprehension instead of applying a higher-order map() function. This is equivalent to the following command snippet: [odd(1), odd(2), odd(3), odd(4)] Here, we've manually applied the odd() function to each value in a sequence. Yes, that's a diesel engine alternator and some hoses: We'll use this alternator as a subject for some concrete examples of higher-order functions. Diesel engine background Some basic diesel engine mechanics. The following some basic information: The engine turns the alternator. The alternator generates pulses that drive the tachometer. Amongst other things, like charging the batteries. The alternator provides an indirect measurement of engine RPMs. Direct measurement would involve connecting to a small geared shaft. It's difficult and expensive. We already have a tachometer; it's just incorrect. The new alternator has new wheels. The ratios between engine and alternator have changed. We're not interested in installing a new tachometer. Instead, we'll create a conversion from a number on the tachometer, which is calibrated to the old alternator, to a proper number of engine RPMs. This has to allow the change in ratio between the original tachometer and the new tach. Let's collect some data and see what we can figure out about engine RPMs. New alternator First approximation: all we did was get new wheels. We can presume that the old tachometer was correct. Since the new wheel is smaller, we'll have higher alternator RPMs. That means higher readings on the old tachometer. Here's the key question: How far wrong are the RPMs? The old wheel was approximately 3.5 RPM and the new wheel is approximately 2.5 RPM. We can compute the potential ratio between what the tach says and what the engine is really doing: >>> 3.5/2.51.4>>> 1/_0.7142857142857143 That's nice. Is it right? Can we really just multiply and display RPMs by .7 to get actual engine RPMs? Let's create the conversion card first, then collect some more data. Use case Given RPM on the tachometer, what's the real RPM of the engine? Use the following command to find the RPM: def eng(r):   return r/1.4 Use it like the following: >>> eng(2100)1500.0 This seems useful. Tach says 2100, engine (theoretically) spinning at 1500, more or less. Let's confirm our hypothesis with some real data. Data collection Over a period of time, we recorded tachometer readings and actual RPMs using a visual RPM measuring device. The visual device requires a strip of reflective tape on one of the engine wheels. It uses a laser and counts returns per minute. Simple. Elegant. Accurate. It's really inconvenient. But it got some data we could digest. Skipping some boring statistics, we wind up with the following function that maps displayed RPMs to actual RPMs, such as this: def eng2(r):   return 0.7724*r**1.0134 Here's a sample result: >>> eng2(2100)1797.1291903589386 When tach says 2100, the engine is measured as spinning at about 1800 RPM. That's not quite the same as the theoretical model. But it's so close that it gives us a lot of confidence in this version. Of course, the number displayed is hideous. All that floating-point cruft is crazy. What can we do? Rounding is only part of the solution. We need to think through the use case. After all, we use this standing at the helm of the boat; how much detail is appropriate? Limits and ranges The engine has governors and only runs between 800 and 2500 RPM. There's a very tight limit here. Realistically, we're talking about this small range of values: >>> list(range(800, 2500, 200))[800, 1000, 1200, 1400, 1600, 1800, 2000, 2200, 2400] There's no sensible reason for proving any more detailed engine RPMs. It's a sailboat; top speed is 7.5 knots (Nautical miles per hour). Wind and current have far more impact on the boat speed than the difference between 1600 and 1700 RPMs. The tach can't be read to closer than 100-200 RPM. It's not digital, it's a red pointer near little tick lines. There's no reason to preserve more than a few bits of precision. Example of Tach translation Given the engine RPMs and the conversion function, we can deduce that the tachometer display will be between 1000 to 3200. This will map to engine RPMs in the range of about 800 to 2500. We can confirm this with a mapping like this: >>> list(map(eng2, range(1000,3200,200)))[847.3098694826986, 1019.258964596305, 1191.5942982618956, 1364.2609728487703, 1537.2178605443924, 1710.4329833319157, 1883.8807562755746, 2057.5402392829747, 2231.3939741669838, 2405.4271806626366, 2579.627182659544] We've applied the eng2() mapping from tach to engine RPM. For tach readings between 1000 and 3200 in steps of 200, we've computed the actual engine RPMs. For those who use spreadsheets a lot, the range() function is like filling a column with values. The map(eng2, …) function is like filling an adjacent column with a calculation. We've created the result of applying a function to each value of a given range. As shown, this is little difficult to use. We need to do a little more cleanup. What other function do we need to apply to the results? Round to 100 Here's a function that will round up to the nearest 100: def next100(n):   return int(round(n, -2)) We could call this a kind of composite function built from a partial application of round() and int() functions. If we map this function to the previous results, we get something a little easier to work with. How does this look? >>> tach= range(1000,3200,200)>>> list(map(next100, map(eng2, tach)))[800, 1000, 1200, 1400, 1500, 1700, 1900, 2100, 2200, 2400, 2600] This expression is a bit complex; let's break it down into three discrete steps: First, map the eng2() function to tach numbers between 1000 and 3200. The result is effectively a sequence of values (it's not actually a list, it's a generator, a potential list) Second, map the next100() function to results of previous mapping Finally, collect a single list object from the results We've applied two functions, eng2() and next100(), to a list of values. In principle, we've created a kind of composite function, next100○eng20(rpm). Python doesn't support function composition directly, hence the complex-looking map of map syntax. Interleave sequences of values The final step is to create a table that shows both the tachometer reading and the computed engine RPMs. We need to interleave the input and output values into a single list of pairs. Here are the tach readings we're working with, as a list: >>> tach= range(1000,3200,200) Here are the engine RPMs: >>> engine= list(map(next100,map(eng2,tach))) Here's how we can interleave the two to create something that shows our tachometer reading and engine RPMs: >>> list(zip(tach, engine))[(1000, 800), (1200, 1000), (1400, 1200), (1600, 1400), (1800, 1500), (2000, 1700),(2200, 1900), (2400, 2100), (2600, 2200), (2800, 2400), (3000, 2600)] The rest is pretty-printing. What's important is that we could take functions like eng() or eng2() and apply it to columns of numbers, creating columns of results. The map() function means that we don't have to write explicit for loops to simply apply a function to a sequence of values. Map is lazy We have a few other observations about the Python higher-order functions. First, these functions are lazy, they don't compute any results until required by other statements or expressions. Because they don't actually create intermediate list objects, they may be quite fast. The laziness feature is true for the built-in higher-order functions map() and filter(). It's also true for many of the functions in the itertools library. Many of these functions don't simply create a list object, they yield values as requested. For debugging purposes, we use list() to see what's being produced. If we don't apply list() to the result of a lazy function, we simply see that it's a lazy function. Here's an example: >>> map(lambda x:x*1.4, range(1000,3200,200))<map object at 0x102130610> We don't see a proper result here, because the lazy map() function didn't do anything. The list(), tuple(), or set() functions will force a lazy map() function to actually get up off the couch and compute something. Function Wrappers There are a number of Python functions which are syntactic sugar for method functions. One example is the len() function. This function behaves as if it had the following definition: def len(obj):   return obj.__len__() The function acts like it's simply invoking the object's built-in __len__() method. There are several Python functions that exist only to make the syntax a little more readable. Post-fix syntax purists would prefer to see syntax such as some_list.len(). Those who like their code to look a little more mathematical prefer len(some_list). Some people will go so far as to claim that the presence of prefix functions means that Python isn't strictly object-oriented. This is false; Python is very strictly object-oriented. It doesn't—however—use only postfix method notation. We can write function wrappers to make some method functions a little more palatable. Another good example is the divmod() function. This relies on two method functions, such as the following: a.__divmod__(b) b.__rdivmod__(a) The usual operator rules apply here. If the class for object a implements __divmod__(), then that's used to compute the result. If not, then the same test is made for the class of object b; if there's an implementation, that will be used to compute the results. Otherwise, it's undefined and we'll get an exception. Why wrap a method? Function wrappers for methods are syntactic sugar. They exist to make object methods look like simple functions. In some cases, the functional view is more succinct and expressive. Sometimes the object involved is obvious. For example, the os module functions provide access to OS-level libraries. The OS object is concealed inside the module. Sometimes the object is implied. For example, the random module makes a Random instance for us. We can simply call random.randint() without worrying about the object that was required for this to work properly. Lambdas A lambda is an anonymous function with a degenerate body. It's like a function in some respects and it's unlike a function because of the following two things: A lambda has no name A lambda has no statements A lambda's body is a single expression, nothing more. This expression can have parameters, however, which is why a lambda is a handy form of a callable function. The syntax is essentially as follows: lambda params : expression Here's a concrete example: lambda r: 0.7724*r**1.0134 You may recognize this as the eng2() function defined previously. We don't always need a complete, formal function. Sometimes, we just need an expression that has parameters. Speaking theoretically, a lambda is a one-argument function. When we have multi-argument functions, we can transform it to a series of one-argument lambda forms. This transformation can be helpful for optimization. None of that applies to Python. We'll move on. Using a Lambda with map Here are two equivalent results: map(eng2, tach) map(lambda r: 0.7724*r**1.0134, tach) Here's a previous example, using the lambda instead of the function: >>> tach= range(1000,3200,200)>>> list( map(lambda r: 0.7724*r**1.0134, tach))[847.3098694826986, 1019.258964596305, 1191.5942982618956, 1364.2609728487703, 1537.2178605443924, 1710.4329833319157, 1883.8807562755746, 2057.5402392829747, 2231.3939741669838, 2405.4271806626366, 2579.627182659544] You could scroll back to see that the results are the same. If we're doing a small thing once only, a lambda object might be more clear than a complete function definition. Emphasis here is on small once only. If we start trying to reuse a lambda object, or feel the need to assign a lambda object to a variable, we should really consider a function definition and the associated docstring and doctest features. Another use of Lambdas A common use of lambdas is with three other higher-order functions: sort(), min(), and max(). We might use one of these with a list object: list.sort(key= lambda x: expr) list.min(key= lambda x: expr) list.max(key= lambda x: expr) In each case, we're using a lambda object to embed an expression into the argument values for a function. In some cases, the expression might be very sophisticated; in other cases, it might be something as trivial as lambda x: x[1]. When the expression is trivial, a lambda object is a good idea. If the expression is going to get reused, however, a lambda object might be a bad idea. You can do this… But… The following kind of statement makes sense: some_name = lambda x: 3*x+1 We've created a callable object that takes a single argument value and returns a numeric value such as the following command snippet: def some_name(x): return 3*x+1. There are some differences. Most notably the following: A lambda object is all on one line of code. A possible advantage. There's no docstring. A disadvantage for lambdas of any complexity. Nor is there any doctest in the missing docstring. A significant problem for a lambda object that requires testing. There are ways to test lambdas with doctest outside a docstring, but it seems simpler to switch to a full function definition. We can't easily apply decorators to it. To do it, we lose the @decorator syntax. We can't use any Python statements in it. In particular, no try-except block is possible. For these reasons, we suggest limiting the use of lambdas to truly trivial situations. Callable Objects A callable object fits the model of a function. The unifying feature of all of the things we've looked at is that they're callable. Functions are the primary example of being callable but objects can also be callable. Callable objects can be subclasses of collections.abc.Callable. Because of Python's flexibility, this isn't a requirement, it's merely a good idea. To be callable, a class only needs to provide a __call__() method. Here's a complete callable class definition: from collections.abc import Callableclass Engine(Callable):   def __call__(self, tach):       return 0.7724*tach**1.0134 We've imported the collections.abc.Callable class. This will provide some assurance that any class that extends this abstract superclass will provide a definition for the __call__() method. This is a handy error-checking feature. Our class extends Callable by providing the needed __call__() method. In this case, the __call__() method performs a calculation on the single parameter value, returning a single result. Here's a callable object built from this class: eng= Engine() This creates a function that we can then use. We can evaluate eng(1000) to get the engine RPMs when the tach reads 1000. Callable objects step-by-step There are two parts to making a function a callable object. We'll emphasize these for folks who are new to object-oriented programming: Define a class. Generally, we make this a subclass of collections.abc.Callable. Technically, we only need to implement a __call__() method. It helps to use the proper superclass because it might help catch a few common mistakes. Create an instance of the class. This instance will be a callable object. The object that's created will be very similar to a defined function. And very similar to a lambda object that's been assigned to a variable. While it will be similar to a def statement, it will have one important additional feature: hysteresis. This can be the source of endless bugs. It can also be a way to improve performance. Callables can have hysteresis Here's an example of a callable object that uses hysteresis as a kind of optimization: class Factorial(Callable):   def __init__(self):       self.previous = {}   def __call__(self, n):       if n not in self.previous:           self.previous[n]= self.compute(n)       return self.previous[n]   def compute(self, n):       if n == 0 : return 1       return n*self.__call__(n-1)Here's how we can use this:>>> fact= Factorial()>>> fact(5)120 We create an instance of the class, and then call the instance to compute a value for us. The initializer The initialization method looks like this:    def __init__(self):       self.previous = {} This function creates a cache of previously computed values. This is a technique called memoization. If we've already computed a result once, it's in the self.previous cache; we don't need to compute it again, we already know the answer. The Callable interface The required __call__() method looks like this:    def __call__(self, n):       if n not in self.previous:           self.previous[n]= self.compute(n)       return self.previous[n] We've checked the memoization cache first. If the value is not there, we're forced to compute the answer, and insert it into the cache. The final answer is always a value in the cache. A common what if question is what if we have a function of multiple arguments? There are two minuscule changes to support more complex arguments. Use def __call__(self, *n): and self.compute(*n). Since we're only computing factorial, there's no need to over-generalize. The Compute method The essential computation has been allocated to a method called compute. It looks like this:    def compute(self, n):       if n == 0: return 1           return n*self.__call__(n-1) This does the real work of the callable object: it computes n!. In this case, we've used a pretty standard recursive factorial definition. This recursion relies on the __call__() method to check the cache for previous values. If we don't expect to compute values larger than 1000! (a 2,568 digit number, by the way) the recursion works nicely. If we think we need to compute really large factorials, we'll need to use a different approach. Execute the following code to compute very large factorials: functools.reduce(operator.mul, range(1,n+1)) Either way, we can depend on the internal memoization to leverage previous results. Note the potential issue Hysteresis—memory of what came before—is available to the callable objects. We call functions and lambdas stateless, where callable objects can be stateful. This may be desirable to optimize performance. We can memoize the previous results or we can design an object that's simply confusing. Consider a function like divmod() that returns two values. We could try to define a callable object that first returns the quotient and on the second call with the same arguments returns the remainder: >>> crazy_divmod(355,113)3>>> crazy_divmod(255,113)16 This is technically possible. But it's crazy. Warning: Stay away. We generally expect idempotence: functions do the same thing each time. Implementing memoization didn't alter the basic idempotence of our factorial function. Generator Functions Here's a fun generator, the Collatz function. The function creates a sequence using a simple pair of rules. We'll could call this rule, Half-Or-Three-Plus-One (HOTPO). We'll call it collatz(): def collatz(n):   if n % 2 == 0:        return n//2   else:       return 3*n+1 Each integer argument yields a next integer. These can form a chain. For example, if we start with collatz(13), we get 40. The value of collatz(40) is 20. Here's the sequence of values: 13 → 40 → 20 → 10 → 5 → 16 → 8 → 4 → 2 → 1At 1, it loops: 1 → 4 → 2 → 1 … Interestingly, all chains seem to lead—eventually—to 1. To explore this, we need a simple function that will build a chain from a given starting value. Successive values Here's a generator function that will build a list object. This iterates through values in the sequence until it reaches 1, when it terminates: def col_list(n):   seq= [n]   while n != 1:       n= collatz(n)       seq.append(n)   return seq This is not wrong. But it's not really the most useful implementation. This always creates a sequence object. In many cases, we don't really want an object, we only want information about the sequence. We might only want the length, or the largest numbers, or the sum. This is where a generator function might be more useful. A generator function yields elements from the sequence instead of building the sequence as a single object. Generator functions To create a generator function, we write a function that has a loop; inside the loop, there's a yield statement. A function with a yield statement is effectively an Iterable object, it can be used in a for statement to produce values. It doesn't create a big list object, it creates the items that can be accumulated into a list or tuple object. A generator function is lazy: it doesn't compute anything unless forced to by another function needing results. We can iterate through as many (or as few) of the results as we need. For example, list(some_generator()) forces all values to be returned. For another example of a lazy generator, look at how range() objects work. If we evaluate range(10), we only get a generator. If we evaluate list(range(10)), we get a list object. The Collatz generator Here's a generator function that will produce sequences of values using the collatz() method rule shown previously: def col_iter(n):   yield n   while n != 1:       n= collatz(n)        yield n When we use this in a for loop or with the list() function, this will yield the argument number. While the number is not 1, it will apply the collatz() function and yield successive values in the chain. When it has yielded 1, it will will terminate. One common pattern for generator functions is to replace all list-accumulation statements with yield statements. Instead of building a list one time at a time, we yield each item. The collatz() function it lazy. We don't get an answer unless we use list() or tuple() or some variation of a for statement context. Using a generator function Here's how this function looks in practice: >>> for i in col_iter(3):…   print(i)3105168421 We've used the generator function in a for loop so that it will yield all of the values until it terminates. Collatz function sequences Now we can do some exploration of this Collatz sequence. Here are a few evaluations of the col_iter() function: >>> list(col_iter(3))[3, 10, 5, 16, 8, 4, 2, 1]>>> list(col_iter(5))[5, 16, 8, 4, 2, 1]>>> list(col_iter(6))[6, 3, 10, 5, 16, 8, 4, 2, 1]>>> list(syracuse_iter(13))[13, 40, 20, 10, 5, 16, 8, 4, 2, 1] There's an interesting pattern here. It seems that from 16, we know the rest. Generalizing this: from any number we've already seen, we know the rest. Wait. This means that memoization might be a big help in exploring the values created by this sequence. When we start combining function design patterns like this, we're doing functional programming. We're stepping outside the box of purely object-oriented Python. Alternate styles Here is an alternative version of the collatz() function: def collatz2(n):   return n//2 if n%2 == 0 else 3*n+1 This simply collapses the if statements into a single if expression and may not help much. We also have this: collatz3= lambda n: n//2 if n%2 == 0 else 3*n+1 We've collapsed the expression into a lambda object. Helpful? Perhaps not. On the other hand, the function doesn't really need all of the overhead of a full function definition and multiple statements. The lambda object seems to capture everything relevant. Functions as object There's a higher-level function that will produce values until some ending condition is met. We can plug in one of the versions of the collatz() function and a termination test into this general-purpose function: def recurse_until(ending, the_function, n):   yield n   while not ending(n):       n= the_function(n)       yield n This requires two plug-in functions, they are as follows: ending() is a function to test to see whether we're done, for example, lambda n: n==1 the_function() is a form of the Collatz function We've completely uncoupled the general idea of recursively applying a function from a specific function and a specific terminating condition. Using the recurs_until() function We can apply this higher-order recurse_until() function like this: >>> recurse_until(lambda n: n==1, syracuse2, 9)<generator object recurse_until at 0x1021278c0> What's that? That's how a lazy generator looks: it didn't return any values because we didn't demand any values. We need to use it in a loop or some kind of expression that iterates through all available values. The list() function, for example, will collect all of the values. Getting the list of values Here's how we make the lazy generator do some work: >>> list(_)[9, 28, 14, 7, 22, 11, 34, 17, 52, 26, 13, 40, 20, 10, 5, 16, 8, 4, 2, 1] The _ variable is the previously computed value. It relieves us from the burden of having to write an assignment statement. We can write an expression, see the results, and know the results were automatically saved in the _ variable. Project Euler #14 Which starting number, under one million, produces the longest chain? Try it without memoization. It's really simple: >>> collatz_len= [len(list(recurse_until(lambda n: n==1, collatz2, i))) ... for i in range(1,11)]>>> results = zip(collatz_len, range(1,11))>>> max(results)(20, 9)>>> list(col_iter(9))[9, 28, 14, 7, 22, 11, 34, 17, 52, 26, 13, 40, 20, 10, 5, 16, 8, 4, 2, 1] We defined collatz_len as a list. We're writing a list comprehension that shows the values built from a generator expression. The generator expression evaluates len(something) for i in range(1,11). This means we'll be collecting ten values into the list, each of which is the length of something. The something is a list object built from the recurse_until(lambda n: n==1, collatz2, i) function that we discussed. This will compute the sequence of Collatz values starting from i and proceeding until the value in the sequence is 1. We've zipped the lengths with the original values of i. This will create pairs of lengths and starting numbers. The maximum length will now have a starting value associated with it so that we can confirm that the results match our expectations. And yes, this Project Euler problem could—in principle—be solved in a single line of code. Will this scale to 100? 1,000? 1,000,000? How much will memoization help? Summary In this article, we've looked at five (or six) kinds of Python callables. They all fit the y = f(x) model of a function to varying degrees. When is it appropriate to use each of these different ways to express the same essential concept? Functions are created with def and return. It shouldn't come as a surprise that this should cover most cases. This allows a docstring comment and doctest test cases. We could call these def functions, since they're built with the def statement. Higher-order functions—map(), filter(), and the itertools library—are generally written as plain-old def functions. They're higher-order because they accept functions as arguments or return functions as results. Otherwise, they're just functions. Function wrappers—len(), divmod(), pow(), str(), and repr()—are function syntax wrapped around object methods. These are def'd functions with very tiny bodies. We use them because a.pow(2) doesn't seem as clear as pow(2,a). Lambdas are appropriate for one-time use of something so simple that it doesn't deserve being wrapped in a def statement body. In some cases, we have a small nugget of code that seems more clear when written as a lambda expression rather than a more complete function definition. Simple filter rules, and simple computations are often more clearly shown as a lambda object. The Callable objects have a special property that other functions lack: hysteresis. They can retain the results of previous calculations. We've used this hysteresis property to implement memoizing. This can be a huge performance boost. Callable objects can be used badly, however, to create objects that have simply bizarre behavior. Most functions should strive for idempotence—the same arguments should yield the same results. Generator functions are created with a def statement and at least one yield statement. These functions are iterable. They can be used in a for statement to examine each resulting value. They can also be used with functions like list(), tuple(), and set() to create an actual object from the iterable sequence of values. We might combine them with higher-order functions to do complex processing, one item at a time. It's important to work with each of these kinds of callables. If you only have one tool—a hammer—then every problem has to be reshaped into a nail before you can solve it. Once you have multiple tools available, you can pick the tools that provides the most succinct and expressive solution to the problem. Resources for Article: Further resources on this subject: Expert Python Programming [article] Python Network Programming Cookbook [article] Learning Python Design Patterns [article]

"One server to satisfy them all" could have been the name of this article by David Lewin, the author of BeagleBone Media Center. We now have a great media server where we can share any media, but we would like to be more independent so that we can choose the functionalities the server can have. The goal of this article is to let you cross the bridge, where you are going to increase your knowledge by getting your hands dirty. After all, you want to build your own services, so why not create your own contents as well. (For more resources related to this topic, see here.) More specifically, here we will begin by building a webcam streaming service from scratch, and we will see how this can interact with what we have implemented previously in the server. We will also see how to set up a service to retrieve RSS feeds. We will discuss the services in the following sections: Installing and running MJPG-Streamer Detecting the hardware device and installing drivers and libraries for a webcam Configuring RSS feeds with Leed Detecting the hardware device and installing drivers and libraries for a webcam Even though today many webcams are provided with hardware encoding capabilities such as the Logitech HD Pro series, we will focus on those without this capability, as we want to have a low budget project. You will then learn how to reuse any webcam left somewhere in a box because it is not being used. At the end, you can then create a low cost video conference system as well. How to know your webcam As you plug in the webcam, the Linux kernel will detect it, so you can read every detail it's able to retrieve about the connected device. We are going to see two ways to retrieve the webcam we have plugged in: the easy one that is not complete and the harder one that is complete. "All magic comes with a price."                                                                                     –Rumpelstiltskin, Once Upon a Time Often, at a certain point in your installation, you have to choose between the easy or the hard way. Most of the time, powerful Linux commands or tools are not thought to be easy at first but after some experiments you'll discover that they really can make your life better. Let's start with the fast and easy way, which is lsusb : debian@arm:~$ lsusb Bus 001 Device 002: ID 046d:0802 Logitech, Inc. Webcam C200 Bus 001 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub Bus 002 Device 001: ID 1d6b:0002 Linux Foundation 2.0 root hub This just confirms that the webcam is running well and is seen correctly from the USB. Most of the time we want more details, because a hardware installation is not exactly as described in books or documentations, so you might encounter slight differences. This is why the second solution comes in. Among some of the advantages, you are able to know each step that has taken place when the USB device was discovered by the board and Linux, such as in a hardware scenario: debian@arm:~$ dmesg A UVC device (here, a Logitech C200) has been used to obtain these messages Most probably, you won't exactly have the same outputs, but they should be close enough so that you can interpret them easily when they are referred to: New USB device found: This is the main message. In case of any issue, we will check its presence elsewhere. This message indicates that this is a hardware error and not a software or configuration error that you need to investigate. idVendor and idProduct: This message indicates that the device has been detected. This information is interesting so you can check the constructor detail. Most recent webcams are compatible with the Linux USB Video Class (UVC), you can check yours at http://www.ideasonboard.org/uvc/#devices. Among all the messages, you should also look for the one that says Registered new interface driver interface because failing to find it can be a clue that Linux could detect the device but wasn't able to install it. The new device will be detected as /dev/video0. Nevertheless, at start, you can see your webcam as a different device name according to your BeagleBone configuration, for example, if a video capable cape is already plugged in. Setting up your webcam Now we know what is seen from the USB level. The next step is to use the crucial Video4Linux driver, which is like a Swiss army knife for anything related to video capture: debian@arm:~$ Install v4l-utils The primary use of this tool is to inquire about what the webcam can provide with some of its capabilities: debian@arm:~$ v4l2-ctl -–all There are four distinctive sections that let you know how your webcam will be used according to the current settings: Driver info (1) : This contains the following information: Name, vendor, and product IDs that we find in the system message The driver info (the kernel's version) Capabilities: the device is able to provide video streaming Video capture supported format(s) (2): This contains the following information: What resolution(s) are to be used. As this example uses an old webcam, there is not much to choose from but you can easily have a lot of choices with devices nowadays. The pixel format is all about how the data is encoded but more details can be retrieved about format capabilities (see the next paragraph). The remaining stuff is relevant only if you want to know in precise detail. Crop capabilities (3): This contains your current settings. Indeed, you can define the video crop window that will be used. If needed, use the crop settings: --set-crop-output=top=<x>,left=<y>,width=<w>,height=<h> Video input (4): This contains the following information: The input number. Here we have used 0, which is the one that we found previously. Its current status. The famous frames per second, which gives you a local ratio. This is not what you will obtain when you'll be using a server, as network latencies will downgrade this ratio value. You can grab capabilities for each parameter. For instance, if you want to see all the video formats the webcam can provide, type this command: debian@arm:~$ v4l2-ctl --list-formats Here, we see that we can also use MJPEG format directly provided by the cam. While this part is not mandatory, such a hardware tour is interesting because you know what you can do with your device. It is also a good habit to be able to retrieve diagnostics when the webcam shows some bad signs. If you would like to get more in depth knowledge about your device, install the uvcdynctrl package, which lets you retrieve all the formats and frame rates supported. Installing and running MJPG-Streamer Now that we have checked the chain from the hardware level up to the driver, we can install the software that will make use of Video4Linux for video streaming. Here comes MJPG-Streamer. This application aims to provide you with a JPEG stream on the network available for browsers and all video applications. Besides this, we are also interested in this solution as it's made for systems with less advanced CPU, so we can start MJPG-Streamer as a service. With this streamer, you can also use the built-hardware compression and even control webcams such as pan, tilt, rotations, zoom capabilities, and so on. Installing MJPG-Streamer Before installing MJPG-Streamer, we will install all the necessary dependencies: debian@arm:~$ install subversion libjpeg8-dev imagemagick Next, we will retrieve the code from the project: debian@arm:~$ svn checkout http://svn.code.sf.net/p/mjpg-streamer/code/ mjpg-streamer-code You can now build the executable from the sources you just downloaded by performing the following steps: Enter the following into the local directory you have downloaded: debian@arm:~$ cd mjpg-streamer-code/mjpg-streamer Then enter the following command: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ make When the compilation is complete, we end up with some new files. From this picture the new green files are produced from the compilation: there are the executables and some plugins as well. That's all that is needed, so the application is now considered ready. We can now try it out. Not so much to do after all, don't you think? Starting the application This section aims at getting you started quickly with MJPG-Streamer. At the end, we'll see how to start it as a service on boot. Before getting started, the server requires some plugins to be copied into the dedicated lib directory for this purpose: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ sudo cp input_uvc.so output_http.so /usr/lib The MJPG-Streamer application has to know the path where these files can be found, so we define the following environment variable: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ export LD_LIBRARY_PATH=/usr/ lib;$LD_LIBRARY_PATH Enough preparation! Time to start streaming: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$./mjpg_streamer -i "input_uvc.so" -o "output_http.so -w www" As the script starts, the input parameters that will be taken into consideration are displayed. You can now identify this information, as they have been explained previously: The detected device from V4L2 The resolution that will be displayed, according to your settings Which port will be opened Some controls that depend on your camera capabilities (tilt, pan, and so on) If you need to change the port used by MJPG-Streamer, add -p xxxx at the end of the command, which is shown as follows: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ ./mjpg_streamer -i "input_uvc.so" -o "output_http.so -w www –p 1234" Let's add some security If you want to add some security, then you should set the credentials: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ ./mjpg-streamer -o "output_http.so -w ./www -c debian:temppwd" Credentials can always be stolen and used without your consent. The best way to ensure that your stream is confidential all along would be to encrypt it. So if you intend to use strong encryption for secured applications, the crypto-cape is worth taking a look at http://datko.net/2013/10/03/howto_crypto_beaglebone_black/. "I'm famous" – your first stream That's it. The webcam is made accessible to everyone across the network from BeagleBone; you can access the video from your browser and connect to http://192.168.0.15:8080/. You will then see the default welcome screen, bravo!: Your first contact with the MJPG-Server You might wonder how you would get informed about which port to use among those already assigned. Using our stream across the network Now that the webcam is available across the network, you have several options to handle this: You can use the direct flow available from the home page. On the left-hand side menu, just click on the stream tab. Using VLC, you can open the stream with the direct link available at http://192.168.0.15:8080/?action=stream.The VideoLAN menu tab is a M3U-playlist link generator that you can click on. This will generate a playlist file you can open thereafter. In this case, VLC is efficient, as you can transcode the webcam stream to any format you need. Although it's not mandatory, this solution is the most efficient, as it frees the BeagleBone's CPU so that your server can focus on providing services. Using MediaDrop, we can integrate this new stream in our shiny MediaDrop server, knowing that currently MediaDrop doesn't support direct local streams. You can create a new post with the related URL link in the message body, as shown in the following screenshot: Starting the streaming service automatically on boot In the beginning, we saw that MJPG-Streamer needs only one command line to be started. We can put it in a bash script, but servicing on boot is far better. For this, use a console text editor – nano or vim – and create a file dedicated to this service. Let's call it start_mjpgstreamer and add the following commands: #! /bin/sh # /etc/init.d/start_mjpgstreamer export LD_LIBRARY_PATH="/home/debian/mjpg-streamer/mjpg-streamer-code/ mjpg-streamer;$LD_LIBRARY_PATH" EXEC_PATH="/home/debian/mjpg-streamer/mjpg-streamer-code/mjpg-streamer" $EXEC_PATH/mjpg_streamer -i "input_uvc.so" -o "output_http.so -w EXEC_PATH /www" You can then use administrator rights to add it to the services: debian@arm:~$ sudo /etc/init.d/start_mjpgstreamer start On the next reboot, MJPG-Streamer will be started automatically. Exploring new capabilities to install For those about to explore, we salute you! Plugins Remember that at the beginning of this article, we began the demonstration with two plugins: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ ./mjpg_streamer -i "input_uvc.so" -o "output_http.so -w www" If we take a moment to look at these plugins, we will understand that the first plugin is responsible for handling the webcam directly from the driver. Simply ask for help and options as follows: debian@beaglebone:~/mjpg-streamer-code/mjpg-streamer$ ./mjpg_streamer --input "input_uvc.so --help" The second plugin is about the web server settings: The path to the directory contains the final web server HTML pages. This implies that you can modify the existing pages with a little effort or create new ones based on those provided. Force a special port to be used. Like I said previously, port use is dedicated for a server. You define here which will be the one for this service. You can discover many others by asking: debian@arm:~$ ./mjpg_streamer --output "output_http.so --help" Apart from input_uvc and output_http, you have other available plugins to play with. Let's take a look at the plugins directory. Another tool for the webcam The Mjpg_streamer project is dedicated for streaming over network, but it is not the only one. For instance, do you have any specific needs such as monitoring your house/son/cat/Jon Snow figurine? buuuuzzz: if you answered yes to the last one, you just defined yourself as a geek. Well, in that case the Motion project is for you; just install the motion package and start it with the default motion.conf configuration. You will then record videos and pictures of any moving object/person that will be detected. As MJPG-Streamer motion aims to be a low CPU consumer, it works very well on BeagleBone Black. Configuring RSS feeds with Leed Our server can handle videos, pictures, and music from any source and it would be cool to have another tool to retrieve news from some RSS providers. This can be done with Leed, a RSS project organized for servers. You can have a final result, as shown in the following screenshot: This project has a "quick and easy" installation spirit, so you can give it a try without harness. Leed (for Light Feed) allows you to you access RSS feeds from any browser, so no RSS reader application is needed, and every user in your network can read them as well. You install it on the server and feeds are automatically updated. Well, the truth behind the scenes is that a cron task does this for you. You will be guided to set some synchronisation after the installation. Creating the environment for Leed in three steps We already have Apache, MySQL, and PHP installed, and we need a few other prerequisites to run Leed: Create a database for Leed Download the project code and set permissions Install Leed itself Creating a database for Leed You will begin by opening a MySQL session: debian@arm:~$ mysql –u root –p What we need here is to have a dedicated Leed user with its database. This user will be connected using the following: create user 'debian_leed'@'localhost' IDENTIFIED BY 'temppwd'; create database leed_db; use leed_db; grant create, insert, update, select, delete on leed_db.* to debian_leed@localhost; exit Downloading the project code and setting permissions We prepared our server to have its environment ready for Leed, so after getting the latest version, we'll get it working with Apache by performing the following steps: From your home, retrieve the latest project's code. It will also create a dedicated directory: debian@arm:~$ git clone https://github.com/ldleman/Leed.git debian@arm:~$ ls mediadrop mjpg-streamer Leed music Now, we need to put this new directory where the Apache server can find it: debian@arm:~$ sudo mv Leed /var/www/ Change the permissions for the application: debian@arm:~$ chmod 777 /var/www/Leed/ -R Installing Leed When you go to the server address (http//192.168.0.15/leed/install.php), you'll get the following installation screen: We now need to fill in the database details that we previously defined and add the Administrator credentials as well. Now save and quit. Don't worry about the explanations, we'll discuss these settings thereafter. It's important that all items from the prerequisites list on the right are green. Otherwise, a warning message will be displayed about the wrong permissions settings, as shown in the following screenshot: After the configuration, the installation is complete: Leed is now ready for you. Setting up a cron job for feed updates If you want automatic updates for your feeds, you'll need to define a synchronization task with cron: Modify cron jobs: debian@arm:~$ sudo crontab –e Add the following line: 0 * * * * wget -q -O /var/www/leed/logsCron "http://192.168.0.15/Leed/action.php?action=synchronize Save it and your feeds will be refreshed every hour. Finally, some little cleanup: remove install.php for security matters: debian@arm:~$ rm /var/www/Leed/install.php Using Leed to add your RSS feed When you need to add some feeds from the Manage menu, in Feed Options (on the right- hand side) select Preferences and you just have to paste the RSS link and add it with the button: You might find it useful to organize your feeds into groups, as we did for movies in MediaDrop. The Rename button will serve to achieve this goal. For example, here a TV Shows category has been created, so every feed related to this type will be organized on the main screen. Some Leed preferences settings in a server environment You will be asked to choose between two synchronisation modes: Complete and Graduated. Complete: This isto be used in a usual computer, as it will update all your feeds in a row, which is a CPU consuming task Graduated: Look for the oldest 10 feeds and update them if required You also have the possibility of allowing anonymous people to read your feeds. Setting Allow anonymous readers to Yeswill let your guests access your feeds but not add any. Extending Leed with plugins If you want to extend Leed capabilities, you can use the Leed Market—as the author defined it—from Feed options in the Manage menu. There, you'll be directed to the Leed Market space. Installation is just a matter of downloading the ZIP file with all plugins: debian@arm:~/Leed$ wget  https://github.com/ldleman/Leed-market/archive/master.zip debian@arm:~/Leed$ sudo unzip master.zip Let's use the AdBlock plugin for this example: Copy the content of the AdBlock plugin directory where Leed can see it: debian@arm:~/Leed$ sudo cp –r Leed-market-master/adblock /var/www/Leed/plugins Connect yourself and set the plugin by navigating to Manage | Available Plugins and then activate adblock withEnable, as follows: In this article, we covered: Some words about the hardware How to know your webcam Configuring RSS feeds with Leed Summary In this article, we had some good experiments with the hardware part of the server "from the ground," to finally end by successfully setting up the webcam service on boot. We discovered hardware detection, a way to "talk" with our local webcam and thus to be able to see what happens when we plug a device in the BeagleBone. Through the topics, we also discovered video4linux to retrieve information about the device, and learned about configuring devices. Along the way, we encountered MJPG-Streamer. Finally, it's better to be on our own instead of being dependent on some GUI interfaces, where you always wonder where you need to click. Finally, our efforts have been rewarded, as we ended up with a web page we can use and modify according to our tastes. RSS news can also be provided by our server so that you can manage all your feeds in one place, read them anywhere, and even organize dedicated groups. Plenty of concepts have been seen for hardware and software. Then think of this article as a concrete example you can use and adapt to understand how Linux works. I hope you enjoyed this freedom of choice, as you drag ideas and drop them in your BeagleBone as services. We entered in the DIY area, showing you ways to explore further. You can argue, saying that we can choose the software but still use off the shelf commercial devices. Resources for Article: Further resources on this subject: Using PVR with Raspbmc [Article] Pulse width modulator [Article] Making the Unit Very Mobile - Controlling Legged Movement [Article]

In this article by Paul Gerrard and Radia M. Johnson, the authors of Mastering Scientific Computation with R, we'll discuss the fundamental ideas underlying structural equation modeling, which are often overlooked in other books discussing structural equation modeling (SEM) in R, and then delve into how SEM is done in R. We will then discuss two R packages, OpenMx and lavaan. We can directly apply our discussion of the linear algebra underlying SEM using OpenMx. Because of this, we will go over OpenMx first. We will then discuss lavaan, which is probably more user friendly because it sweeps the matrices and linear algebra representations under the rug so that they are invisible unless the user really goes looking for them. Both packages continue to be developed and there will always be some features better supported in one of these packages than in the other. (For more resources related to this topic, see here.) SEM model fitting and estimation methods To ultimately find a good solution, software has to use trial and error to come up with an implied covariance matrix that matches the observed covariance matrix as well as possible. The question is what does "as well as possible" mean? The answer to this is that the software must try to minimize some particular criterion, usually some sort of discrepancy function. Just what that criterion is depends on the estimation method used. The most commonly used estimation methods in SEM include: Ordinary least squares (OLS) also called unweighted least squares Generalized least squares (GLS) Maximum likelihood (ML) There are a number of other estimation methods as well, some of which can be done in R, but here we will stick with describing the most common ones. In general, OLS is the simplest and computationally cheapest estimation method. GLS is computationally more demanding, and ML is computationally more intensive. We will see why this is, as we discuss the details of these estimation methods. Any SEM estimation method seeks to estimate model parameters that recreate the observed covariance matrix as well as possible. To evaluate how closely an implied covariance matrix matches an observed covariance matrix, we need a discrepancy function. If we assume multivariate normality of the observed variables, the following function can be used to assess discrepancy: In the preceding figure, R is the observed covariance matrix, C is the implied covariance matrix, and V is a weight matrix. The tr function refers to the trace function, which sums the elements of the main diagonal. The choice of V varies based on the SEM estimation method: For OLS, V = I For GLS, V = R-1 In the case of an ML estimation, we seek to minimize one of a number of similar criteria to describe ML, as follows: In the preceding figure, n is the number of variables. There are a couple of points worth noting here. GLS estimation inverts the observed correlation matrix, something computationally demanding with large matrices, but something that must only be done once. Alternatively, ML requires inversion of the implied covariance matrix, which changes with each iteration. Thus, each iteration requires the computationally demanding step of matrix inversion. With modern fast computers, this difference may not be noticeable, but with large SEM models, this might start to be quite time-consuming. Assessing SEM model fit The final question in an SEM model is how well the model explains the data. This is answered with the use of SEM measures of fit. Most of these measures are based on a chi-squared distribution. The fit criteria for GLS and ML (as well as a number of other estimation procedures such as asymptotic distribution-free methods) multiplied by N-1 is approximately chi-square distributed. Here, the capital N represents the number of observations in the dataset, as opposed to lower case n, which gives the number of variables. We compute degrees of freedom as the difference between the number of estimated parameters and the number of known covariances (that is, the total number of values in one triangle of an observed covariance matrix). This gives way to the first test statistic for SEM models, a chi-squared significance level comparing our chi-square value to some minimum chi-square threshold to achieve statistical significance. As with conventional chi-square testing, a chi-square value that is higher than some minimal threshold will reject the null hypothesis. Most experimental science features such as rejection supports the hypothesis of the experiment. This is not the case in SEM, where the null hypothesis is that the model fits the data. Thus, a non-significant chi-square is an indicator of model fit, whereas a significant chi-square rejects model fit. A notable limitation of this is that a greater sample size, greater N, will increase the chi-square value and will therefore increase the power to reject model fit. Thus, using conventional chi-squared testing will tend to support models developed in small samples and reject models developed in large samples. The choice an interpretation of fit measures is a contentious one in SEM literature. However, as can be seen, chi-square has limitations. As such, other model fit criteria were developed that do not penalize models that fit in large samples (some may penalize models fit to small samples though). There are over a dozen indices, but the most common fit indices and interpretation information are as follows: Comparative fit index: In this index, a higher value is better. Conventionally, a value of greater than 0.9 was considered an indicator of good model fit, but some might argue that a value of at least 0.95 is needed. This is relatively sample size insensitive. Root mean square error of approximation: A value of under 0.08 (smaller is better) is often considered necessary to achieve model fit. However, this fit measure is quite sample size sensitive, penalizing small sample studies. Tucker-Lewis index (Non-normed fit index): This is interpreted in a similar manner as the comparative fit index. Also, this is not very sample size sensitive. Standardized root mean square residual: In this index, a lower value is better. A value of 0.06 or less is considered needed for model fit. Also, this may penalize small samples. In the next section, we will show you how to actually fit SEM models in R and how to evaluate fit using fit measures. Using OpenMx and matrix specification of an SEM We went through the basic principles of SEM and discussed the basic computational approach by which this can be achieved. SEM remains an active area of research (with an entire journal devoted to it, Structural Equation Modeling), so there are many additional peculiarities, but rather than delving into all of them, we will start by delving into actually fitting an SEM model in R. OpenMx is not in the CRAN repository, but it is easily obtainable from the OpenMx website, by typing the following in R: source('http://openmx.psyc.virginia.edu/getOpenMx.R')" Summarizing the OpenMx approach In this example, we will use OpenMx by specifying matrices as mentioned earlier. To fit an OpenMx model, we need to first specify the model and then tell the software to attempt to fit the model. Model specification involves four components: Specifying the model matrices; this has two parts: Declare starting values for the estimation Declaring which values can be estimated and which are fixed Telling OpenMx the algebraic relationship of the matrices that should produce an implied covariance matrix Giving an instruction for the model fitting criterion Providing a source of data The R commands that correspond to each of these steps are: mxMatrix mxAlgebra mxMLObjective mxData We will then pass the objects created with each of these commands to create an SEM model using mxModel. Explaining an entire example First, to make things simple, we will store the FALSE and TRUE logical values in single letter variables, which will be convenient when we have matrices full of TRUE and FALSE values as follows: F <- FALSE T <- TRUE Specifying the model matrices Specifying matrices is done with the mxMatrix function, which returns an MxMatrix object. (Note that the object starts with a capital "M" while the function starts with a lowercase "m.") Specifying an MxMatrix is much like specifying a regular R matrix, but MxMatrices has some additional components. The most notable difference is that there are actually two different matrices used to create an MxMatrix. The first is a matrix of starting values, and the second is a matrix that tells which starting values are free to be estimated and which are not. If a starting value is not freely estimable, then it is a fixed constant. Since the actual starting values that we choose do not really matter too much in this case, we will just pick one as a starting value for all parameters that we would like to be estimated. Let's take a look at the following example: mx.A <- mxMatrix( type = "Full", nrow=14, ncol=14, #Provide the Starting Values values = c(    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0 ), #Tell R which values are free to be estimated    free = c(    F, F, F, F, F, F, F, F, F, F, F, F, F, F,    F, F, F, F, F, F, F, F, F, F, F, F, T, F,    F, F, F, F, F, F, F, F, F, F, F, F, T, F,    F, F, F, F, F, F, F, F, F, F, F, F, T, F,    F, F, F, F, F, F, F, F, F, F, F, F, F, F,    F, F, F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, F, F,    F, F, F, F, F, F, F, F, F, F, F, T, F, F,    F, F, F, F, F, F, F, F, F, F, F, T, F, F,    F, F, F, F, F, F, F, F, F, F, F, F, F, F,    F, F, F, F, F, F, F, F, F, F, F, T, F, F,    F, F, F, F, F, F, F, F, F, F, F, T, T, F ), byrow=TRUE,   #Provide a matrix name that will be used in model fitting name="A", ) We will now apply this same technique to the S matrix. Here, we will create two S matrices, S1 and S2. They differ simply in the starting values that they supply. We will later try to fit an SEM model using one matrix, and then the other to address problems with the first one. The difference is that S1 uses starting variances of 1 in the diagonal, and S2 uses starting variances of 5. Here, we will use the "symm" matrix type, which is a symmetric matrix. We could use the "full" matrix type, but by using "symm", we are saved from typing all of the symmetric values in the upper half of the matrix. Let's take a look at the following matrix: mx.S1 <- mxMatrix("Symm", nrow=14, ncol=14, values = c(    1,    0, 1,    0, 0, 1,    0, 1, 0, 1,    1, 0, 0, 0, 1,    0, 1, 0, 0, 0, 1,    0, 0, 1, 0, 0, 0, 1,    0, 0, 0, 1, 0, 1, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1 ),      free = c(    T,    F, T,    F, F, T,    F, T, F, T,    T, F, F, F, T,    F, T, F, F, F, T,    F, F, T, F, F, F, T,    F, F, F, T, F, T, F, T,    F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, F, T ), byrow=TRUE, name="S" )   #The alternative, S2 matrix: mx.S2 <- mxMatrix("Symm", nrow=14, ncol=14, values = c(    5,    0, 5,    0, 0, 5,    0, 1, 0, 5,    1, 0, 0, 0, 5,    0, 1, 0, 0, 0, 5,    0, 0, 1, 0, 0, 0, 5,    0, 0, 0, 1, 0, 1, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5,    0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5 ),         free = c(    T,    F, T,    F, F, T,    F, T, F, T,    T, F, F, F, T,    F, T, F, F, F, T,    F, F, T, F, F, F, T,    F, F, F, T, F, T, F, T,    F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, T,    F, F, F, F, F, F, F, F, F, F, F, F, F, T ), byrow=TRUE, name="S" ) mx.Filter <- mxMatrix("Full", nrow=11, ncol=14, values= c(        1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,      0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0    ),    free=FALSE,    name="Filter",    byrow = TRUE ) And finally, we will create our identity and filter matrices the same way, as follows: mx.I <- mxMatrix("Full", nrow=14, ncol=14,    values= c(        1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1    ),    free=FALSE,    byrow = TRUE,    name="I" ) Fitting the model Now, it is time to declare the model that we would like to fit using the mxModel command. This part includes steps 2 through step 4 mentioned earlier. Here, we will tell mxModel which matrices to use. We will then use the mxAlgegra command to tell R how the matrices should be combined to reproduce the implied covariance matrix. We will tell R to use ML estimation with the mxMLObjective command, and we will tell it to apply the estimation to a particular matrix algebra, which we named "C". This is simply the right-hand side of the McArdle McDonald equation. Finally, we will tell R where to get the data to use in model fitting using the following code: factorModel.1 <- mxModel("Political Democracy Model", #Model Matrices mx.A, mx.S1, mx.Filter, mx.I, #Model Fitting Instructions mxAlgebra(Filter %*% solve(I-A) %*% S %*% t(solve(I - A)) %*% t(Filter), name="C"),      mxMLObjective("C", dimnames = names(PoliticalDemocracy)),    #Data to fit mxData(cov(PoliticalDemocracy), type="cov", numObs=75) ) Now, let's tell R to fit the model and summarize the results using mxRun, as follows: summary(mxRun(factorModel.1)) Running Political Democracy Model Error in summary(mxRun(factorModel.1)) : error in evaluating the argument 'object' in selecting a method for function 'summary': Error: The job for model 'Political Democracy Model' exited abnormally with the error message: Expected covariance matrix is non-positive-definite. Uh oh! We got an error message telling us that the expected covariance matrix is not positive definite. Our observed covariance matrix is positive definite but the implied covariance matrix (at least at first) is not. This is an effect of the fact that if we multiply our starting value matrices together as specified by the McArdle McDonald equation, we get a starting implied covariance matrix. If we perform an eigenvalue decomposition of this starting implied covariance matrix, then we will find that the last eigenvalue is negative. This means a negative variance does not make much sense, and this is what "not positive definite" refers to. The good news is that this is simply our starting values, so we can fix this if we modify our starting values. In this case, we can choose values of five along the diagonal of the S matrix, and get a positive definite starting implied covariance matrix. We can rerun this using the mx.S2 matrix specified earlier and the software will proceed as follows: #Rerun with a positive definite matrix   factorModel.2 <- mxModel("Political Democracy Model", #Model Matrices mx.A, mx.S2, mx.Filter, mx.I, #Model Fitting Instructions mxAlgebra(Filter %*% solve(I-A) %*% S %*% t(solve(I - A)) %*% t(Filter), name="C"),    mxMLObjective("C", dimnames = names(PoliticalDemocracy)),    #Data to fit mxData(cov(PoliticalDemocracy), type="cov", numObs=75) )   summary(mxRun(factorModel.2)) This should provide a solution. As can be seen from the previous code, the parameters solved in the model are returned as matrix components. Just like we had to figure out how to go from paths to matrices, we now have to figure out how to go from matrices to paths (the reverse problem). In the following screenshot, we show just the first few free parameters: The preceding screenshot tells us that the parameter estimated in the position of the tenth row and twelfth column in the matrix A is 2.18. This corresponds to a path from the twelfth variable in the A matrix ind60, to the 10th variable in the matrix x2. Thus, the path coefficient from ind60 to x2 is 2.18. There are a few other pieces of information here. The first one tells us that the model has not converged but is "Mx status Green." This means that the model was still converging when it stopped running (that is, it did not converge), but an optimal solution was still found and therefore, the results are likely reliable. Model fit information is also provided suggesting a pretty good model fit with CFI of 0.99 and RMSEA of 0.032. This was a fair amount of work, and creating model matrices by hand from path diagrams can be quite tedious. For this reason, SEM fitting programs have generally adopted the ability to fit SEM by declaring paths rather than model matrices. OpenMx has the ability to allow declaration by paths, but applying model matrices has a few advantages. Principally, we get under the hood of SEM fitting. If we step back, we can see that OpenMx actually did very little for us that is specific to SEM. We told OpenMx how we wanted matrices multiplied together and which parameters of the matrix were free to be estimated. Instead of using the RAM specification, we could have passed the matrices of the LISREL or Bentler-Weeks models with the corresponding algebra methods to recreate an implied covariance matrix. This means that if we are trying to come up with our matrix specification, reproduce prior research, or apply a new SEM matrix specification method published in the literature, OpenMx gives us the power to do it. Also, for educators wishing to teach the underlying mathematical ideas of SEM, OpenMx is a very powerful tool. Fitting SEM models using lavaan If we were to describe OpenMx as the SEM equivalent of having a well-stocked pantry and full kitchen to create whatever you want, and you have the time and know how to do it, we might regard lavaan as a large freezer full of prepackaged microwavable dinners. It does not allow quite as much flexibility as OpenMx because it sweeps much of the work that we did by hand in OpenMx under the rug. Lavaan does use an internal matrix representation, but the user never has to see it. It is this sweeping under the rug that makes lavaan generally much easier to use. It is worth adding that the list of prepackaged features that are built into lavaan with minimal additional programming challenge many commercial SEM packages. The lavaan syntax The key to describing lavaan models is the model syntax, as follows: X =~ Y: Y is a manifestation of the latent variable X Y ~ X: Y is regressed on X Y ~~ X: The covariance between Y and X can be estimated Y ~ 1: This estimates the intercept for Y (implicitly requires mean structure) Y | a*t1 + b*t2: Y has two thresholds that is a and b Y ~ a * X: Y is regressed on X with coefficient a Y ~ start(a) * X: Y is regressed on X; the starting value used for estimation is a It may not be evident at first, but this model description language actually makes lavaan quite powerful. Wherever you have seen a or b in the previous examples, a variable or constant can be used in their place. The beauty of this is that multiple parameters can be constrained to be equal simply by assigning a single parameter name to them. Using lavaan, we can fit a factor analysis model to our physical functioning dataset with only a few lines of code: phys.func.data <- read.csv('phys_func.csv')[-1] names(phys.func.data) <- LETTERS[1:20] R has a built-in vector named LETTERS, which contains all of the capital letters of the English alphabet. The lower case vector letters contains the lowercase alphabet. We will then describe our model using the lavaan syntax. Here, we have a model of three latent variables, our factors, and each of them has manifest variables. Let's take a look at the following example: model.definition.1 <- ' #Factors    Cognitive =~ A + Q + R + S    Legs =~ B + C + D + H + I + J + M + N    Arms =~ E + F+ G + K +L + O + P + T    #Correlations Between Factors    Cognitive ~~ Legs    Cognitive ~~ Arms    Legs ~~ Arms ' We then tell lavaan to fit the model as follows: fit.phys.func <- cfa(model.definition.1, data=phys.func.data, ordered= c('A','B', 'C','D', 'E','F','G', 'H','I','J', 'K', 'L','M','N','O','P','Q','R', 'S', 'T')) In the previous code, we add an ordered = argument, which tells lavaan that some variables are ordinal in nature. In response, lavaan estimates polychoric correlations for these variables. Polychoric correlations assume that we binned a continuous variable into discrete categories, and attempts to explicitly model correlations assuming that there is some continuous underlying variable. Part of this requires finding thresholds (placed on an arbitrary scale) between each categorical response. (for example, threshold 1 falls between the response of 1 and 2, and so on). By telling lavaan to treat some variables as categorical, lavaan will also know to use a special estimation method. Lavaan will use diagonally weighted least squares, which does not assume normality and uses the diagonals of the polychoric correlation matrix for weights in the discrepancy function. With five response options, it is questionable as to whether polychoric correlations are truly needed. Some analysts might argue that with many response options, the data can be treated as continuous, but here we use this method to show off lavaan's capabilities. All SEM models in lavaan use the lavaan command. Here, we use the cfa command, which is one of a number of wrapper functions for the lavaan command. Others include sem and growth. These commands differ in the default options passed to the lavaan command. (For full details, see the package documentation.) Summarizing the data, we can see the loadings of each item on the factor as well as the factor intercorrelations. We can also see the thresholds between each category from the polychoric correlations as follows: summary(fit.phys.func) We can also assess things such as model fit using the fitMeasures command, which has most of the popularly used fit measures and even a few obscure ones. Here, we tell lavaan to simply extract three measures of model fit as follows: fitMeasures(fit.phys.func, c('rmsea', 'cfi', 'srmr')) Collectively, these measures suggest adequate model fit. It is worth noting here that the interpretation of fit measures largely comes from studies using maximum likelihood estimation, and there is some debate as to how well these generalize other fitting methods. The lavaan package also has the capability to use other estimators that treat the data as truly continuous in nature. For this, a particular dataset is far from multivariate normal distributed, so an estimator such as ML is appropriate to use. However, if we wanted to do so, the syntax would be as follows: fit.phys.func.ML <- cfa(model.definition.1, data=phys.func.data, estimator = 'ML') Comparing OpenMx to lavaan It can be seen that lavaan has a much simpler syntax that allows to rapidly model basic SEM models. However, we were a bit unfair to OpenMx because we used a path model specification for lavaan and a matrix specification for OpenMx. The truth is that OpenMx is still probably a bit wordier than lavaan, but let's apply a path model specification in each to do a fair head-to-head comparison. We will use the famous Holzinger-Swineford 1939 dataset here from the lavaan package to do our modeling, as follows: hs.dat <- HolzingerSwineford1939 We will create a new dataset with a shorter name so that we don't have to keep typing HozlingerSwineford1939. Explaining an example in lavaan We will learn to fit the Holzinger-Swineford model in this section. We will start by specifying the SEM model using the lavaan model syntax: hs.model.lavaan <- ' visual =~ x1 + x2 + x3 textual =~ x4 + x5 + x6 speed   =~ x7 + x8 + x9   visual ~~ textual visual ~~ speed textual ~~ speed '   fit.hs.lavaan <- cfa(hs.model.lavaan, data=hs.dat, std.lv = TRUE) summary(fit.hs.lavaan) Here, we add the std.lv argument to the fit function, which fixes the variance of the latent variables to 1. We do this instead of constraining the first factor loading on each variable to 1. Only the model coefficients are included for ease of viewing in this book. The result is shown in the following model: > summary(fit.hs.lavaan) …                      Estimate Std.err Z-value P(>|z|) Latent variables: visual =~    x1               0.900   0.081   11.127   0.000    x2               0.498   0.077   6.429   0.000    x3              0.656   0.074   8.817   0.000 textual =~    x4               0.990   0.057   17.474   0.000    x5               1.102   0.063   17.576   0.000    x6               0.917   0.054   17.082   0.000 speed =~    x7               0.619   0.070   8.903   0.000    x8               0.731   0.066   11.090   0.000    x9               0.670   0.065   10.305   0.000   Covariances: visual ~~    textual           0.459   0.064   7.189   0.000    speed             0.471   0.073   6.461   0.000 textual ~~    speed             0.283   0.069   4.117   0.000 Let's compare these results with a model fit in OpenMx using the same dataset and SEM model. Explaining an example in OpenMx The OpenMx syntax for path specification is substantially longer and more explicit. Let's take a look at the following model: hs.model.open.mx <- mxModel("Holzinger Swineford", type="RAM",      manifestVars = names(hs.dat)[7:15], latentVars = c('visual', 'textual', 'speed'),    # Create paths from latent to observed variables mxPath(        from = 'visual',        to = c('x1', 'x2', 'x3'),    free = c(TRUE, TRUE, TRUE),    values = 1          ), mxPath(        from = 'textual',        to = c('x4', 'x5', 'x6'),        free = c(TRUE, TRUE, TRUE),        values = 1      ), mxPath(    from = 'speed',    to = c('x7', 'x8', 'x9'),    free = c(TRUE, TRUE, TRUE),    values = 1      ), # Create covariances among latent variables mxPath(    from = 'visual',    to = 'textual',    arrows=2,    free=TRUE      ), mxPath(        from = 'visual',        to = 'speed',        arrows=2,        free=TRUE      ), mxPath(        from = 'textual',        to = 'speed',        arrows=2,        free=TRUE      ), #Create residual variance terms for the latent variables mxPath(    from= c('visual', 'textual', 'speed'),    arrows=2, #Here we are fixing the latent variances to 1 #These two lines are like st.lv = TRUE in lavaan    free=c(FALSE,FALSE,FALSE),    values=1 ), #Create residual variance terms mxPath( from= c('x1', 'x2', 'x3', 'x4', 'x5', 'x6', 'x7', 'x8', 'x9'),    arrows=2, ),    mxData(        observed=cov(hs.dat[,c(7:15)]),        type="cov",        numObs=301    ) )     fit.hs.open.mx <- mxRun(hs.model.open.mx) summary(fit.hs.open.mx) Here are the results of the OpenMx model fit, which look very similar to lavaan's. This gives a long output. For ease of viewing, only the most relevant parts of the output are included in the following model (the last column that R prints giving the standard error of estimates is also not shown here): > summary(fit.hs.open.mx) …   free parameters:                            name matrix     row     col Estimate Std.Error 1   Holzinger Swineford.A[1,10]     A     x1 visual 0.9011177 2   Holzinger Swineford.A[2,10]     A     x2 visual 0.4987688 3   Holzinger Swineford.A[3,10]     A     x3 visual 0.6572487 4   Holzinger Swineford.A[4,11]     A     x4 textual 0.9913408 5   Holzinger Swineford.A[5,11]     A     x5 textual 1.1034381 6   Holzinger Swineford.A[6,11]     A     x6 textual 0.9181265 7   Holzinger Swineford.A[7,12]     A     x7   speed 0.6205055 8   Holzinger Swineford.A[8,12]     A     x8 speed 0.7321655 9   Holzinger Swineford.A[9,12]     A     x9   speed 0.6710954 10   Holzinger Swineford.S[1,1]     S     x1     x1 0.5508846 11   Holzinger Swineford.S[2,2]     S     x2     x2 1.1376195 12   Holzinger Swineford.S[3,3]     S    x3     x3 0.8471385 13   Holzinger Swineford.S[4,4]     S     x4     x4 0.3724102 14   Holzinger Swineford.S[5,5]     S     x5     x5 0.4477426 15   Holzinger Swineford.S[6,6]     S     x6     x6 0.3573899 16   Holzinger Swineford.S[7,7]      S     x7     x7 0.8020562 17   Holzinger Swineford.S[8,8]     S     x8     x8 0.4893230 18   Holzinger Swineford.S[9,9]     S     x9     x9 0.5680182 19 Holzinger Swineford.S[10,11]     S visual textual 0.4585093 20 Holzinger Swineford.S[10,12]     S visual   speed 0.4705348 21 Holzinger Swineford.S[11,12]     S textual   speed 0.2829848 In summary, the results agree quite closely. For example, looking at the coefficient for the path going from the latent variable visual to the observed variable x1, lavaan gives an estimate of 0.900 while OpenMx computes a value of 0.901. Summary The lavaan package is user friendly, pretty powerful, and constantly adding new features. Alternatively, OpenMx has a steeper learning curve but tremendous flexibility in what it can do. Thus, lavaan is a bit like a large freezer full of prepackaged microwavable dinners, whereas OpenMx is like a well-stocked pantry with no prepared foods but a full kitchen that will let you prepare it if you have the time and the know-how. To run a quick analysis, it is tough to beat the simplicity of lavaan, especially given its wide range of capabilities. For large complex models, OpenMx may be a better choice. The methods covered here are useful to analyze statistical relationships when one has all of the data from events that have already occurred. Resources for Article: Further resources on this subject: Creating your first heat map in R [article] Going Viral [article] Introduction to S4 Classes [article]

In this article by Marco Schwartz and Oliver Manickum authors of the book Programming Arduino with LabVIEW, we will see how to design an interfave using LabVIEW. (For more resources related to this topic, see here.) At this stage, we know that we have our two sensors working and that they were interfaced correctly with the LabVIEW interface. However, we can do better; for now, we simply have a text display of the measurements, which is not elegant to read. Also, the light-level measurement goes from 0 to 5, which doesn't mean anything for somebody who will look at the interface for the first time. Therefore, we will modify the interface slightly. We will add a temperature gauge to display the data coming from the temperature sensor, and we will modify the output of the reading from the photocell to display the measurement from 0 (no light) to 100 percent (maximum brightness). We first need to place the different display elements. To do this, perform the following steps: Start with Front Panel. You can use a temperature gauge for the temperature and a simple slider indicator for Light Level. You will find both in the Indicators submenu of LabVIEW. After that, simply place them on the right-hand side of the interface and delete the other indicators we used earlier. Also, name the new indicators accordingly so that we can know to which element we have to connect them later. Then, it is time to go back to Block Diagram to connect the new elements we just added in Front Panel. For the temperature element, it is easy: you can simply connect the temperature gauge to the TMP36 output pin. For the light level, we will make slightly more complicated changes. We will divide the measured value beside the Analog Read element by 5, thus obtaining an output value between 0 and 1. Then, we will multiply this value by 100, to end up with a value going from 0 to 100 percent of the ambient light level. To do so perform the following steps: The first step is to place two elements corresponding to the two mathematical operations we want to do: a divide operator and a multiply operator. You can find both of them in the Functions panel of LabVIEW. Simply place them close to the Analog Read element in your program. After that, right-click on one of the inputs of each operator element, and go to Create | Constant to create a constant input for each block. Add a value of 5 for the division block, and add a value of 100 for the multiply block. Finally, connect the output of the Analog Read element to the input of the division block, the output of this block to the input of the multiply block, and the output of the multiply block to the input of the Light Level indicator. You can now go back to Front Panel to see the new interface in action. You can run the program again by clicking on the little arrow on the toolbar. You should immediately see that Temperature is now indicated by the gauge on the right and Light Level is immediately changing on the slider, depending on how you cover the sensor with your hand. Summary In this article, we connected a temperature sensor and a light-level sensor to Arduino and built a simple LabVIEW program to read data from these sensors. Then, we built a nice graphical interface to visualize the data coming from these sensors. There are many ways you can build other projects based on what you learned in this article. You can, for example, connect higher temperatures and/or more light-level sensors to the Arduino board and display these measurements in the interface. You can also connect other kinds of sensors that are supported by LabVIEW, for example, other analog sensors. For example, you can add a barometric pressure sensor or a humidity sensor to the project to build an even more complete weather-measurement station. One other interesting extension of this article will be to use the storage and plotting capabilities of LabVIEW to dynamically plot the history of the measured data inside the LabVIEW interface. Resources for Article: Further resources on this subject: The Arduino Mobile Robot [article] Using the Leap Motion Controller with Arduino [article] Avoiding Obstacles Using Sensors [article]

In this article by Alan Thorn author of the book Mastering Unity Scripting will cover the following topics: Events Event management (For more resources related to this topic, see here.) The Update events for MonoBehaviour objects seem to offer a convenient place for executing code that should perform regularly over time, spanning multiple frames, and possibly multiple scenes. When creating sustained behaviors over time, such as artificial intelligence for enemies or continuous motion, it may seem that there are almost no alternatives to filling an Update function with many if and switch statements, branching your code in different directions depending on what your objects need to do at the current time. But, when the Update events are seen this way, as a default place to implement prolonged behaviors, it can lead to severe performance problems for larger and more complex games. On deeper analysis, it's not difficult to see why this would be the case. Typically, games are full of so many behaviors, and there are so many things happening at once in any one scene that implementing them all through the Update functions is simply unfeasible. Consider the enemy characters alone, they need to know when the player enters and leaves their line of sight, when their health is low, when their ammo has expired, when they're standing on harmful terrain, when they're taking damage, when they're moving or not, and lots more. On thinking initially about this range of behaviors, it seems that all of them require constant and continuous attention because enemies should always know, instantly, when changes in these properties occur as a result of the player input. That is, perhaps, the main reason why the Update function seems to be the most suitable place in these situations but there are better alternatives, namely, event-driven programming. By seeing your game and your application in terms of events, you can make considerable savings in performance. This article then considers the issue of events and how to manage them game wide. Events Game worlds are fully deterministic systems; in Unity, the scene represents a shared 3D Cartesian space and timeline inside which finite GameObjects exist. Things only happen within this space when the game logic and code permits them to. For example, objects can only move when there is code somewhere that tells them to do so, and under specific conditions, such as when the player presses specific buttons on the keyboard. Notice from the example that behaviors are not simply random but are interconnected; objects move only when keyboard events occur. There is an important connection established between the actions, where one action entails another. These connections or linkages are referred to as events; each unique connection being a single event. Events are not active but passive; they represent moments of opportunity but not action in themselves, such as a key press, a mouse click, an object entering a collider volume, the player being attacked, and so on. These are examples of events and none of them say what the program should actually do, but only the kind of scenario that just happened. Event-driven programming starts with the recognition of events as a general concept and comes to see almost every circumstance in a game as an instantiation of an event; that is, as an event situated in time, not just an event concept but as a specific event that happens at a specific time. Understanding game events like these is helpful because all actions in a game can then be seen as direct responses to events as and when they happen. Specifically, events are connected to responses; an event happens and triggers a response. Further, the response can go on to become an event that triggers further responses and so on. In other words, the game world is a complete, integrated system of events and responses. Once the world is seen this way, the question then arises as to how it can help us improve performance over simply relying on the Update functions to move behaviors forward on every frame. And the method is simply by finding ways to reduce the frequency of events. Now, stated in this way, it may sound a crude strategy, but it's important. To illustrate, let's consider the example of an enemy character firing a weapon at the player during combat. Throughout the gameplay, the enemy will need to keep track of many properties. Firstly, their health, because when it runs low the enemy should seek out medical kits and aids to restore their health again. Secondly, their ammo, because when it runs low the enemy should seek to collect more and also the enemy will need to make reasoned judgments about when to fire at the player, such as only when they have a clear line of sight. Now, by simply thinking about this scenario, we've already identified some connections between actions that might be identified as events. But before taking this consideration further, let's see how we might implement this behavior using an Update function, as shown in the following code sample 4-1. Then, we'll look at how events can help us improve on that implementation: // Update is called once per frame void Update () {    //Check enemy health    //Are we dead?    if(Health <= 0)    {          //Then perform die behaviour          Die();          return;    }    //Check for health low    if(health <= 20)    {        //Health is low, so find first-aid          RunAndFindHealthRestore();          return;    }    //Check ammo    //Have we run out of ammo?    if(Ammo <= 0)    {          //Then find more          SearchMore();          return;    }    //Health and ammo are fine. Can we see player? If so, shoot    if(HaveLineOfSight)    {            FireAtPlayer();    } } The preceding code sample 4-1 shows a heavy Update function filled with lots of condition checking and responses. In essence, the Update function attempts to merge event handling and response behaviors into one and the results in an unnecessarily expensive process. If we think about the event connections between these different processes (the health and ammo check), we see how the code could be refactored more neatly. For example, ammo only changes on two occasions: when a weapon is fired or when new ammo is collected. Similarly, health only changes on two occasions: when an enemy is successfully attacked by the player or when an enemy collects a first-aid kit. In the first case, there is a reduction, and in the latter case, an increase. Since these are the only times when the properties change (the events), these are the only points where their values need to be validated. See the following code sample 4-2 for a refactored enemy, which includes C# properties and a much reduced Update function: using UnityEngine; using System.Collections; public class EnemyObject : MonoBehaviour {    //-------------------------------------------------------    //C# accessors for private variables    public int Health    {          get{return _health;}          set          {                //Clamp health between 0-100                _health = Mathf.Clamp(value, 0, 100);               //Check if dead                if(_health <= 0)                {                      OnDead();                      return;                }                //Check health and raise event if required                if(_health <= 20)               {                      OnHealthLow();                      return;                }          }    }    //-------------------------------------------------------    public int Ammo    {          get{return _ammo;}          set          {              //Clamp ammo between 0-50              _ammo = Mathf.Clamp(value,0,50);                //Check if ammo empty                if(_ammo <= 0)                {                      //Call expired event                      OnAmmoExpired();                      return;                }          }    }    //-------------------------------------------------------    //Internal variables for health and ammo    private int _health = 100;    private int _ammo = 50;    //-------------------------------------------------------    // Update is called once per frame    void Update ()    {    }    //-------------------------------------------------------    //This event is called when health is low    void OnHealthLow()    {          //Handle event response here    }    //-------------------------------------------------------    //This event is called when enemy is dead    void OnDead()    {        //Handle event response here    }    //-------------------------------------------------------    //Ammo run out event    void OnAmmoExpired()    {        //Handle event response here    }    //------------------------------------------------------- } The enemy class in the code sample 4-2 has been refactored to an event-driven design, where properties such as Ammo and Health are validated not inside the Update function but on assignment. From here, events are raised wherever appropriate based on the newly assigned values. By adopting an event-driven design, we introduce performance optimization and cleanness into our code; we reduce the excess baggage and value checks as found with the Update function in the code sample 4-1, and instead we only allow value-specific events to drive our code, knowing they'll be invoked only at the relevant times. Event management Event-driven programming can make our lives a lot easier. But no sooner than we accept events into the design do we come across a string of new problems that require a thoroughgoing resolution. Specifically, we saw in the code sample 4-2 how C# properties for health and ammo are used to validate and detect for relevant changes and then to raise events (such as OnDead) where appropriate. This works fine in principle, at least when the enemy must be notified about events that happen to itself. However, what if an enemy needed to know about the death of another enemy or needed to know when a specified number of other enemies had been killed? Now, of course, thinking about this specific case, we could go back to the enemy class in the code sample 4-2 and amend it to call an OnDead event not just for the current instance but for all other enemies using functions such as SendMessage. But this doesn't really solve our problem in the general sense. In fact, let's state the ideal case straight away; we want every object to optionally listen for every type of event and to be notified about them as and when they happen, just as easily as if the event had happened to them. So the question that we face now is about how to code an optimized system to allow easy event management like this. In short, we need an EventManager class that allows objects to listen to specific events. This system relies on three central concepts, as follows: Event Listener: A listener refers to any object that wants to be notified about an event when it happens, even its own events. In practice, almost every object will be a listener for at least one event. An enemy, for example, may want notifications about low health and low ammo among others. In this case, it's a listener for at least two separate events. Thus, whenever an object expects to be told when an event happens, it becomes a listener. Event Poster: In contrast to listeners, when an object detects that an event has occurred, it must announce or post a public notification about it that allows all other listeners to be notified. In the code sample 4-2, the enemy class detects the Ammo and Health events using properties and then calls the internal events, if required. But to be a true poster in this sense, we require that the object must raise events at a global level. Event Manager: Finally, there's an overarching singleton Event Manager object that persists across levels and is globally accessible. This object effectively links listeners to posters. It accepts notifications of events sent by posters and then immediately dispatches the notifications to all appropriate listeners in the form of events. Starting event management with interfaces The first or original entity in the event handling system is the listener—the thing that should be notified about specific events as and when they happen. Potentially, a listener could be any kind of object or any kind of class; it simply expects to be notified about specific events. In short, the listener will need to register itself with the Event Manager as a listener for one or more specific events. Then, when the event actually occurs, the listener should be notified directly by a function call. So, technically, the listener raises a type-specificity issue for the Event Manager about how the manager should invoke an event on the listener if the listener could potentially be an object of any type. Of course, this issue can be worked around, as we've seen, using either SendMessage or BroadcastMessage. Indeed, there are event handling systems freely available online, such as NotificationCenter that rely on these functions. However, we'll avoid them using interfaces and use polymorphism instead, as both SendMessage and BroadcastMessage rely heavily on reflection. Specifically, we'll create an interface from which all listener objects derive. More information on the freely available NotificationCenter (C# version) is available from the Unity wiki at http://wiki.unity3d.com/index.php?title=CSharpNotificationCenter. In C#, an interface is like a hollow abstract base class. Like a class, an interface brings together a collection of methods and functions into a single template-like unit. But, unlike a class, an interface only allows you to define function prototypes such as the name, return type, and arguments for a function. It doesn't let you define a function body. The reason being that an interface simply defines the total set of functions that a derived class will have. The derived class may implement the functions however necessary, and the interface simply exists so that other objects can invoke the functions via polymorphism without knowing the specific type of each derived class. This makes interfaces a suitable candidate to create a Listener object. By defining a Listener interface from which all objects will be derived, every object has the ability to be a listener for events. The following code sample 4-3 demonstrates a sample Listener interface: 01 using UnityEngine; 02 using System.Collections; 03 //----------------------------------------------------------- 04 //Enum defining all possible game events 05 //More events should be added to the list 06 public enum EVENT_TYPE {GAME_INIT, 07                                GAME_END, 08                                 AMMO_EMPTY, 09                                 HEALTH_CHANGE, 10                                 DEAD}; 11 //----------------------------------------------------------- 12 //Listener interface to be implemented on Listener classes 13 public interface IListener 14 { 15 //Notification function invoked when events happen 16 void OnEvent(EVENT_TYPE Event_Type, Component Sender,    Object Param = null); 17 } 18 //----------------------------------------------------------- The following are the comments for the code sample 4-3: Lines 06-10: This enumeration should define a complete list of all possible game events that could be raised. The sample code lists only five game events: GAME_INIT, GAME_END, AMMO_EMPTY, HEALTH_CHANGE, and DEAD. Your game will presumably have many more. You don't actually need to use enumerations for encoding events; you could just use integers. But I've used enumerations to improve event readability in code. Lines 13-17: The Listener interface is defined as IListener using the C# interfaces. It supports just one event, namely OnEvent. This function will be inherited by all derived classes and will be invoked by the manager whenever an event occurs for which the listener is registered. Notice that OnEvent is simply a function prototype; it has no body. More information on C# interfaces can be found at http://msdn.microsoft.com/en-us/library/ms173156.aspx. Using the IListener interface, we now have the ability to make a listener from any object using only class inheritance; that is, any object can now declare itself as a listener and potentially receive events. For example, a new MonoBehaviour component can be turned into a listener with the following code sample 4-4. This code uses multiple inheritance, that is, it inherits from two classes. More information on multiple inheritance can be found at http://www.dotnetfunda.com/articles/show/1185/multiple-inheritance-in-csharp: using UnityEngine; using System.Collections; public class MyCustomListener : MonoBehaviour, IListener {    // Use this for initialization    void Start () {}    // Update is called once per frame    void Update () {}    //---------------------------------------    //Implement OnEvent function to receive Events    public void OnEvent(EVENT_TYPE Event_Type, Component Sender, Object Param = null)    {    }    //--------------------------------------- } Creating an EventManager Any object can now be turned into a listener, as we've seen. But still the listeners must register themselves with a manager object of some kind. Thus, it is the duty of the manager to call the events on the listeners when the events actually happen. Let's now turn to the manager itself and its implementation details. The manager class will be called EventManager, as shown in the following code sample 4-5. This class, being a persistent singleton object, should be attached to an empty GameObject in the scene where it will be directly accessible to every other object through a static instance property. More on this class and its usage is considered in the subsequent comments: 001 using UnityEngine; 002 using System.Collections; 003 using System.Collections.Generic; 004 //----------------------------------- 005 //Singleton EventManager to send events to listeners 006 //Works with IListener implementations 007 public class EventManager : MonoBehaviour 008 { 009     #region C# properties 010 //----------------------------------- 011     //Public access to instance 012     public static EventManager Instance 013       { 014             get{return instance;} 015            set{} 016       } 017   #endregion 018 019   #region variables 020       // Notifications Manager instance (singleton design pattern) 021   private static EventManager instance = null; 022 023     //Array of listeners (all objects registered for events) 024     private Dictionary<EVENT_TYPE, List<IListener>> Listeners          = new Dictionary<EVENT_TYPE, List<IListener>>(); 025     #endregion 026 //----------------------------------------------------------- 027     #region methods 028     //Called at start-up to initialize 029     void Awake() 030     { 031             //If no instance exists, then assign this instance 032             if(instance == null) 033           { 034                   instance = this; 035                   DontDestroyOnLoad(gameObject); 036           } 037             else 038                   DestroyImmediate(this); 039     } 040//----------------------------------------------------------- 041     /// <summary> 042     /// Function to add listener to array of listeners 043     /// </summary> 044     /// <param name="Event_Type">Event to Listen for</param> 045     /// <param name="Listener">Object to listen for event</param> 046     public void AddListener(EVENT_TYPE Event_Type, IListener        Listener) 047    { 048           //List of listeners for this event 049           List<IListener> ListenList = null; 050 051           // Check existing event type key. If exists, add to list 052           if(Listeners.TryGetValue(Event_Type,                out ListenList)) 053           { 054                   //List exists, so add new item 055                   ListenList.Add(Listener); 056                   return; 057           } 058 059           //Otherwise create new list as dictionary key 060           ListenList = new List<IListener>(); 061           ListenList.Add(Listener); 062           Listeners.Add(Event_Type, ListenList); 063     } 064 //----------------------------------------------------------- 065       /// <summary> 066       /// Function to post event to listeners 067       /// </summary> 068       /// <param name="Event_Type">Event to invoke</param> 069       /// <param name="Sender">Object invoking event</param> 070       /// <param name="Param">Optional argument</param> 071       public void PostNotification(EVENT_TYPE Event_Type,          Component Sender, Object Param = null) 072       { 073           //Notify all listeners of an event 074 075           //List of listeners for this event only 076           List<IListener> ListenList = null; 077 078           //If no event exists, then exit 079           if(!Listeners.TryGetValue(Event_Type,                out ListenList)) 080                   return; 081 082             //Entry exists. Now notify appropriate listeners 083             for(int i=0; i<ListenList.Count; i++) 084             { 085                   if(!ListenList[i].Equals(null)) 086                   ListenList[i].OnEvent(Event_Type, Sender, Param); 087             } 088     } 089 //----------------------------------------------------------- 090     //Remove event from dictionary, including all listeners 091     public void RemoveEvent(EVENT_TYPE Event_Type) 092     { 093           //Remove entry from dictionary 094           Listeners.Remove(Event_Type); 095     } 096 //----------------------------------------------------------- 097       //Remove all redundant entries from the Dictionary 098     public void RemoveRedundancies() 099     { 100             //Create new dictionary 101             Dictionary<EVENT_TYPE, List<IListener>>                TmpListeners = new Dictionary                <EVENT_TYPE, List<IListener>>(); 102 103             //Cycle through all dictionary entries 104             foreach(KeyValuePair<EVENT_TYPE, List<IListener>>                Item in Listeners) 105             { 106                   //Cycle all listeners, remove null objects 107                   for(int i = Item.Value.Count-1; i>=0; i--) 108                   { 109                         //If null, then remove item 110                         if(Item.Value[i].Equals(null)) 111                                 Item.Value.RemoveAt(i); 112                   } 113 114           //If items remain in list, then add to tmp dictionary 115                   if(Item.Value.Count > 0) 116                         TmpListeners.Add (Item.Key,                              Item.Value); 117             } 118 119             //Replace listeners object with new dictionary 120             Listeners = TmpListeners; 121     } 122 //----------------------------------------------------------- 123       //Called on scene change. Clean up dictionary 124       void OnLevelWasLoaded() 125       { 126           RemoveRedundancies(); 127       } 128 //----------------------------------------------------------- 129     #endregion 130 } More information on the OnLevelWasLoaded event can be found at http://docs.unity3d.com/ScriptReference/MonoBehaviour.OnLevelWasLoaded.html. The following are the comments for the code sample 4-5: Line 003: Notice the addition of the System.Collections.Generic namespace giving us access to additional mono classes, including the Dictionary class. This class will be used throughout the EventManager class. In short, the Dictionary class is a special kind of 2D array that allows us to store a database of values based on key-value pairing. More information on the Dictionary class can be found at http://msdn.microsoft.com/en-us/library/xfhwa508%28v=vs.110%29.aspx. Line 007: The EventManager class is derived from MonoBehaviour and should be attached to an empty GameObject in the scene where it will exist as a persistent singleton. Line 024: A private member variable Listeners is declared using a Dictionary class. This structure maintains a hash-table array of key-value pairs, which can be looked up and searched like a database. The key-value pairing for the EventManager class takes the form of EVENT_TYPE and List<Component>. In short, this means that a list of event types can be stored (such as HEALTH_CHANGE), and for each type there could be none, one, or more components that are listening and which should be notified when the event occurs. In effect, the Listeners member is the primary data structure on which the EventManager relies to maintain who is listening for what. Lines 029-039: The Awake function is responsible for the singleton functionality, that is, to make the EventManager class into a singleton object that persists across scenes. Lines 046-063: The AddListener method of EventManager should be called by a Listener object once for each event for which it should listen. The method accepts two arguments: the event to listen for (Event_Type) and a reference to the listener object itself (derived from IListener), which should be notified if and when the event happens. The AddListener function is responsible for accessing the Listeners dictionary and generating a new key-value pair to store the connection between the event and the listener. Lines 071-088: The PostNotification function can be called by any object, whether a listener or not, whenever an event is detected. When called, the EventManager cycles all matching entries in the dictionary, searching for all listeners connected to the current event, and notifies them by invoking the OnEvent method through the IListener interface. Lines 098-127: The final methods for the EventManager class are responsible for maintaining data integrity of the Listeners structure when a scene change occurs and the EventManager class persists. Although the EventManager class persists across scenes, the listener objects themselves in the Listeners variable may not do so. They may get destroyed on scene changes. If so, scene changes will invalidate some listeners, leaving the EventManager with invalid entries. Thus, the RemoveRedundancies method is called to find and eliminate all invalid entries. The OnLevelWasLoaded event is invoked automatically by Unity whenever a scene change occurs. More information on the OnLevelWasLoaded event can be found online at: http://docs.unity3d.com/ScriptReference/MonoBehaviour.OnLevelWasLoaded.html. #region and #endregion The two preprocessor directives #region and #endregion (in combination with the code folding feature) can be highly useful for improving the readability of your code and also for improving the speed with which you can navigate the source file. They add organization and structure to your source code without affecting its validity or execution. Effectively, #region marks the top of a code block and #endregion marks the end. Once a region is marked, it becomes foldable, that is, it becomes collapsible using the MonoDevelop code editor, provided the code folding feature is enabled. Collapsing a region of code is useful for hiding it from view, which allows you to concentrate on reading other areas relevant to your needs, as shown in the following screenshot: Enabling code folding in MonoDevelop To enable code folding in MonoDevelop, select Options in Tools from the application menu. This displays the Options window. From here, choose the General tab in the Text Editor option and click on Enable code folding as well as Fold #regions by default. Using EventManager Now, let's see how to put the EventManager class to work in a practical context from the perspective of listeners and posters in a single scene. First, to listen for an event (any event) a listener must register itself with the EventManager singleton instance. Typically, this will happen once and at the earliest opportunity, such as the Start function. Do not use the Awake function; this is reserved for an object's internal initialization as opposed to the functionality that reaches out beyond the current object to the states and setup of others. See the following code sample 4-6 and notice that it relies on the Instance static property to retrieve a reference to the active EventManager singleton: //Called at start-up void Start() { //Add myself as listener for health change events EventManager.Instance.AddListener(EVENT_TYPE.HEALTH_CHANGE, this); } Having registered listeners for one or more events, objects can then post notifications to EventManager as events are detected, as shown in the following code sample 4-7: public int Health { get{return _health;} set {    //Clamp health between 0-100    _health = Mathf.Clamp(value, 0, 100);    //Post notification - health has been changed   EventManager.Instance. PostNotification(EVENT_TYPE.HEALTH_CHANGE, this, _health); } } Finally, after a notification is posted for an event, all the associated listeners are updated automatically through EventManager. Specifically, EventManager will call the OnEvent function of each listener, giving listeners the opportunity to parse event data and respond where needed, as shown in the following code sample 4-7: //Called when events happen public void OnEvent(EVENT_TYPE Event_Type, Component Sender, object Param = null) { //Detect event type switch(Event_Type) {    case EVENT_TYPE.HEALTH_CHANGE:          OnHealthChange(Sender, (int)Param);    break; } } Summary This article focused on the manifold benefits available for your applications by adopting an event-driven framework consistently through the EventManager class. In implementing such a manager, we were able to rely on either interfaces or delegates, and either method is powerful and extensible. Specifically, we saw how it's easy to add more and more functionality into an Update function but how doing this can lead to severe performance issues. Better is to analyze the connections between your functionality to refactor it into an event-driven framework. Essentially, events are the raw material of event-driven systems. They represent a necessary connection between one action (the cause) and another (the response). To manage events, we created the EventManager class—an integrated class or system that links posters to listeners. It receives notifications from posters about events as and when they happen and then immediately dispatches a function call to all listeners for the event. Resources for Article: Further resources on this subject: Customizing skin with GUISkin [Article] 2D Twin-stick Shooter [Article] Components in Unity [Article]

In article by Alberto Paro, the author of ElasticSearch Cookbook Second Edition, we will cover about the following recipes: (For more resources related to this topic, see here.) Installing additional script plugins Managing scripts Sorting data using scripts Computing return fields with scripting Filtering a search via scripting Introduction ElasticSearch has a powerful way of extending its capabilities with custom scripts, which can be written in several programming languages. The most common ones are Groovy, MVEL, JavaScript, and Python. In this article, we will see how it's possible to create custom scoring algorithms, special processed return fields, custom sorting, and complex update operations on records. The scripting concept of ElasticSearch can be seen as an advanced stored procedures system in the NoSQL world; so, for an advanced usage of ElasticSearch, it is very important to master it. Installing additional script plugins ElasticSearch provides native scripting (a Java code compiled in JAR) and Groovy, but a lot of interesting languages are also available, such as JavaScript and Python. In older ElasticSearch releases, prior to version 1.4, the official scripting language was MVEL, but due to the fact that it was not well-maintained by MVEL developers, in addition to the impossibility to sandbox it and prevent security issues, MVEL was replaced with Groovy. Groovy scripting is now provided by default in ElasticSearch. The other scripting languages can be installed as plugins. Getting ready You will need a working ElasticSearch cluster. How to do it... In order to install JavaScript language support for ElasticSearch (1.3.x), perform the following steps: From the command line, simply enter the following command: bin/plugin --install elasticsearch/elasticsearch-lang-javascript/2.3.0 This will print the following result: -> Installing elasticsearch/elasticsearch-lang-javascript/2.3.0... Trying http://download.elasticsearch.org/elasticsearch/elasticsearch-lang-javascript/ elasticsearch-lang-javascript-2.3.0.zip... Downloading ....DONE Installed lang-javascript If the installation is successful, the output will end with Installed; otherwise, an error is returned. To install Python language support for ElasticSearch, just enter the following command: bin/plugin -install elasticsearch/elasticsearch-lang-python/2.3.0 The version number depends on the ElasticSearch version. Take a look at the plugin's web page to choose the correct version. How it works... Language plugins allow you to extend the number of supported languages to be used in scripting. During the ElasticSearch startup, an internal ElasticSearch service called PluginService loads all the installed language plugins. In order to install or upgrade a plugin, you need to restart the node. The ElasticSearch community provides common scripting languages (a list of the supported scripting languages is available on the ElasticSearch site plugin page at http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-plugins.html), and others are available in GitHub repositories (a simple search on GitHub allows you to find them). The following are the most commonly used languages for scripting: Groovy (http://groovy.codehaus.org/): This language is embedded in ElasticSearch by default. It is a simple language that provides scripting functionalities. This is one of the fastest available language extensions. Groovy is a dynamic, object-oriented programming language with features similar to those of Python, Ruby, Perl, and Smalltalk. It also provides support to write a functional code. JavaScript (https://github.com/elasticsearch/elasticsearch-lang-javascript): This is available as an external plugin. The JavaScript implementation is based on Java Rhino (https://developer.mozilla.org/en-US/docs/Rhino) and is really fast. Python (https://github.com/elasticsearch/elasticsearch-lang-python): This is available as an external plugin, based on Jython (http://jython.org). It allows Python to be used as a script engine. Considering several benchmark results, it's slower than other languages. There's more... Groovy is preferred if the script is not too complex; otherwise, a native plugin provides a better environment to implement complex logic and data management. The performance of every language is different; the fastest one is the native Java. In the case of dynamic scripting languages, Groovy is faster, as compared to JavaScript and Python. In order to access document properties in Groovy scripts, the same approach will work as in other scripting languages: doc.score: This stores the document's score. doc['field_name'].value: This extracts the value of the field_name field from the document. If the value is an array or if you want to extract the value as an array, you can use doc['field_name'].values. doc['field_name'].empty: This returns true if the field_name field has no value in the document. doc['field_name'].multivalue: This returns true if the field_name field contains multiple values. If the field contains a geopoint value, additional methods are available, as follows: doc['field_name'].lat: This returns the latitude of a geopoint. If you need the value as an array, you can use the doc['field_name'].lats method. doc['field_name'].lon: This returns the longitude of a geopoint. If you need the value as an array, you can use the doc['field_name'].lons method. doc['field_name'].distance(lat,lon): This returns the plane distance, in miles, from a latitude/longitude point. If you need to calculate the distance in kilometers, you should use the doc['field_name'].distanceInKm(lat,lon) method. doc['field_name'].arcDistance(lat,lon): This returns the arc distance, in miles, from a latitude/longitude point. If you need to calculate the distance in kilometers, you should use the doc['field_name'].arcDistanceInKm(lat,lon) method. doc['field_name'].geohashDistance(geohash): This returns the distance, in miles, from a geohash value. If you need to calculate the same distance in kilometers, you should use doc['field_name'] and the geohashDistanceInKm(lat,lon) method. By using these helper methods, it is possible to create advanced scripts in order to boost a document by a distance that can be very handy in developing geolocalized centered applications. Managing scripts Depending on your scripting usage, there are several ways to customize ElasticSearch to use your script extensions. In this recipe, we will see how to provide scripts to ElasticSearch via files, indexes, or inline. Getting ready You will need a working ElasticSearch cluster populated with the populate script (chapter_06/populate_aggregations.sh), available at https://github.com/aparo/ elasticsearch-cookbook-second-edition. How to do it... To manage scripting, perform the following steps: Dynamic scripting is disabled by default for security reasons; we need to activate it in order to use dynamic scripting languages such as JavaScript or Python. To do this, we need to turn off the disable flag (script.disable_dynamic: false) in the ElasticSearch configuration file (config/elasticseach.yml) and restart the cluster. To increase security, ElasticSearch does not allow you to specify scripts for non-sandbox languages. Scripts can be placed in the scripts directory inside the configuration directory. To provide a script in a file, we'll put a my_script.groovy script in the config/scripts location with the following code content: doc["price"].value * factor If the dynamic script is enabled (as done in the first step), ElasticSearch allows you to store the scripts in a special index, .scripts. To put my_script in the index, execute the following command in the command terminal: curl -XPOST localhost:9200/_scripts/groovy/my_script -d '{ "script":"doc["price"].value * factor" }' The script can be used by simply referencing it in the script_id field; use the following command: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "match_all": {} }, "sort": {    "_script" : {      "script_id" : "my_script",      "lang" : "groovy",      "type" : "number",      "ignore_unmapped" : true,      "params" : {        "factor" : 1.1      },      "order" : "asc"    } } }' How it works... ElasticSearch allows you to load your script in different ways; each one of these methods has their pros and cons. The most secure way to load or import scripts is to provide them as files in the config/scripts directory. This directory is continuously scanned for new files (by default, every 60 seconds). The scripting language is automatically detected by the file extension, and the script name depends on the filename. If the file is put in subdirectories, the directory path becomes part of the filename; for example, if it is config/scripts/mysub1/mysub2/my_script.groovy, the script name will be mysub1_mysub2_my_script. If the script is provided via a filesystem, it can be referenced in the code via the "script": "script_name" parameter. Scripts can also be available in the special .script index. These are the REST end points: To retrieve a script, use the following code: GET http://<server>/_scripts/<language>/<id"> To store a script use the following code: PUT http://<server>/_scripts/<language>/<id> To delete a script use the following code: DELETE http://<server>/_scripts/<language>/<id> The indexed script can be referenced in the code via the "script_id": "id_of_the_script" parameter. The recipes that follow will use inline scripting because it's easier to use it during the development and testing phases. Generally, a good practice is to develop using the inline dynamic scripting in a request, because it's faster to prototype. Once the script is ready and no changes are needed, it can be stored in the index since it is simpler to call and manage. In production, a best practice is to disable dynamic scripting and store the script on the disk (generally, dumping the indexed script to disk). See also The scripting page on the ElasticSearch website at http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-scripting.html Sorting data using script ElasticSearch provides scripting support for the sorting functionality. In real world applications, there is often a need to modify the default sort by the match score using an algorithm that depends on the context and some external variables. Some common scenarios are given as follows: Sorting places near a point Sorting by most-read articles Sorting items by custom user logic Sorting items by revenue Getting ready You will need a working ElasticSearch cluster and an index populated with the script, which is available at https://github.com/aparo/ elasticsearch-cookbook-second-edition. How to do it... In order to sort using scripting, perform the following steps: If you want to order your documents by the price field multiplied by a factor parameter (that is, sales tax), the search will be as shown in the following code: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "match_all": {} }, "sort": {    "_script" : {      "script" : "doc["price"].value * factor",      "lang" : "groovy",      "type" : "number",      "ignore_unmapped" : true,    "params" : {        "factor" : 1.1      },            "order" : "asc"        }    } }' In this case, we have used a match_all query and a sort script. If everything is correct, the result returned by ElasticSearch should be as shown in the following code: { "took" : 7, "timed_out" : false, "_shards" : {    "total" : 5,    "successful" : 5,    "failed" : 0 }, "hits" : {    "total" : 1000,    "max_score" : null,    "hits" : [ {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "161",      "_score" : null, "_source" : … truncated …,      "sort" : [ 0.0278578661440021 ]    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "634",      "_score" : null, "_source" : … truncated …,     "sort" : [ 0.08131364254827411 ]    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "465",      "_score" : null, "_source" : … truncated …,      "sort" : [ 0.1094966959069832 ]    } ] } } How it works... The sort scripting allows you to define several parameters, as follows: order (default "asc") ("asc" or "desc"): This determines whether the order must be ascending or descending. script: This contains the code to be executed. type: This defines the type to convert the value. params (optional, a JSON object): This defines the parameters that need to be passed. lang (by default, groovy): This defines the scripting language to be used. ignore_unmapped (optional): This ignores unmapped fields in a sort. This flag allows you to avoid errors due to missing fields in shards. Extending the sort with scripting allows the use of a broader approach to score your hits. ElasticSearch scripting permits the use of every code that you want. You can create custom complex algorithms to score your documents. There's more... Groovy provides a lot of built-in functions (mainly taken from Java's Math class) that can be used in scripts, as shown in the following table: Function Description time() The current time in milliseconds sin(a) Returns the trigonometric sine of an angle cos(a) Returns the trigonometric cosine of an angle tan(a) Returns the trigonometric tangent of an angle asin(a) Returns the arc sine of a value acos(a) Returns the arc cosine of a value atan(a) Returns the arc tangent of a value toRadians(angdeg) Converts an angle measured in degrees to an approximately equivalent angle measured in radians toDegrees(angrad) Converts an angle measured in radians to an approximately equivalent angle measured in degrees exp(a) Returns Euler's number raised to the power of a value log(a) Returns the natural logarithm (base e) of a value log10(a) Returns the base 10 logarithm of a value sqrt(a) Returns the correctly rounded positive square root of a value cbrt(a) Returns the cube root of a double value IEEEremainder(f1, f2) Computes the remainder operation on two arguments, as prescribed by the IEEE 754 standard ceil(a) Returns the smallest (closest to negative infinity) value that is greater than or equal to the argument and is equal to a mathematical integer floor(a) Returns the largest (closest to positive infinity) value that is less than or equal to the argument and is equal to a mathematical integer rint(a) Returns the value that is closest in value to the argument and is equal to a mathematical integer atan2(y, x) Returns the angle theta from the conversion of rectangular coordinates (x,y_) to polar coordinates (r,_theta) pow(a, b) Returns the value of the first argument raised to the power of the second argument round(a) Returns the closest integer to the argument random() Returns a random double value abs(a) Returns the absolute value of a value max(a, b) Returns the greater of the two values min(a, b) Returns the smaller of the two values ulp(d) Returns the size of the unit in the last place of the argument signum(d) Returns the signum function of the argument sinh(x) Returns the hyperbolic sine of a value cosh(x) Returns the hyperbolic cosine of a value tanh(x) Returns the hyperbolic tangent of a value hypot(x,y) Returns sqrt(x^2+y^2) without an intermediate overflow or underflow acos(a) Returns the arc cosine of a value atan(a) Returns the arc tangent of a value If you want to retrieve records in a random order, you can use a script with a random method, as shown in the following code: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "match_all": {} }, "sort": {    "_script" : {      "script" : "Math.random()",      "lang" : "groovy",      "type" : "number",      "params" : {}    } } }' In this example, for every hit, the new sort value is computed by executing the Math.random() scripting function. See also The official ElasticSearch documentation at http://www.elasticsearch.org/guide/en/elasticsearch/reference/current/modules-scripting.html Computing return fields with scripting ElasticSearch allows you to define complex expressions that can be used to return a new calculated field value. These special fields are called script_fields, and they can be expressed with a script in every available ElasticSearch scripting language. Getting ready You will need a working ElasticSearch cluster and an index populated with the script (chapter_06/populate_aggregations.sh), which is available at https://github.com/aparo/ elasticsearch-cookbook-second-edition. How to do it... In order to compute return fields with scripting, perform the following steps: Return the following script fields: "my_calc_field": This concatenates the text of the "name" and "description" fields "my_calc_field2": This multiplies the "price" value by the "discount" parameter From the command line, execute the following code: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/ _search?&pretty=true&size=3' -d '{ "query": {    "match_all": {} }, "script_fields" : {    "my_calc_field" : {      "script" : "doc["name"].value + " -- " + doc["description"].value"    },    "my_calc_field2" : {      "script" : "doc["price"].value * discount",      "params" : {       "discount" : 0.8      }    } } }' If everything works all right, this is how the result returned by ElasticSearch should be: { "took" : 4, "timed_out" : false, "_shards" : {    "total" : 5,    "successful" : 5,    "failed" : 0 }, "hits" : {    "total" : 1000,    "max_score" : 1.0,    "hits" : [ {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "4",      "_score" : 1.0,      "fields" : {        "my_calc_field" : "entropic -- accusantium",        "my_calc_field2" : 5.480038242170081      }    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "9",      "_score" : 1.0,      "fields" : {        "my_calc_field" : "frankie -- accusantium",        "my_calc_field2" : 34.79852410178313      }    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "11",      "_score" : 1.0,      "fields" : {        "my_calc_field" : "johansson -- accusamus",        "my_calc_field2" : 11.824173084636591      }    } ] } } How it works... The scripting fields are similar to executing an SQL function on a field during a select operation. In ElasticSearch, after a search phase is executed and the hits to be returned are calculated, if some fields (standard or script) are defined, they are calculated and returned. The script field, which can be defined with all the supported languages, is processed by passing a value to the source of the document and, if some other parameters are defined in the script (in the discount factor example), they are passed to the script function. The script function is a code snippet; it can contain everything that the language allows you to write, but it must be evaluated to a value (or a list of values). See also The Installing additional script plugins recipe in this article to install additional languages for scripting The Sorting using script recipe to have a reference of the extra built-in functions in Groovy scripts Filtering a search via scripting ElasticSearch scripting allows you to extend the traditional filter with custom scripts. Using scripting to create a custom filter is a convenient way to write scripting rules that are not provided by Lucene or ElasticSearch, and to implement business logic that is not available in the query DSL. Getting ready You will need a working ElasticSearch cluster and an index populated with the (chapter_06/populate_aggregations.sh) script, which is available at https://github.com/aparo/ elasticsearch-cookbook-second-edition. How to do it... In order to filter a search using a script, perform the following steps: Write a search with a filter that filters out a document with the value of age less than the parameter value: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "filtered": {      "filter": {        "script": {          "script": "doc["age"].value > param1",          "params" : {            "param1" : 80          }        }      },      "query": {        "match_all": {}      }    } } }' In this example, all the documents in which the value of age is greater than param1 are qualified to be returned. If everything works correctly, the result returned by ElasticSearch should be as shown here: { "took" : 30, "timed_out" : false, "_shards" : {    "total" : 5,    "successful" : 5,    "failed" : 0 }, "hits" : {    "total" : 237,    "max_score" : 1.0,    "hits" : [ {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "9",      "_score" : 1.0, "_source" :{ … "age": 83, … }    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "23",      "_score" : 1.0, "_source" : { … "age": 87, … }    }, {      "_index" : "test-index",      "_type" : "test-type",      "_id" : "47",      "_score" : 1.0, "_source" : {…. "age": 98, …}    } ] } } How it works... The script filter is a language script that returns a Boolean value (true/false). For every hit, the script is evaluated, and if it returns true, the hit passes the filter. This type of scripting can only be used as Lucene filters, not as queries, because it doesn't affect the search (the exceptions are constant_score and custom_filters_score). These are the scripting fields: script: This contains the code to be executed params: These are optional parameters to be passed to the script lang (defaults to groovy): This defines the language of the script The script code can be any code in your preferred and supported scripting language that returns a Boolean value. There's more... Other languages are used in the same way as Groovy. For the current example, I have chosen a standard comparison that works in several languages. To execute the same script using the JavaScript language, use the following code: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "filtered": {      "filter": {        "script": {          "script": "doc["age"].value > param1",          "lang":"javascript",          "params" : {            "param1" : 80          }        }      },      "query": {        "match_all": {}      }    } } }' For Python, use the following code: curl -XGET 'http://127.0.0.1:9200/test-index/test-type/_search?&pretty=true&size=3' -d '{ "query": {    "filtered": {      "filter": {        "script": {          "script": "doc["age"].value > param1",          "lang":"python",          "params" : {            "param1" : 80          }        }      },      "query": {        "match_all": {}      }    } } }' See also The Installing additional script plugins recipe in this article to install additional languages for scripting The Sorting data using script recipe in this article to get a reference of the extra built-in functions in Groovy scripts Summary In this article you have learnt the ways you can use scripting to extend the ElasticSearch functional capabilities using different programming languages. Resources for Article: Further resources on this subject: Indexing the Data [Article] Low-Level Index Control [Article] Designing Puppet Architectures [Article]

In this article, written by Jayant Varma, the author of Xcode 6 Essentials, we shall look at Xcode closely as this is going to be the tool you would use quite a lot for all aspects of your app development for Apple devices. It is a good idea to know and be familiar with the interface, the sections, shortcut keys, and so on. (For more resources related to this topic, see here.) Starting Xcode Xcode, like many other Mac applications, is found in the Applications folder or the Launchpad. On starting Xcode, you will be greeted with the launch screen that offers some entry points for working with Xcode. Mostly, you will select Create a new Xcode project or Check out an existing project , if you have an existing project to continue work on. Xcode remembers what it was doing last, so if you had a project or file open, it will open up those windows again. Creating a new project After selecting the Create a new project option, we are guided via a wizard that helps us get started. Selecting the project type The first step is to select what type of project you want to create. At the moment, there are two distinct types of projects, mobile (iOS) or desktop (OS X) that you can create. Within each of those types, you can select the type of project you want. The screenshot displays a standard configuration for iOS application projects. The templates used when the selected type of project is created are self sufficient, that is, when the Run button is pressed, the app compiles and runs. It might do nothing, as this is a minimalistic template. On selecting the type of project, we can select the next step: Setting the project options This step allows selecting the options, namely setting the application name, the organization name, identifier, language, and devices to support. In the past, the language was always set to Objective-C, however with Xcode 6, there are two options: objective-C and Swift Setting the project properties On creation, the main screen is displayed. Here it offers the option to change other details related to the application such as the version number and build. It also allows you to configure the team ID and certificates used for signing the application to test on a mobile device or for distribution to the App Store. It also allows you to set the compatibility for earlier versions. The orientation and app icons, splash screens, and so on are also set from this screen. If you want to set these up later on in the project, it is fine, this can be accessed at any time and does not stop you from development. It needs to be set prior to deploying it on a device or creating an App Store ready application. Xcode overview Let us have a look at the Xcode interface to familiarize ourselves with the same as it would help improve productivity when building your application. The top section immediately following the traffic light (window chrome) displays a Play and Stop button. This allows the project to run and stop. The breadcrumb toolbar displays the project-specific settings with respect to the product and the target. With an iOS project, it could be a particular simulator for iPhone, iPad, and so on, or a physical device (number 5 in the following screenshot). Just under this are vertical areas that are the main content area with all the files, editors, UI, and so on. These can be displayed or hidden as required and can be stacked vertically or horizontally. The distinct areas in Xcode are as follows: Project navigation (number1) Editor and assistant editor (number 2 ) and (number 3 ) Utility/inspector (number 4 ) The toolbar (number 5 ) and (number 6 ) These sections can be switched on and off (shown or hidden) as required to make space for other sections or more screen space to work with: Sections in Xcode The project section The project navigation section has three sub sections, the topmost being the project toolbar that has eight icons. These can be seen as in the following screenshot. The next sub section contains the project files and all the assets required for this project. The bottom most section consists of recently edited files and filters: You can use the keyboard shortcuts to access these areas quickly with the CMD + 1...8 keys. The eight areas available under project navigation are key and for the beginner to Xcode, this could be a bit daunting. When you run the project, the current section might change and display another where you might wonder how to get back to the project (file) navigator. Getting familiar with these is always helpful and the easiest way to navigate between these is the CMD + 1..8 keys. Project navigator ( CMD + 1 ): This displays all of the files, folders, assets, frameworks, and so on that are part of this project. This is displayed as a hierarchical view and is the way that a majority of developers access their files, folders, and so on. Symbol navigator ( CMD + 2 ): This displays all of the classes, members, and methods that are available in them. This is the easiest way to navigate quickly to a method/function, attribute/property. Search navigator ( CMD + 3 ): This allows you to search the project for a particular match. This is quite useful to find and replace text. Issues navigator ( CMD + 4 ): This displays the warning and errors that occur while typing your code or on building and running it. This also displays the results of the static analyzer. Tests navigator ( CMD + 5 ); This displays the tests that you have present in your code either added by yourself or the default ones created with the project. Debug navigator ( CMD + 6 ): This displays the information about the application when you choose to run it. It has some amazing detailed information on CPU usage, memory usage, disk usage, threads, and so on. Breakpoint navigator ( CMD + 7 ): This displays all the breakpoints in your project from all files. This also allows you to create exception and symbolic breakpoints. Log navigator ( CMD + 8 ): This displays a log of all actions carried out, namely compiling, building, and running. This is more useful when used to determine the results of automated builds The editor and assistant editor sections The second area contains the editor and assistant editor sections. These display the code, the XIB (as appropriate), storyboard files, device previews, and so on. Each of the sub sections have a jump bar on the top that relates to files and allow for navigating back and forth in the files and display the location of the file in the workspace. To the right from this is a mini issues navigator that displays all warnings and errors. In the case of the assistant editors, it also displays two buttons: one to add a new assistant editor area and another to close it.   Source code editors While we are looking at the interface, it is worth noting that the Xcode code editor is a very advanced editor with a lot of features, which is now seen as standard with a lot of text editors. Some of the features that make working with Xcode easier are as follows: Code folding : This feature helps to hide code at points such as the function declaration, loops, matching brace brackets, and so on. When a function or portion of code is folded, it hides it from view, thereby allowing you to view other areas of the code that would not be visible unless you scrolled. Syntax highlighting : This is one of the most useful features as it helps you, the developer, to visually, at a glance, differentiate your source code from variables, constants, and strings. Xcode has syntax highlighting for a couple of languages as mentioned earlier. Context help : This is one of the best features whereby when you hover over a word in the source code with OPT pressed, it shows a dotted underline and the cursor changes to a question mark. When you click on a word with the dotted underline and the question mark cursor, it displays a popup with details about that word. It also highlights all instances of that word in the file. The popup details as much information as available. If it is a variable or a function that you have added to the code, then it will display the name of the file where it was declared. If it is a word that is contained in the Apple libraries, then it displays the description and other additional details. Context jump : This is another cool feature that allows jumping to the point of declaration of that word. This is achieved by clicking on a word while keeping the CMD button pressed. In many cases, this is mainly helpful to know how the function is declared and what parameters it expects. It can also be useful to get information on other enumerators and constants used with that function. The jump could be in the same file as where you are editing the code or it could be to the header files where they are declared. Edit all in scope : This is a cool feature where you can edit all of the instances of the word together rather than using search and replace. A case scenario is if you want to change the name of a variable and ensure that all instances you are using in the file are changed but not the ones that are text, then you can use this option to quickly change it. Catching mistakes with fix-it : This is another cool feature in Xcode that will save you a lot of time and hassle. As you type text, Xcode keeps analyzing the code and looking for errors. If you have declared a variable and not used it in your code, Xcode immediately draws attention to it suggesting that the variable is an unused variable. However, if it was supposed to be a pointer and you have declared it without *; Xcode immediately flags it as an error that the interface type cannot be statically allocated. It offers a fix-it solution of inserting * and the code has a greyed * character showing where it will be added. This helps the developer fix commonly overlooked issues such as missing semicolons, missing declarations, or misspelled variable names. Code completion : This is the bit that makes writing code so much easier, type in a few letters of the function name and Xcode pops up a list of functions, constants, methods, and so on that start with those letters and displays all of the required parameters (as applicable) including the return type. When selected, it adds the token placeholders that can be replaced with the actual parameter values. The results might vary from person to person depending on the settings and the speed of the system you run Xcode on. The assistant editor The assistant editor is mainly used to display the counterparts and related files to the file open in the primary editor (generally used when working with Objective-C where the .h or.m files are the related files). The assistant editors track the contents of the editor. Xcode is quite intelligent and knows the corresponding sections and counterparts. When you click on a file, it opens up in the editor. However, pressing the OPT + Shift while clicking on the file, you would be provided with an interactive dialog to select where to open the file. The options include the primary editor or the assistant editor. You can also add assistant editors as required.   Another way to open a file quickly is to use the Open Quickly option, which has a shortcut key of CMD + Shift + O . This displays a textbox that allows accessing a file from the project. The utility/inspector section The last section contains the inspector and library. This section changes based on the type of file selected in the current editor. The inspector has 6 tabs/sections and they are as follows: The file inspector ( CMD + OPT + 1 ): This displays the physical file information for the file selected. For code files, it is the text encoding, the targets that it belongs to, and the physical file path. While for the storyboard, it is the physical file path and allows setting attributes such as auto layout and size classes (new in Xcode 6). The quick help inspector ( CMD + OPT + 2 ): This displays information about the class or object selected. The identity inspector ( CMD + OPT + 3 ): This displays the class name, ID, and others that identify the object selected. The attributes inspector ( CMD + OPT + 4 ): This displays the attributes for the object selected as if it is the initial root view controller, does it extend under the top bars or not, if it has a navigation bar or not, and others. This also displays the user-defined attributes (a new feature with Xcode 6). The size inspector ( CMD + OPT + 5 ): This displays the size of the control selected and the associated constraints that help position it on the container. The connections inspector ( CMD + OPT + 6 ): This displays the connections created in the Interface Builder between the UI and the code. The lower half of this inspector contains four options that help you work efficiently, they are as follows: The file template library : This contains the options to create a new class, protocol. The options that are available when selecting the File | New option from the menu. The code snippets library : This is a wonderful but not widely used option. This can hold code snippets that can help you avoid writing repetitive blocks of code in your app. You can drag and drop the snippet to your code in the editor. This also offers features such as shortcuts, scopes, platforms, and languages. So you can have a shortcut such as appDidLoad (for example) that inserts the code to create and populate a button. This is achieved simply by setting the platform as appropriate to iOS or OS X. After creating a code snippet, as soon as you type the first few characters, the code snippet shows up in the list of autocomplete options; The object library : This is the toolbox that contains all of the controls that you need for creating your UI, be it a button, a label, a Table View, view, View Controller, or anything else. Adding a code snippet is as easy as dragging the selected code from the editor onto the snippet area. It is a little tricky because the moment you start dragging, it could break your selection highlight. You need to select the text, click (hold) and then drag it. The media library : This contains the list of all images and other media types that are available to this project/workspace. Summary In this article, you have seen a quick tour of Xcode, keeping the shortcuts and tips handy as they really do help get things done faster. The code snippets are a wonderful feature that allow for quickly setting up commonly used code with shortcut keywords. Resources for Article: Further resources on this subject: Introducing Xcode Tools for iPhone Development [article] Xcode 4 ios: Displaying Notification Messages [article] Linking OpenCV to an iOS project [article]

In this article by Matt Kaufman and Michael Wicherski, authors of the book Learning Apex Programming, we will see how we can use Apex to extend the Salesforce1 Platform. We will also see how to create a customized Force.com page. (For more resources related to this topic, see here.) Apex on its own is a powerful tool to extend the Salesforce1 Platform. It allows you to define your own database logic and fully customize the behavior of the platform. Sometimes, controlling "what happens behind the scenes isn't enough. You might have a complex process that needs to step users through a wizard or need to present data in a format that isn't native to the Salesforce1 Platform, or maybe even make things look like your corporate website. Anytime you need to go beyond custom logic and implement a custom interface, you can turn to Visualforce. Visualforce is the user interface framework for the Salesforce1 Platform. It supports the use of HTML, JavaScript, CSS, and Flash—all of which enable you to build your own custom web pages. These web pages are stored and hosted by the Salesforce1 Platform and can be exposed to just your internal users, your external community users, or publicly to the world. But wait, there's more! Also included with Visualforce is a robust markup language. This markup language (which is also referred to as Visualforce) allows you to bind your web pages to data and actions stored on the platform. It also allows you to leverage Apex for code-based objects and actions. Like the rest of the platform, the markup portion of Visualforce is upgraded three times a year with new tags and features. All of these features mean that Visualforce is very powerful. s-con-what? Before the "introduction of Visualforce, the Salesforce1 Platform had a feature called s-controls. These were simple files where you could write HTML, CSS, and JavaScript. There was no custom markup language included. In order to make things look like the Force.com GUI, a lot of HTML was required. If you wanted to create just a simple input form for a new Account record, so much HTML code was required. The following is just a" small, condensed excerpt of what the HTML would look like if you wanted to recreate such a screen from scratch: <div class="bPageTitle"><div class="ptBody"><div class="content"> <img src="/s.gif" class="pageTitleIcon" title="Account" /> <h1 class="pageType">    Account Edit<span class="titleSeparatingColon">:</span> </h1> <h2 class="pageDescription"> New Account</h2> <div class="blank">&nbsp;</div> </div> <div class="links"></div></div><div   class="ptBreadcrumb"></div></div> <form action="/001/e" method="post" onsubmit="if   (window.ffInAlert) { return false; }if (window.sfdcPage   &amp;&amp; window.sfdcPage.disableSaveButtons) { return   window.sfdcPage.disableSaveButtons(); }"> <div class="bPageBlock brandSecondaryBrd bEditBlock   secondaryPalette"> <div class="pbHeader">    <table border="0" cellpadding="0" cellspacing="0"><tbody>      <tr>      <td class="pbTitle">      <img src="/s.gif" width="12" height="1" class="minWidth"         style="margin-right: 0.25em;margin-right: 0.25em;margin-       right: 0.25em;">      <h2 class="mainTitle">Account Edit</h2>      </td>      <td class="pbButton" id="topButtonRow">      <input value="Save" class="btn" type="submit">      <input value="Cancel" class="btn" type="submit">      </td>      </tr>    </tbody></table> </div> <div class="pbBody">    <div class="pbSubheader brandTertiaryBgr first       tertiaryPalette" >    <span class="pbSubExtra"><span class="requiredLegend       brandTertiaryFgr"><span class="requiredExampleOuter"><span       class="requiredExample">&nbsp;</span></span>      <span class="requiredMark">*</span>      <span class="requiredText"> = Required Information</span>      </span></span>      <h3>Account Information<span         class="titleSeparatingColon">:</span> </h3>    </div>    <div class="pbSubsection">    <table class="detailList" border="0" cellpadding="0"     cellspacing="0"><tbody>      <tr>        <td class="labelCol requiredInput">        <label><span class="requiredMark">*</span>Account         Name</label>      </td>      <td class="dataCol col02">        <div class="requiredInput"><div         class="requiredBlock"></div>        <input id="acc2" name="acc2" size="20" type="text">        </div>      </td>      <td class="labelCol">        <label>Website</label>      </td>      <td class="dataCol">        <span>        <input id="acc12" name="acc12" size="20" type="text">        </span>      </td>      </tr>    </tbody></table>    </div> </div> <div class="pbBottomButtons">    <table border="0" cellpadding="0" cellspacing="0"><tbody>    <tr>      <td class="pbTitle"><img src="/s.gif" width="12" height="1"       class="minWidth" style="margin-right: 0.25em;margin-right:       0.25em;margin-right: 0.25em;">&nbsp;</td>      <td class="pbButtonb" id="bottomButtonRow">      <input value=" Save " class="btn" title="Save"         type="submit">      <input value="Cancel" class="btn" type="submit">      </td>    </tr>    </tbody></table> </div> <div class="pbFooter secondaryPalette"><div class="bg"> </div></div> </div> </form> We did our best to trim down this HTML to as little as possible. Despite all of our efforts, it still "took up more space than we wanted. The really sad part is that all of that code only results in the following screenshot: Not only was it time consuming to write all this HTML, but odds were that we wouldn't get it exactly right the first time. Worse still, every time the business requirements changed, we had to go through the exhausting effort of modifying the HTML code. Something had to change in order to provide us relief. That something was the introduction of Visualforce and its markup language. Your own personal Force.com The markup "tags in Visualforce correspond to various parts of the Force.com GUI. These tags allow you to quickly generate HTML markup without actually writing any HTML. It's really one of the greatest tricks of the Salesforce1 Platform. You can easily create your own custom screens that look just like the built-in ones with less effort than it would take you to create a web page for your corporate website. Take a look at the Visualforce markup that corresponds to the HTML and screenshot we showed you earlier: <apex:page standardController="Account" > <apex:sectionHeader title="Account Edit" subtitle="New Account"     /> <apex:form>    <apex:pageBlock title="Account Edit" mode="edit" >      <apex:pageBlockButtons>        <apex:commandButton value="Save" action="{!save}" />        <apex:commandButton value="Cancel" action="{!cancel}" />      </apex:pageBlockButtons>      <apex:pageBlockSection title="Account Information" >        <apex:inputField value="{!account.Name}" />        <apex:inputField value="{!account.Website}" />      </apex:pageBlockSection>    </apex:pageBlock> </apex:form> </apex:page> Impressive! With "merely these 15 lines of markup, we can render nearly 100 lines of earlier HTML. Don't believe us, you can try it out yourself. Creating a Visualforce page Just like" triggers and classes, Visualforce pages can "be created and edited using the Force.com IDE. The Force.com GUI also includes a web-based editor to work with Visualforce pages. To create a new Visualforce page, perform these simple steps: Right-click on your project and navigate to New | Visualforce Page. The Create New Visualforce Page window appears as shown: Enter" the label and name for your "new page in the Label and Name fields, respectively. For this example, use myTestPage. Select the API version for the page. For this example, keep it at the default value. Click on Finish. A progress bar will appear followed by your new Visualforce page. Remember that you always want to create your code in a Sandbox or Developer Edition org, not directly in Production. It is technically possible to edit Visualforce pages in Production, but you're breaking all sorts of best practices when you do. Similar to other markup languages, every tag in a Visualforce page must be closed. Tags and their corresponding closing tags must also occur in a proper order. The values of tag attributes are enclosed by double quotes; however, single quotes can be used inside the value to denote text values. Every Visualforce page starts with the <apex:page> tag and ends with </apex:page> as shown: <apex:page> <!-- Your content goes here --> </apex:page> Within "the <apex:page> tags, you can paste "your existing HTML as long as it is properly ordered and closed. The result will be a web page hosted by the Salesforce1 Platform. Not much to see here If you are" a web developer, then there's a lot you can "do with Visualforce pages. Using HTML, CSS, and images, you can create really pretty web pages that educate your users. If you have some programming skills, you can also use JavaScript in your pages to allow for interaction. If you have access to web services, you can use JavaScript to call the web services and make a really powerful application. Check out the following Visualforce page for an example of what you can do: <apex:page> <script type="text/javascript"> function doStuff(){    var x = document.getElementById("myId");    console.log(x); } </script> <img src="http://www.thisbook.com/logo.png" /> <h1>This is my title</h1> <h2>This is my subtitle</h2> <p>In a world where books are full of code, there was only one     that taught you everything you needed to know about Apex!</p> <ol>    <li>My first item</li>    <li>Etc.</li> </ol> <span id="myId"></span> <iframe src="http://www.thisbook.com/mypage.html" /> <form action="http://thisbook.com/submit.html" >    <input type="text" name="yoursecret" /> </form> </apex:page> All of this code is standalone and really has nothing to do with the Salesforce1 Platform other than being hosted by it. However, what really makes Visualforce powerful is its ability to interact with your data, which allows your pages to be more dynamic. Even better, you" can write Apex code to control how "your pages behave, so instead of relying on client-side JavaScript, your logic can run server side. Summary In this article we learned how a few features of Apex and how we can use it to extend the SalesForce1 Platform. We also created a custom Force.com page. Well, you've made a lot of progress. Not only can you write code to control how the database behaves, but you can create beautiful-looking pages too. You're an Apex rock star and nothing is going to hold you back. It's time to show your skills to the world. If you want to dig deeper, buy the book and read Learning Apex Programming in a simple step-by-step fashion by using Apex, the language for extension of the Salesforce1 Platform. Resources for Article: Further resources on this subject: Learning to Fly with Force.com [article] Building, Publishing, and Supporting Your Force.com Application [article] Adding a Geolocation Trigger to the Salesforce Account Object [article]

This article by Vinith Menon, the author of Microsoft Hyper-V PowerShell Automation, delves into the basics of Hyper-V, right from installing Hyper-V to resizing virtual hard disks. The Hyper-V PowerShell module includes several significant features that extend its use, improve its usability, and allow you to control and manage your Hyper-V environment with more granular control. Various organizations have moved on from Hyper-V (V2) to Hyper-V (V3). In Hyper-V (V2), the Hyper-V management shell was not built-in and the PowerShell module had to be manually installed. In Hyper-V (V3), Microsoft has provided an exhaustive set of cmdlets that can be used to manage and automate all configuration activities of the Hyper-V environment. The cmdlets are executed across the network using Windows Remote Management. In this article, we will cover: The basics of setting up a Hyper-V environment using PowerShell The fundamental concepts of Hyper-V management with the Hyper-V management shell The updated features in Hyper-V (For more resources related to this topic, see here.) Here is a list of all the new features introduced in Hyper-V in Windows Server 2012 R2. We will be going in depth through the important changes that have come into the Hyper-V PowerShell module with the following features and functions: Shared virtual hard disk Resizing the live virtual hard disk Installing and configuring your Hyper-V environment Installing and configuring Hyper-V using PowerShell Before you proceed with the installation and configuration of Hyper-V, there are some prerequisites that need to be taken care of: The user account that is used to install the Hyper-V role should have administrative privileges on the computer There should be enough RAM on the server to run newly created virtual machines Once the prerequisites have been taken care of, let's start with installing the Hyper-V role: Open a PowerShell prompt in Run as Administrator mode: Type the following into the PowerShell prompt to install the Hyper-V role along with the management tools; once the installation is complete, the Hyper-V Server will reboot and the Hyper-V role will be successfully installed: Install-WindowsFeature –Name Hyper-V -IncludeManagementTools - Restart Once the server boots up, verify the installation of Hyper-V using the Get-WindowsFeature cmdlet: Get-WindowsFeature -Name hyper* You will be able to see that the Hyper-V role, Hyper-V PowerShell management shell, and the GUI management tools are successfully installed:   Fundamental concepts of Hyper-V management with the Hyper-V management shell In this section, we will look at some of the fundamental concepts of Hyper-V management with the Hyper-V management shell. Once you get the Hyper-V role installed as per the steps illustrated in the previous section, a PowerShell module to manage your Hyper-V environment will also get installed. Now, perform the following steps: Open a PowerShell prompt in the Run as Administrator mode. PowerShell uses cmdlets that are built using a verb-noun naming system (for more details, refer to Learning Windows PowerShell Names at http://technet.microsoft.com/en-us/library/dd315315.aspx). Type the following command into the PowerShell prompt to get a list of all the cmdlets in the Hyper-V PowerShell module: Get-Command -Module Hyper-V Hyper-V in Windows Server 2012 R2 ships with about 178 cmdlets. These cmdlets allow a Hyper-V administrator to handle very simple, basic tasks to advanced ones such as setting up a Hyper-V replica for virtual machine disaster recovery. To get the count of all the available Hyper-V cmdlets, you can type the following command in PowerShell: Get-Command -Module Hyper-V | Measure-Object The Hyper-V PowerShell cmdlets follow a very simple approach and are very user friendly. The cmdlet name itself indirectly communicates with the Hyper-V administrator about its functionality. The following screenshot shows the output of the Get command: For example, in the following screenshot, the Remove-VMSwitch cmdlet itself says that it's used to delete a previously created virtual machine switch: If the administrator is still not sure about the task that can be performed by the cmdlet, he or she can get help with detailed examples using the Get-Help cmdlet. To get help on the cmdlet type, type the cmdlet name in the prescribed format. To make sure that the latest version of help files are installed on the server, run the Update-Help cmdlet before executing the following cmdlet: Get-Help <Hyper-V cmdlet> -Full The following screenshot is an example of the Get-Help cmdlet: Shared virtual hard disks This new and improved feature in Windows Server 2012 R2 allows an administrator to share a virtual hard disk file (the .vhdx file format) between multiple virtual machines. These .vhdx files can be used as shared storage for a failover cluster created between virtual machines (also known as guest clustering). A shared virtual hard disk allows you to create data disks and witness disks using .vhdx files with some advantages: Shared disks are ideal for SQL database files and file servers Shared disks can be run on generation 1 and generation 2 virtual machines This new feature allows you to save on storage costs and use the .vhdx files for guest clustering, enabling easier deployment rather than using virtual Fibre Channel or Internet Small Computer System Interface (iSCSI), which are complicated and require storage configuration changes such as zoning and Logic Unit Number (LUN) masking. In Windows Server 2012 R2, virtual iSCSI disks (both shared and unshared virtual hard disk files) show up as virtual SAS disks when you add an iSCSI hard disk to a virtual machine. Shared virtual hard disks (.vhdx) files can be placed on Cluster Shared Volumes (CSV) or a Scale-Out File Server cluster Let's look at the ways you can automate and manage your shared .vhdx guest clustering configuration using PowerShell. In the following example, we will demonstrate how you can create a two-node file server cluster using the shared VHDX feature. After that, let's set up a testing environment within which we can start learning these new features. The steps are as follows: We will start by creating two virtual machines each with 50 GB OS drives, which contains a sysprep image of Windows Server 2012 R2. Each virtual machine will have 4 GB RAM and four virtual CPUs. D:vhdbase_1.vhdx and D:vhdbase_2.vhdx are already existing VHDX files with sysprepped image of Windows Server 2012 R2. The following code is used to create two virtual machines: New-VM –Name "Fileserver_VM1" –MemoryStartupBytes 4GB – NewVHDPath d:vhdbase_1.vhdx -NewVHDSizeBytes 50GB New-VM –Name "Fileserver_VM2" –MemoryStartupBytes 4GB –NewVHDPath d:vhdbase_2.vhdx -NewVHDSizeBytes 50GB Next, we will install the file server role and configure a failover cluster on both the virtual machines using PowerShell. You need to enable PowerShell remoting on both the file servers and also have them joined to a domain. The following is the code: Install-WindowsFeature -computername Fileserver_VM1 File- Services, FS-FileServer, Failover-Clustering   Install-WindowsFeature -computername Fileserver_VM1 RSAT- Clustering –IncludeAllSubFeature   Install-WindowsFeature -computername Fileserver_VM2 File- Services, FS-FileServer, Failover-Clustering   Install-WindowsFeature -computername Fileserver_VM2 RSAT- Clustering -IncludeAllSubFeature Once we have the virtual machines created and the file server and failover clustering features installed, we will create the failover cluster as per Microsoft's best practices using the following set of cmdlets: New-Cluster -Name Cluster1 -Node FileServer_VM1,   FileServer_VM2 -StaticAddress 10.0.0.59 -NoStorage – Verbose You will need to choose a name and IP address that fits your organization. Next, we will create two vhdx files named sharedvhdx_data.vhdx (which will be used as a data disk) and sharedvhdx_quorum.vhdx (which will be used as the quorum or the witness disk). To do this, the following commands need to be run on the Hyper-V cluster: New-VHD -Path   c:ClusterStorageVolume1sharedvhdx_data.VHDX -Fixed - SizeBytes 10GB   New-VHD -Path   c:ClusterStorageVolume1sharedvhdx_quorum.VHDX -Fixed - SizeBytes 1GB Once we have created these virtual hard disk files, we will add them as shared .vhdx files. We will attach these newly created VHDX files to the Fileserver_VM1 and Fileserver_VM2 virtual machines and specify the parameter-shared VHDX files for guest clustering: Add-VMHardDiskDrive –VMName Fileserver_VM1 -Path   c:ClusterStorageVolume1sharedvhdx_data.VHDX – ShareVirtualDisk   Add-VMHardDiskDrive –VMName Fileserver_VM2 -Path   c:ClusterStorageVolume1sharedvhdx_data.VHDX – ShareVirtualDisk Finally, we will be making the disks available online and adding them to the failover cluster using the following command: Get-ClusterAvailableDisk | Add-ClusterDisk Once we have executed the preceding set of steps, we will have a highly available file server infrastructure using shared VHD files. Live virtual hard disk resizing With Windows Server 2012 R2, a newly added feature in Hyper-V allows the administrators to expand or shrink the size of a virtual hard disk attached to the SCSI controller while the virtual machines are still running. Hyper-V administrators can now perform maintenance operations on a live VHD and avoid any downtime by not temporarily shutting down the virtual machine for these maintenance activities. Prior to Windows Server 2012 R2, to resize a VHD attached to the virtual machine, it had to be turned off leading to costly downtime. Using the GUI controls, the VHD resize can be done by using only the Edit Virtual Hard Disk wizard. Also, note that the VHDs that were previously expanded can be shrunk. The Windows PowerShell way of doing a VHD resize is by using the Resize-VirtualDisk cmdlet. Let's look at the ways you can automate a VHD resize using PowerShell. In the next example, we will demonstrate how you can expand and shrink a virtual hard disk connected to a VM's SCSI controller. We will continue using the virtual machine that we created for our previous example. We have a pre-created VHD of 50 GB that is connected to the virtual machine's SCSI controller. Expanding the virtual hard disk Let's resize the aforementioned virtual hard disk to 57 GB using the Resize-Virtualdisk cmdlet: Resize-VirtualDisk -Name "scsidisk" -Size (57GB) Next, if we open the VM settings and perform an inspect disk operation, we'll be able to see that the VHDX file size has become 57 GB: Also, one can verify this when he or she logs into the VM, opens disk management, and extends the unused partition. You can see that the disk size has increased to 57 GB: Resizing the virtual hard disk Let's resize the earlier mentioned VHD to 57 GB using the Resize-Virtualdisk cmdlet: For this exercise, the primary requirement is to shrink the disk partition by logging in to the VM using disk management, as you can see in the following screenshot; we're shrinking the VHDX file by 7 GB: Next, click on Shrink. Once you complete this step, you will see that the unallocated space is 7 GB. You can also execute this step using the Resize-Partition Powershell cmdlet: Get-Partition -DiskNumber 1 | Resize-Partition -Size 50GB The following screenshot shows the partition: Next, we will resize/shrink the VHD to 50 GB: Resize-VirtualDisk -Name "scsidisk" -Size (50GB) Once the previous steps have been executed successfully, run a re-scan disk using disk management and you will see that the disk size is 50 GB: Summary In this article, we went through the basics of setting up a Hyper-V environment using PowerShell. We also explored the fundamental concepts of Hyper-V management with Hyper-V management shell. Resources for Article: Further resources on this subject: Hyper-V building blocks for creating your Microsoft virtualization platform [article] The importance of Hyper-V Security [article] Network Access Control Lists [article]

In this article by Jordan Krause, author of the book Windows Server 2012 R2 Administrator Cookbook, we will see how Windows Server 2012 R2 by Microsoft brings a whole new way of looking at remote access. Companies have historically relied on third-party tools to connect remote users into the network, such as traditional and SSL VPN provided by appliances from large networking vendors. I'm here to tell you those days are gone. Those of us running Microsoft-centric shops can now rely on Microsoft technologies to connect our remote workforce. Better yet is that these technologies are included with the Server 2012 R2 operating system, and have functionality that is much improved over anything that a traditional VPN can provide. Regular VPN does still have a place in the remote access space, and the great news is that you can also provide it with Server 2012 R2. Our primary focus for this article will be DirectAccess (DA). DA is kind of like automatic VPN. There is nothing the user needs to do in order to be connected to work. Whenever they are on the Internet, they are also connected automatically to the corporate network. DirectAccess is an amazing way to have your Windows 7 and Windows 8 domain joined systems connected back to the network for data access and for management of those traveling machines. DirectAccess has actually been around since 2008, but the first version came with some steep infrastructure requirements and was not widely used. Server 2012 R2 brings a whole new set of advantages and makes implementation much easier than in the past. I still find many server and networking admins who have never heard of DirectAccess, so let's spend some time together exploring some of the common tasks associated with it. In this article, we will cover the following recipes: Configuring DirectAccess, VPN, or a combination of the two Pre-staging Group Policy Objects (GPOs) to be used by DirectAccess Enhancing the security of DirectAccess by requiring certificate authentication Building your Network Location Server (NLS) on its own system  (For more resources related to this topic, see here.) There are two "flavors" of remote access available in Windows Server 2012 R2. The most common way to implement the Remote Access role is to provide DirectAccess for your Windows 7 and Windows 8 domain joined client computers, and VPN for the rest. The DirectAccess machines are typically your company-owned corporate assets. One of the primary reasons that DirectAccess is usually only for company assets is that the client machines must be joined to your domain, because the DirectAccess configuration settings are brought down to the client through a GPO. I doubt you want home and personal computers joining your domain. VPN is therefore used for down level clients such as Windows XP, and for home and personal devices that want to access the network. Since this is a traditional VPN listener with all regular protocols available such as PPTP, L2TP, SSTP, it can even work to connect devices such as smartphones. There is a third function available within the Server 2012 R2 Remote Access role, called the Web Application Proxy ( WAP ). This function is not used for connecting remote computers fully into the network as DirectAccess and VPN are; rather, WAP is used for publishing internal web resources out to the internet. For example, if you are running Exchange and Lync Server inside your network and want to publish access to these web-based resources to the internet for external users to connect to, WAP would be a mechanism that could publish access to these resources. The term for publishing out to the internet like this is Reverse Proxy, and WAP can act as such. It can also behave as an ADFS Proxy. For further information on the WAP role, please visit: http://technet.microsoft.com/en-us/library/dn584107.aspx One of the most confusing parts about setting up DirectAccess is that there are many different ways to do it. Some are good ideas, while others are not. Before we get rolling with recipes, we are going to cover a series of questions and answers to help guide you toward a successful DA deployment. The first question that always presents itself when setting up DA is "How do I assign IP addresses to my DirectAccess server?". This is quite a loaded question, because the answer depends on how you plan to implement DA, which features you plan to utilize, and even upon how secure you believe your DirectAccess server to be. Let me ask you some questions, pose potential answers to those questions, and discuss the effects of making each decision. DirectAccess Planning Q&A Which client operating systems can connect using DirectAccess? Answer: Windows 7 Ultimate, Windows 7 Enterprise, and Windows 8.x Enterprise. You'll notice that the Professional SKU is missing from this list. That is correct, Windows 7 and Windows 8 Pro do not contain the DirectAccess connectivity components. Yes, this does mean that Surface Pro tablets cannot utilize DirectAccess out of the box. However, I have seen many companies now install Windows 8 Enterprise onto their Surface tablets, effectively turning them into "Surface Enterprises." This works fine and does indeed enable them to be DirectAccess clients. In fact, I am currently typing this text on a DirectAccess-connected Surface "Pro turned Enterprise" tablet. Do I need one or two NICs on my DirectAccess server? Answer: Technically, you could set it up either way. In practice however, it really is designed for dual-NIC implementation. Single NIC DirectAccess works okay sometimes to establish a proof-of-concept to test out the technology. But I have seen too many problems with single NIC implementation in the field to ever recommend it for production use. Stick with two network cards, one facing the internal network and one facing the Internet. Do my DirectAccess servers have to be joined to the domain? Answer: Yes. Does DirectAccess have site-to-site failover capabilities? Answer: Yes, though only Windows 8.x client computers can take advantage of it. This functionality is called Multi-Site DirectAccess. Multiple DA servers that are spread out geographically can be joined together in a multi-site array. Windows 8 client computers keep track of each individual entry point and are able to swing between them as needed or at user preference. Windows 7 clients do not have this capability and will always connect through their primary site. What are these things called 6to4, Teredo, and IP-HTTPS I have seen in the Microsoft documentation? Answer: 6to4, Teredo, and IP-HTTPS are all IPv6 transition tunneling protocols. All DirectAccess packets that are moving across the internet between DA client and DA server are IPv6 packets. If your internal network is IPv4, then when those packets reach the DirectAccess server they get turned down into IPv4 packets, by some special components called DNS64 and NAT64. While these functions handle the translation of packets from IPv6 into IPv4 when necessary inside the corporate network, the key point here is that all DirectAccess packets that are traveling over the Internet part of the connection are always IPv6. Since the majority of the Internet is still IPv4, this means that we must tunnel those IPv6 packets inside something to get them across the Internet. That is the job of 6to4, Teredo, and IP-HTTPS. 6to4 encapsulates IPv6 packets into IPv4 headers and shuttles them around the internet using protocol 41. Teredo similarly encapsulates IPv6 packets inside IPv4 headers, but then uses UDP port 3544 to transport them. IP-HTTPS encapsulates IPv6 inside IPv4 and then inside HTTP encrypted with TLS, essentially creating an HTTPS stream across the Internet. This, like any HTTPS traffic, utilizes TCP port 443. The DirectAccess traffic traveling inside either kind of tunnel is always encrypted, since DirectAccess itself is protected by IPsec. Do I want to enable my clients to connect using Teredo? Answer: Most of the time, the answer here is yes. Probably the biggest factor that weighs on this decision is whether or not you are still running Windows 7 clients. When Teredo is enabled in an environment, this gives the client computers an opportunity to connect using Teredo, rather than all clients connecting in over the IP-HTTPS protocol. IP-HTTPS is sort of the "catchall" for connections, but Teredo will be preferred by clients if it is available. For Windows 7 clients, Teredo is quite a bit faster than IP-HTTPS. So enabling Teredo on the server side means your Windows 7 clients (the ones connecting via Teredo) will have quicker response times, and the load on your DirectAccess server will be lessened. This is because Windows 7 clients who are connecting over IP-HTTPS are encrypting all of the traffic twice. This also means that the DA server is encrypting/decrypting everything that comes and goes twice. In Windows 8, there is an enhancement that brings IP-HTTPS performance almost on par with Teredo, and so environments that are fully cut over to Windows 8 will receive less benefit from the extra work that goes into making sure Teredo works. Can I place my DirectAccess server behind a NAT? Answer: Yes, though there is a downside. Teredo cannot work if the DirectAccess server is sitting behind a NAT. For Teredo to be available, the DA server must have an External NIC that has two consecutive public IP addresses. True public addresses. If you place your DA server behind any kind of NAT, Teredo will not be available and all clients will connect using the IP-HTTPS protocol. Again, if you are using Windows 7 clients, this will decrease their speed and increase the load on your DirectAccess server. How many IP addresses do I need on a standalone DirectAccess server? Answer: I am going to leave single NIC implementation out of this answer since I don't recommend it anyway. For scenarios where you are sitting the External NIC behind a NAT or, for any other reason, are limiting your DA to IP-HTTPS only, then we need one external address and one internal address. The external address can be a true public address or a private NATed DMZ address. Same with the internal; it could be a true internal IP or a DMZ IP. Make sure both NICs are not plugged into the same DMZ, however. For a better installation scenario that allows Teredo connections to be possible, you would need two consecutive public IP addresses on the External NIC and a single internal IP on the Internal NIC. This internal IP could be either true internal or DMZ. But the public IPs would really have to be public for Teredo to work. Do I need an internal PKI? Answer: Maybe. If you want to connect Windows 7 clients, then the answer is yes. If you are completely Windows 8, then technically you do not need internal PKI. But you really should use it anyway. Using an internal PKI, which can be a single, simple Windows CA server, increases the security of your DirectAccess infrastructure. You'll find out during this article just how easy it is to require certificates as part of the tunnel building authentication process. Configuring DirectAccess, VPN, or a combination of the two Now that we have some general ideas about how we want to implement our remote access technologies, where do we begin? Most services that you want to run on a Windows Server begin with a role installation, but the implementation of remote access begins before that. Let's walk through the process of taking a new server and turning it into a Microsoft Remote Access server. Getting ready All of our work will be accomplished on a new Windows Server 2012 R2. We are taking the two-NIC approach to networking, and so we have two NICs installed on this server. The Internal NIC is plugged into the corporate network and the External NIC is plugged into the Internet for the sake of simplicity. The External NIC could just as well be plugged into a DMZ. How to do it... Follow these steps to turn your new server into a Remote Access server: Assign IP addresses to your server. Remember, the most important part is making sure that the Default Gateway goes on the External NIC only. Join the new server to your domain. Install an SSL certificate onto your DirectAccess server that you plan to use for the IP-HTTPS listener. This is typically a certificate purchased from a public CA. If you're planning to use client certificates for authentication, make sure to pull down a copy of the certificate to your DirectAccess server. You want to make sure certificates are in place before you start with the configuration of DirectAccess. This way the wizards will be able to automatically pull in information about those certificates in the first run. If you don't, DA will set itself up to use self-signed certificates, which are a security no-no. Use Server Manager to install the Remote Access role. You should only do this after completing the steps listed earlier. If you plan to load balance multiple DirectAccess servers together at a later time, make sure to also install the feature called Network Load Balancing . After selecting your role and feature, you will be asked which Remote Access role services you want to install. For our purposes in getting the remote workforce connected back into the corporate network, we want to choose DirectAccess and VPN (RAS) .  Now that the role has been successfully installed, you will see a yellow exclamation mark notification near the top of Server Manager indicating that you have some Post-deployment Configuration that needs to be done. Do not click on Open the Getting Started Wizard ! Unfortunately, Server Manager leads you to believe that launching the Getting Started Wizard (GSW) is the logical next step. However, using the GSW as the mechanism for configuring your DirectAccess settings is kind of like roasting a marshmallow with a pair of tweezers. In order to ensure you have the full range of options available to you as you configure your remote access settings, and that you don't get burned later, make sure to launch the configuration this way: Click on the Tools menu from inside Server Manager and launch the Remote Access Management Console . In the left window pane, click on Configuration | DirectAccess and VPN . Click on the second link, the one that says Run the Remote Access Setup Wizard . Please note that once again the top option is to run that pesky Getting Started Wizard. Don't do it! I'll explain why in the How it works… section of this recipe. Now you have a choice that you will have to answer for yourself. Are you configuring only DirectAccess, only VPN, or a combination of the two? Simply click on the option that you want to deploy. Following your choice, you will see a series of steps (steps 1 through 4) that need to be accomplished. This series of mini-wizards will guide you through the remainder of the DirectAccess and VPN particulars. This recipe isn't large enough to cover every specific option included in those wizards, but at least you now know the correct way to bring a DirectAccess/VPN server into operation. How it works... The remote access technologies included in Server 2012 R2 have great functionality, but their initial configuration can be confusing. Following the procedure listed in this recipe will set you on the right path to be successful in your deployment, and prevent you from running into issues down the road. The reason that I absolutely recommend you stay away from using the "shortcut" deployment method provided by the Getting Started Wizard is twofold: GSW skips a lot of options as it sets up DirectAccess, so you don't really have any understanding of how it works after finishing. You may have DA up and running, but have no idea how it's authenticating or working under the hood. This holds so much potential for problems later, should anything suddenly stop working. GSW employs a number of bad security practices in order to save time and effort in the setup process. For example, using the GSW usually means that your DirectAccess server will be authenticating users without client certificates, which is not a best practice. Also, it will co-host something called the NLS website on itself, which is also not a best practice. Those who utilize the GSW to configure DirectAccess will find that their GPO, which contains the client connectivity settings, will be security-filtered to the Domain Computers group. Even though it also contains a WMI filter that is supposed to limit that policy application to mobile hardware such as laptops, this is a terribly scary thing to see inside GPO filtering settings. You probably don't want all of your laptops to immediately start getting DA connectivity settings, but that is exactly what the GSW does for you. Perhaps worst, the GSW will create and make use of self-signed SSL certificates to validate its web traffic, even the traffic coming in from the Internet! This is a terrible practice and is the number one reason that should convince you that clicking on the Getting Started Wizard is not in your best interests. Pre-staging Group Policy Objects (GPOs) to be used by DirectAccess One of the great things about DirectAccess is that all of the connectivity settings the client computers need in order to connect are contained within a Group Policy Object (GPO). This means that you can turn new client computers into DirectAccess-connected clients without ever touching that system. Once configured properly, all you need to do is add the new computer account to an Active Directory security group, and during the next automatic Group Policy refresh cycle (usually within 90 minutes), that new laptop will be connecting via DirectAccess whenever outside the corporate network. You can certainly choose not to pre-stage anything with the GPOs and DirectAccess will still work. When you get to the end of the DA configuration wizards, it will inform you that two new GPOs are about to be created inside Active Directory. One GPO is used to contain the DirectAccess server settings and the other GPO is used to contain the DirectAccess client settings. If you allow the wizard to handle the generation of these GPOs, it will create them, link them, filter them, and populate them with settings automatically. About half of the time I see folks do it this way and they are forever happy with letting the wizard manage those GPOs now and in the future. The other half of the time, it is desired that we maintain a little more personal control over the GPOs. If you are setting up a new DA environment but your credentials don't have permission to create GPOs, the wizard is not going to be able to create them either. In this case, you will need to work with someone on your Active Directory team to get them created. Another reason to manage the GPOs manually is to have better control over placement of these policies. When you let the DirectAccess wizard create the GPOs, it will link them to the top level of your domain. It also sets Security Filtering on those GPOs so they are not going to be applied to everything in your domain, but when you open up the Group Policy Management Console you will always see those DirectAccess policies listed right up there at the top level of the domain. Sometimes this is simply not desirable. So for this reason also, you may want to choose to create and manage the GPOs by hand, so that we can secure placement and links where we specifically want them to be located. The key factors here are to make sure your DirectAccess Server Settings GPO applies to only the DirectAccess server or servers in your environment. And that the DirectAccess Client Settings GPO applies to only the DA client computers that you plan to enable in your network. The best practice here is to specify this GPO to only apply to a specific Active Directory security group so that you have full control over which computer accounts are in that group. I have seen some folks do it based only on the OU links and include whole OUs in the filtering for the clients GPO (foregoing the use of an AD group at all), but doing it this way makes it quite a bit more difficult to add or remove machines from the access list in the future. Requiring certificates as part of your DirectAccess tunnel authentication process is a good idea in any environment. It makes the solution more secure, and enables advanced functionality. The primary driver for most companies to require these certificates is the enablement of Windows 7 clients to connect via DirectAccess, but I suggest that anyone using DirectAccess in any capacity make use of these certs. They are simple to deploy, easy to configure, and give you some extra peace of mind that only computers who have a certificate issued directly to them from your own internal CA server are going to be able to connect through your DirectAccess entry point. Getting ready While the DirectAccess wizards themselves are run from the DirectAccess server, our work with this recipe is not. The Group Policy settings that we will be configuring are all accomplished within Active Directory, and we will be doing the work from a Domain Controller in our environment. How to do it... To pre-stage Group Policy Objects (GPOs) for use with DirectAccess: On your Domain Controller, launch the Group Policy Management Console . Expand Forest | Domains | Your Domain Name . There should be a listing here called Group Policy Object . Right-click on that and choose New . Name your new GPO something like DirectAccess Server Settings. Click on the new DirectAccess Server Settings GPO and it should open up automatically to the Scope tab. We need to adjust the Security Filtering section so that this GPO only applies to our DirectAccess server. This is a critical step for each GPO to ensure the settings that are going to be placed here do not get applied to the wrong computers. Remove Authenticated Users that is prepopulated in that list. The list should now be empty. Click the Add… button and search for the computer account of your DirectAccess server. Mine is called RA-01. By default this window will only search user accounts, so you will need to adjust Object Types to include Computers before it will allow you to add your server into this filtering list. Your Security Filtering list should now look like this:  Now click on the Details tab of your GPO. Change the GPO Status to be User configuration settings disabled . We do this because our GPO is only going to contain computer-level settings, nothing at the user level. The last thing to do is link your GPO to an appropriate container. Since we have Security Filtering enabled, our GPO is only ever going to apply its settings to the RA-01 server; however, without creating a link, the GPO will not even attempt to apply itself to anything. My RA-01 server is sitting inside the OU called Remote Access Servers . So I will right-click on my Remote Access Servers OU and choose Link an Existing GPO… .  Choose the new DirectAccess Server Settings from the list of available GPOs and click on the OK button. This creates the link and puts the GPO into action. Since there are not yet any settings inside the GPO, it won't actually make any changes on the server. The DirectAccess configuration wizards take care of populating the GPO with the settings that are needed. Now we simply need to rinse and repeat all of these steps to create another GPO, something like DirectAccess Client Settings . You want to set up the client settings GPO in the same way. Make sure that it is filtering to only the Active Directory Security Group that you created to contain your DirectAccess client computers. And make sure to link it to an appropriate container that will include those computer accounts. So maybe your clients GPO will look something like this:  How it works... Creating GPOs in Active Directory is a simple enough task, but it is critical that you configure the Links and Security Filtering correctly. If you do not take care to ensure that these DirectAccess connection settings are only going to apply to the machines that actually need the settings, you could create a world of trouble by internal servers getting remote access connection settings and cause them issues with connection while inside the network. Enhancing the security of DirectAccess by requiring certificate authentication When a DirectAccess client computer builds its IPsec tunnels back to the corporate network, it has the ability to require a certificate as part of that authentication process. In earlier versions of DirectAccess, the one in Server 2008 R2 and the one provided by Unified Access Gateway ( UAG ), these certificates were required in order to make DirectAccess work. Setting up the certificates really isn't a big deal at all; as long as there is a CA server in your network you are already prepared to issue the certs needed at no cost. Unfortunately, though, there must have been enough complaints back to Microsoft in order for them to make these certificates "recommended" instead of "required" and they created a new mechanism in Windows 8 and Server 2012 called KerberosProxy that can be used to authenticate the tunnels instead. This allows the DirectAccess tunnels to build without the computer certificate, making that authentication process less secure. I'm here to strongly recommend that you still utilize certificates in your installs! They are not difficult to set up, and using them makes your tunnel authentication stronger. Further, many of you may not have a choice and will still be required to install these certificates. Only simple DirectAccess scenarios that are all Windows 8 on the client side can get away with the shortcut method of foregoing certs. Anybody who still wants to connect Windows 7 via DirectAccess will need to use certificates on all of their client computers, both Windows 7 and Windows 8. In addition to Windows 7 access, anyone who intends to use the advanced features of DirectAccess such as load balancing, multi-site, or two-factor authentication will also need to utilize these certificates. With any of these scenarios, certificates become a requirement again, not a recommendation. In my experience, almost everyone still has Windows 7 clients that would benefit from being DirectAccess connected, and it's always a good idea to make your DA environment redundant by having load balanced servers. This further emphasizes the point that you should just set up certificate authentication right out of the gate, whether or not you need it initially. You might decide to make a change later that would require certificates and it would be easier to have them installed from the get-go rather than trying to incorporate them later into a running DA environment. Getting ready In order to distribute certificates, you will need a CA server running in your network. Once certificates are distributed to the appropriate places, the rest of our work will be accomplished from our Server 2012 R2 DirectAccess server. How to do it... Follow these steps to make use of certificates as part of the DirectAccess tunnel authentication process: The first thing that you need to do is distribute certificates to your DirectAccess servers and all DirectAccess client computers. The easiest way to do this is by using the built-in Computer template provided by default in a Windows CA server. If you desire to build a custom certificate template for this purpose, you can certainly do so. I recommend that you duplicate the Computer template and build it from there. Whenever I create a custom template for use with DirectAccess, I try to make sure that it meets the following criterias: The Subject Name of the certificate should match the Common Name of the computer (which is also the FQDN of the computer). The Subject Alternative Name ( SAN ) of the certificate should match the DNS Name of the computer (which is also the FQDN of the computer). The certificate should serve the Intended Purposes of both Client Authentication and Server Authentication . You can issue the certificates manually using Microsoft Management Console (MMC). Otherwise, you can lessen your hands-on administrative duties by enabling Autoenrollment. Now that we have certificates distributed to our DirectAccess clients and servers, log in to your primary DirectAccess server and open up the Remote Access Management Console . Click on Configuration in the top-left corner. You should now see steps 1 through 4 listed. Click Edit… listed under Step 2 . Now you can either click Next twice or click on the word Authentication to jump directly to the authentication screen. Check the box that says Use computer certificates . Now we have to specify the Certification Authority server that issued our client certificates. If you used an intermediary CA to issue your certs, make sure to check the appropriate checkbox. Otherwise, most of the time, certificates are issued from a root CA and in this case you would simply click on the Browse… button and look for your CA in the list. This screen is sometimes confusing because people expect to have to choose the certificate itself from the list. This is not the case. What you are actually choosing from this list is the Certificate Authority server that issued the certificates. Make any other appropriate selections on the Authentication screen. For example, many times when we require client certificates for authentication, it is because we have Windows 7 computers that we want to connect via DirectAccess. If that is the case for you, select the checkbox for Enable Windows 7 client computers to connect via DirectAccess .  How it works... Requiring certificates as part of your DirectAccess tunnel authentication process is a good idea in any environment. It makes the solution more secure, and enables advanced functionality. The primary driver for most companies to require these certificates is the enablement of Windows 7 clients to connect via DirectAccess, but I suggest that anyone using DirectAccess in any capacity make use of these certs. They are simple to deploy, easy to configure, and give you some extra peace of mind that only computers who have a certificate issued directly to them from your own internal CA server are going to be able to connect through your DirectAccess entry point. Building your Network Location Server (NLS) on its own system If you zipped through the default settings when configuring DirectAccess, or worse used the Getting Started Wizard, chances are that your Network Location Server ( NLS ) is running right on the DirectAccess server itself. This is not the recommended method for using NLS, it really should be running on a separate web server. In fact, if you later want to do something more advanced such as setting up load balanced DirectAccess servers, you're going to have to move NLS off onto a different server anyway. So you might as well do it right the first time. NLS is a very simple requirement, yet a critical one. It is just a website, it doesn't matter what content the site has, and it only has to run inside your network. Nothing has to be externally available. In fact, nothing should be externally available, because you only want this site being accessed internally. This NLS website is a large part of the mechanism by which DirectAccess client computers figure out when they are inside the office and when they are outside. If they can see the NLS website, they know they are inside the network and will disable DirectAccess name resolution, effectively turning off DA. If they do not see the NLS website, they will assume they are outside the corporate network and enable DirectAccess name resolution. There are two gotchas with setting up an NLS website: The first is that it must be HTTPS, so it does need a valid SSL certificate. Since this website is only running inside the network and being accessed from domain-joined computers, this SSL certificate can easily be one that has been issued from your internal CA server. So no cost associated there. The second catch that I have encountered a number of times is that for some reason the default IIS splash screen page doesn't make for a very good NLS website. If you set up a standard IIS web server and use the default site as NLS, sometimes it works to validate the connections and sometimes it doesn't. Given that, I always set up a specific site that I create myself, just to be on the safe side. So let's work together to follow the exact process I always take when setting up NLS websites in a new DirectAccess environment. Getting ready Our NLS website will be hosted on an IIS server we have that runs Server 2012 R2. Most of the work will be accomplished from this web server, but we will also be creating a DNS record and will utilize a Domain Controller for that task. How to do it... Let's work together to set up our new Network Location Server website: First decide on an internal DNS name to use for this website and set it up in DNS of your domain. I am going to use nls.mydomain.local and am creating a regular Host (A) record that points nls.mydomain.local at the IP address of my web server. Now log in to that web server and let's create some simple content for this new website. Create a new folder called C:NLS. Inside your new folder, create a new Default.htm file. Edit this file and throw some simple text in there. I usually say something like This is the NLS website used by DirectAccess. Please do not delete or modify me!.  Remember, this needs to be an HTTPS website, so before we try setting up the actual website, we should acquire the SSL certificate that we need to use with this site. Since this certificate is coming from my internal CA server, I'm going to open up MMC on my web server to accomplish this task. Once MMC is opened, snap-in the Certificates module. Make sure to choose Computer account and then Local computer when it prompts you for which certificate store you want to open. Expand Certificates (Local Computer) | Personal | Certificates . Right-click on this Certificates folder and choose All Tasks | Request New Certificate… . Click Next twice and you should see your list of certificate templates that are available on your internal CA server. If you do not see one that looks appropriate for requesting a website certificate, you may need to check over the settings on your CA server to make sure the correct templates are configured for issuing. My template is called Custom Web Server . Since this is a web server certificate, there is some additional information that I need to provide in my request in order to successfully issue a certificate. So I go ahead and click on that link that says More information is required to enroll for this certificate. Click here to configure settings. .  Drop-down the Subject name | Type menu and choose the option Common name . Enter a common name for our website into the Value field, which in my case is nls.mydomain.local. Click the Add button and your CN should move over to the right side of the screen like this:  Click on OK then click on the Enroll button. You should now have an SSL certificate sitting in your certificates store that can be used to authenticate traffic moving to our nls.mydomain.local name. Open up Internet Information Services (IIS) Manager , and browse to the Sites folder. Go ahead and remove the default website that IIS automatically set up, so that we can create our own NLS website without any fear of conflict. Click on the Add Website… action. Populate the information as shown in the following screenshot. Make sure to choose your own IP address and SSL certificate from the lists, of course:  Click the OK button and you now have an NLS website running successfully in your network. You should be able to open up a browser on a client computer sitting inside the network and successfully browse to https://nls.mydomain.local. How it works... In this recipe, we configured a basic Network Location Server website for use with our DirectAccess environment. This site will do exactly what we need it to when our DA client computers try to validate whether they are inside or outside the corporate network. While this recipe meets our requirements for NLS, and in fact puts us into a good practice of installing DirectAccess with NLS being hosted on its own web server, there is yet another step you could take to make it even better. Currently this web server is a single point of failure for NLS. If this web server goes down or has a problem, we would have DirectAccess client computers inside the office who would think they are outside, and they would have some major name resolution problems until we sorted out the NLS problem. Given that, it is a great idea to make NLS redundant. You could cluster servers together, use Microsoft Network Load Balancing ( NLB ), or even use some kind of hardware load balancer if you have one available in your network. This way you could run the same NLS website on multiple web servers and know that your clients will still work properly in the event of a web server failure. Summary This article encourages you to use Windows Server 2012 R2 as the connectivity platform that brings your remote computers into the corporate network. We discussed DirectAccess and VPN in this article. We also saw how to configure DirectAccess and VPN, and how to secure DirectAccess using certificate authentication. Resources for Article: Further resources on this subject: Cross-premise Connectivity [article] Setting Up and Managing E-mails and Batch Processing [article] Upgrading from Previous Versions [article]

In this article by Chitij Chauhan, author of the book PostgreSQL Cookbook, we will talk about various high availability and replication solutions, including some popular third-party replication tools such as Slony-I and Londiste. In this article, we will cover the following recipes: Setting up hot streaming replication Replication using Slony-I Replication using Londiste The important components for any production database is to achieve fault tolerance, 24/7 availability, and redundancy. It is for this purpose that we have different high availability and replication solutions available for PostgreSQL. From a business perspective, it is important to ensure 24/7 data availability in the event of a disaster situation or a database crash due to disk or hardware failure. In such situations, it becomes critical to ensure that a duplicate copy of the data is available on a different server or a different database, so that seamless failover can be achieved even when the primary server/database is unavailable. Setting up hot streaming replication In this recipe, we are going to set up a master-slave streaming replication. Getting ready For this exercise, you will need two Linux machines, each with the latest version of PostgreSQL installed. We will be using the following IP addresses for the master and slave servers: Master IP address: 192.168.0.4 Slave IP address: 192.168.0.5 Before you start with the master-slave streaming setup, it is important that the SSH connectivity between the master and slave is setup. How to do it... Perform the following sequence of steps to set up a master-slave streaming replication: First, we are going to create a user on the master, which will be used by the slave server to connect to the PostgreSQL database on the master server: psql -c "CREATE USER repuser REPLICATION LOGIN ENCRYPTED PASSWORD 'charlie';" Next, we will allow the replication user that was created in the previous step to allow access to the master PostgreSQL server. This is done by making the necessary changes as mentioned in the pg_hba.conf file: Vi pg_hba.conf host   replication   repuser   192.168.0.5/32   md5 In the next step, we are going to configure parameters in the postgresql.conf file. These parameters need to be set in order to get the streaming replication working: Vi /var/lib/pgsql/9.3/data/postgresql.conf listen_addresses = '*' wal_level = hot_standby max_wal_senders = 3 wal_keep_segments = 8 archive_mode = on       archive_command = 'cp %p /var/lib/pgsql/archive/%f && scp %p postgres@192.168.0.5:/var/lib/pgsql/archive/%f' checkpoint_segments = 8 Once the parameter changes have been made in the postgresql.conf file in the previous step, the next step will be to restart the PostgreSQL server on the master server, in order to let the changes take effect: pg_ctl -D /var/lib/pgsql/9.3/data restart Before the slave can replicate the master, we will need to give it the initial database to build off. For this purpose, we will make a base backup by copying the primary server's data directory to the standby. The rsync command needs to be run as a root user: psql -U postgres -h 192.168.0.4 -c "SELECT pg_start_backup('label', true)" rsync -a /var/lib/pgsql/9.3/data/ 192.168.0.5:/var/lib/pgsql/9.3/data/ --exclude postmaster.pid psql -U postgres -h 192.168.0.4 -c "SELECT pg_stop_backup()" Once the data directory, mentioned in the previous step, is populated, the next step is to enable the following parameter in the postgresql.conf file on the slave server: hot_standby = on The next step will be to copy the recovery.conf.sample file in the $PGDATA location on the slave server and then configure the following parameters: cp /usr/pgsql-9.3/share/recovery.conf.sample /var/lib/pgsql/9.3/data/recovery.conf standby_mode = on primary_conninfo = 'host=192.168.0.4 port=5432 user=repuser password=charlie' trigger_file = '/tmp/trigger.replication′ restore_command = 'cp /var/lib/pgsql/archive/%f "%p"' The next step will be to start the slave server: service postgresql-9.3 start Now that the above mentioned replication steps are set up, we will test for replication. On the master server, log in and issue the following SQL commands: psql -h 192.168.0.4 -d postgres -U postgres -W postgres=# create database test;   postgres=# c test;   test=# create table testtable ( testint int, testchar varchar(40) );   CREATE TABLE test=# insert into testtable values ( 1, 'What A Sight.' ); INSERT 0 1 On the slave server, we will now check whether the newly created database and the corresponding table, created in the previous step, are replicated: psql -h 192.168.0.5 -d test -U postgres -W test=# select * from testtable; testint | testchar ---------+--------------------------- 1 | What A Sight. (1 row) How it works... The following is the explanation for the steps performed in the preceding section. In the initial step of the preceding section, we create a user called repuser, which will be used by the slave server to make a connection to the primary server. In the second step of the preceding section, we make the necessary changes in the pg_hba.conf file to allow the master server to be accessed by the slave server using the repuser user ID that was created in step 1. We then make the necessary parameter changes on the master in step 3 of the preceding section to configure a streaming replication. The following is a description of these parameters: listen_addresses: This parameter is used to provide the IP address associated with the interface that you want to have PostgreSQL listen to. A value of * indicates all available IP addresses. wal_level: This parameter determines the level of WAL logging done. Specify hot_standby for streaming replication. wal_keep_segments: This parameter specifies the number of 16 MB WAL files to be retained in the pg_xlog directory. The rule of thumb is that more such files might be required to handle a large checkpoint. archive_mode: Setting this parameter enables completed WAL segments to be sent to the archive storage. archive_command: This parameter is basically a shell command that is executed whenever a WAL segment is completed. In our case, we are basically copying the file to the local machine and then using the secure copy command to send it across to the slave. max_wal_senders: This parameter specifies the total number of concurrent connections allowed from the slave servers. checkpoint_segments: This parameter specifies the maximum number of logfile segments between automatic WAL checkpoints. Once the necessary configuration changes have been made on the master server, we then restart the PostgreSQL server on the master in order to let the new configuration changes take effect. This is done in step 4 of the preceding section. In step 5 of the preceding section, we are basically building the slave by copying the primary server's data directory to the slave. Now, with the data directory available on the slave, the next step is to configure it. We will now make the necessary parameter replication related parameter changes on the slave in the postgresql.conf directory on the slave server. We set the following parameters on the slave: hot_standby: This parameter determines whether you can connect and run queries when the server is in the archive recovery or standby mode. In the next step, we are configuring the recovery.conf file. This is required to be set up so that the slave can start receiving logs from the master. The parameters explained next are configured in the recovery.conf file on the slave. standby_mode: This parameter, when enabled, causes PostgreSQL to work as a standby in a replication configuration. primary_conninfo: This parameter specifies the connection information used by the slave to connect to the master. For our scenario, our master server is set as 192.168.0.4 on port 5432 and we are using the repuser userid with the password charlie to make a connection to the master. Remember that repuser was the userid which was created in the initial step of the preceding section for this purpose, that is, connecting to the master from the slave. trigger_file: When a slave is configured as a standby, it will continue to restore the XLOG records from the master. The trigger_file parameter specifies what is used to trigger a slave, in order to switch over its duties from standby and take over as master or primary server. At this stage, the slave is fully configured now and we can start the slave server; then, the replication process begins. This is shown in step 8 of the preceding section. In steps 9 and 10 of the preceding section, we are simply testing our replication. We first begin by creating a test database, then we log in to the test database and create a table by the name testtable, and then we begin inserting some records into the testtable table. Now, our purpose is to see whether these changes are replicated across the slave. To test this, we log in to the slave on the test database and then query the records from the testtable table, as seen in step 10 of the preceding section. The final result that we see is that all the records that are changed/inserted on the primary server are visible on the slave. This completes our streaming replication's setup and configuration. Replication using Slony-I Here, we are going to set up replication using Slony-I. We will be setting up the replication of table data between two databases on the same server. Getting ready The steps performed in this recipe are carried out on a CentOS Version 6 machine. It is also important to remove the directives related to hot streaming replication prior to setting up replication using Slony-I. We will first need to install Slony-I. The following steps need to be performed in order to install Slony-I: First, go to http://slony.info/downloads/2.2/source/ and download the given software. Once you have downloaded the Slony-I software, the next step is to unzip the .tar file and then go the newly created directory. Before doing this, please ensure that you have the postgresql-devel package for the corresponding PostgreSQL version installed before you install Slony-I: tar xvfj slony1-2.2.3.tar.bz2  cd slony1-2.2.3 In the next step, we are going to configure, compile, and build the software: ./configure --with-pgconfigdir=/usr/pgsql-9.3/bin/ make make install How to do it... You need to perform the following sequence of steps, in order to replicate data between two tables using Slony-I replication: First, start the PostgreSQL server if you have not already started it: pg_ctl -D $PGDATA start In the next step, we will be creating two databases, test1 and test2, which will be used as the source and target databases respectively: createdb test1 createdb test2 In the next step, we will create the t_test table on the source database, test1, and insert some records into it: psql -d test1 test1=# create table t_test (id numeric primary key, name varchar);   test1=# insert into t_test values(1,'A'),(2,'B'), (3,'C'); We will now set up the target database by copying the table definitions from the test1 source database: pg_dump -s -p 5432 -h localhost test1 | psql -h localhost -p 5432 test2 We will now connect to the target database, test2, and verify that there is no data in the tables of the test2 database: psql -d test2 test2=# select * from t_test; We will now set up a slonik script for the master-slave, that is source/target, setup. In this scenario, since we are replicating between two different databases on the same server, the only different connection string option will be the database name: cd /usr/pgsql-9.3/bin vi init_master.slonik   #!/bin/sh cluster name = mycluster; node 1 admin conninfo = 'dbname=test1 host=localhost port=5432 user=postgres password=postgres'; node 2 admin conninfo = 'dbname=test2 host=localhost port=5432 user=postgres password=postgres'; init cluster ( id=1); create set (id=1, origin=1); set add table(set id=1, origin=1, id=1, fully qualified name = 'public.t_test'); store node (id=2, event node = 1); store path (server=1, client=2, conninfo='dbname=test1 host=localhost port=5432 user=postgres password=postgres'); store path (server=2, client=1, conninfo='dbname=test2 host=localhost port=5432 user=postgres password=postgres'); store listen (origin=1, provider = 1, receiver = 2);  store listen (origin=2, provider = 2, receiver = 1); We will now create a slonik script for subscription to the slave, that is, target: cd /usr/pgsql-9.3/bin vi init_slave.slonik #!/bin/sh cluster name = mycluster; node 1 admin conninfo = 'dbname=test1 host=localhost port=5432 user=postgres password=postgres'; node 2 admin conninfo = 'dbname=test2 host=localhost port=5432 user=postgres password=postgres'; subscribe set ( id = 1, provider = 1, receiver = 2, forward = no); We will now run the init_master.slonik script created in step 6 and run this on the master, as follows: cd /usr/pgsql-9.3/bin   slonik init_master.slonik We will now run the init_slave.slonik script created in step 7 and run this on the slave, that is, target: cd /usr/pgsql-9.3/bin   slonik init_slave.slonik In the next step, we will start the master slon daemon: nohup slon mycluster "dbname=test1 host=localhost port=5432 user=postgres password=postgres" & In the next step, we will start the slave slon daemon: nohup slon mycluster "dbname=test2 host=localhost port=5432 user=postgres password=postgres" & Next, we will connect to the master, that is, the test1 source database, and insert some records in the t_test table: psql -d test1 test1=# insert into t_test values (5,'E'); We will now test for the replication by logging on to the slave, that is, the test2 target database, and see whether the inserted records in the t_test table are visible: psql -d test2   test2=# select * from t_test; id | name ----+------ 1 | A 2 | B 3 | C 5 | E (4 rows) How it works... We will now discuss the steps performed in the preceding section: In step 1, we first start the PostgreSQL server if it is not already started. In step 2, we create two databases, namely test1 and test2, that will serve as our source (master) and target (slave) databases. In step 3, we log in to the test1 source database, create a t_test table, and insert some records into the table. In step 4, we set up the target database, test2, by copying the table definitions present in the source database and loading them into test2 using the pg_dump utility. In step 5, we log in to the target database, test2, and verify that there are no records present in the t_test table because in step 4, we only extracted the table definitions into the test2 database from the test1 database. In step 6, we set up a slonik script for the master-slave replication setup. In the init_master.slonik file, we first define the cluster name as mycluster. We then define the nodes in the cluster. Each node will have a number associated with a connection string, which contains database connection information. The node entry is defined both for the source and target databases. The store_path commands are necessary, so that each node knows how to communicate with the other. In step 7, we set up a slonik script for the subscription of the slave, that is, the test2 target database. Once again, the script contains information such as the cluster name and the node entries that are designated a unique number related to connection string information. It also contains a subscriber set. In step 8, we run the init_master.slonik file on the master. Similarly, in step 9, we run the init_slave.slonik file on the slave. In step 10, we start the master slon daemon. In step 11, we start the slave slon daemon. The subsequent steps, 12 and 13, are used to test for replication. For this purpose, in step 12 of the preceding section, we first log in to the test1 source database and insert some records into the t_test table. To check whether the newly inserted records have been replicated in the target database, test2, we log in to the test2 database in step 13. The result set obtained from the output of the query confirms that the changed/inserted records on the t_test table in the test1 database are successfully replicated across the target database, test2. For more information on Slony-I replication, go to http://slony.info/documentation/tutorial.html. There's more... If you are using Slony-I for replication between two different servers, in addition to the steps mentioned in the How to do it… section, you will also have to enable authentication information in the pg_hba.conf file existing on both the source and target servers. For example, let's assume that the source server's IP is 192.168.16.44 and the target server's IP is 192.168.16.56 and we are using a user named super to replicate the data. If this is the situation, then in the source server's pg_hba.conf file, we will have to enter the information, as follows: host         postgres         super     192.168.16.44/32           md5 Similarly, in the target server's pg_hba.conf file, we will have to enter the authentication information, as follows: host         postgres         super     192.168.16.56/32           md5 Also, in the shell scripts that were used for Slony-I, wherever the connection information for the host is localhost that entry will need to be replaced by the source and target server's IP addresses. Replication using Londiste In this recipe, we are going to show you how to replicate data using Londiste. Getting ready For this setup, we are using the same host CentOS Linux machine to replicate data between two databases. This can also be set up using two separate Linux machines running on VMware, VirtualBox, or any other virtualization software. It is assumed that the latest version of PostgreSQL, version 9.3, is installed. We used CentOS Version 6 as the Linux operating system for this exercise. To set up Londiste replication on the Linux machine, perform the following steps: Go to http://pgfoundry.org/projects/skytools/ and download the latest version of Skytools 3.2, that is, tarball skytools-3.2.tar.gz. Extract the tarball file, as follows: tar -xvzf skytools-3.2.tar.gz Go to the new location and build and compile the software: cd skytools-3.2 ./configure --prefix=/var/lib/pgsql/9.3/Sky –with-pgconfig=/usr/pgsql-9.3/bin/pg_config   make   make install Also, set the PYTHONPATH environment variable, as shown here. Alternatively, you can also set it in the .bash_profile script: export PYTHONPATH=/opt/PostgreSQL/9.2/Sky/lib64/python2.6/site-packages/ How to do it... We are going to perform the following sequence of steps to set up replication between two different databases using Londiste. First, create the two databases between which replication has to occur: createdb node1 createdb node2 Populate the node1 database with data using the pgbench utility: pgbench -i -s 2 -F 80 node1 Add any primary key and foreign keys to the tables in the node1 database that are needed for replication. Create the following .sql file and add the following lines to it: Vi /tmp/prepare_pgbenchdb_for_londiste.sql -- add primary key to history table ALTER TABLE pgbench_history ADD COLUMN hid SERIAL PRIMARY KEY;   -- add foreign keys ALTER TABLE pgbench_tellers ADD CONSTRAINT pgbench_tellers_branches_fk FOREIGN KEY(bid) REFERENCES pgbench_branches; ALTER TABLE pgbench_accounts ADD CONSTRAINT pgbench_accounts_branches_fk FOREIGN KEY(bid) REFERENCES pgbench_branches; ALTER TABLE pgbench_history ADD CONSTRAINT pgbench_history_branches_fk FOREIGN KEY(bid) REFERENCES pgbench_branches; ALTER TABLE pgbench_history ADD CONSTRAINT pgbench_history_tellers_fk FOREIGN KEY(tid) REFERENCES pgbench_tellers; ALTER TABLE pgbench_history ADD CONSTRAINT pgbench_history_accounts_fk FOREIGN KEY(aid) REFERENCES pgbench_accounts; We will now load the .sql file created in the previous step and load it into the database: psql node1 -f /tmp/prepare_pgbenchdb_for_londiste.sql We will now populate the node2 database with table definitions from the tables in the node1 database: pg_dump -s -t 'pgbench*' node1 > /tmp/tables.sql psql -f /tmp/tables.sql node2 Now starts the process of replication. We will first create the londiste.ini configuration file with the following parameters in order to set up the root node for the source database, node1: Vi londiste.ini   [londiste3] job_name = first_table db = dbname=node1 queue_name = replication_queue logfile = /home/postgres/log/londiste.log pidfile = /home/postgres/pid/londiste.pid In the next step, we are going to use the londiste.ini configuration file created in the previous step to set up the root node for the node1 database, as shown here: [postgres@localhost bin]$ ./londiste3 londiste3.ini create-root node1 dbname=node1   2014-12-09 18:54:34,723 2335 WARNING No host= in public connect string, bad idea 2014-12-09 18:54:35,210 2335 INFO plpgsql is installed 2014-12-09 18:54:35,217 2335 INFO pgq is installed 2014-12-09 18:54:35,225 2335 INFO pgq.get_batch_cursor is installed 2014-12-09 18:54:35,227 2335 INFO pgq_ext is installed 2014-12-09 18:54:35,228 2335 INFO pgq_node is installed 2014-12-09 18:54:35,230 2335 INFO londiste is installed 2014-12-09 18:54:35,232 2335 INFO londiste.global_add_table is installed 2014-12-09 18:54:35,281 2335 INFO Initializing node 2014-12-09 18:54:35,285 2335 INFO Location registered 2014-12-09 18:54:35,447 2335 INFO Node "node1" initialized for queue "replication_queue" with type "root" 2014-12-09 18:54:35,465 2335 INFO Don We will now run the worker daemon for the root node: [postgres@localhost bin]$ ./londiste3 londiste3.ini worker 2014-12-09 18:55:17,008 2342 INFO Consumer uptodate = 1 In the next step, we will create a slave.ini configuration file in order to make a leaf node for the node2 target database: Vi slave.ini [londiste3] job_name = first_table_slave db = dbname=node2 queue_name = replication_queue logfile = /home/postgres/log/londiste_slave.log pidfile = /home/postgres/pid/londiste_slave.pid We will now initialize the node in the target database: ./londiste3 slave.ini create-leaf node2 dbname=node2 –provider=dbname=node1 2014-12-09 18:57:22,769 2408 WARNING No host= in public connect string, bad idea 2014-12-09 18:57:22,778 2408 INFO plpgsql is installed 2014-12-09 18:57:22,778 2408 INFO Installing pgq 2014-12-09 18:57:22,778 2408 INFO   Reading from /var/lib/pgsql/9.3/Sky/share/skytools3/pgq.sql 2014-12-09 18:57:23,211 2408 INFO pgq.get_batch_cursor is installed 2014-12-09 18:57:23,212 2408 INFO Installing pgq_ext 2014-12-09 18:57:23,213 2408 INFO   Reading from /var/lib/pgsql/9.3/Sky/share/skytools3/pgq_ext.sql 2014-12-09 18:57:23,454 2408 INFO Installing pgq_node 2014-12-09 18:57:23,455 2408 INFO   Reading from /var/lib/pgsql/9.3/Sky/share/skytools3/pgq_node.sql 2014-12-09 18:57:23,729 2408 INFO Installing londiste 2014-12-09 18:57:23,730 2408 INFO   Reading from /var/lib/pgsql/9.3/Sky/share/skytools3/londiste.sql 2014-12-09 18:57:24,391 2408 INFO londiste.global_add_table is installed 2014-12-09 18:57:24,575 2408 INFO Initializing node 2014-12-09 18:57:24,705 2408 INFO Location registered 2014-12-09 18:57:24,715 2408 INFO Location registered 2014-12-09 18:57:24,744 2408 INFO Subscriber registered: node2 2014-12-09 18:57:24,748 2408 INFO Location registered 2014-12-09 18:57:24,750 2408 INFO Location registered 2014-12-09 18:57:24,757 2408 INFO Node "node2" initialized for queue "replication_queue" with type "leaf" 2014-12-09 18:57:24,761 2408 INFO Done We will now launch the worker daemon for the target database, that is, node2: [postgres@localhost bin]$ ./londiste3 slave.ini worker 2014-12-09 18:58:53,411 2423 INFO Consumer uptodate = 1 We will now create the configuration file, that is pgqd.ini, for the ticker daemon: vi pgqd.ini   [pgqd] logfile = /home/postgres/log/pgqd.log pidfile = /home/postgres/pid/pgqd.pid Using the configuration file created in the previous step, we will launch the ticker daemon: [postgres@localhost bin]$ ./pgqd pgqd.ini 2014-12-09 19:05:56.843 2542 LOG Starting pgqd 3.2 2014-12-09 19:05:56.844 2542 LOG auto-detecting dbs ... 2014-12-09 19:05:57.257 2542 LOG node1: pgq version ok: 3.2 2014-12-09 19:05:58.130 2542 LOG node2: pgq version ok: 3.2 We will now add all the tables to the replication on the root node: [postgres@localhost bin]$ ./londiste3 londiste3.ini add-table --all 2014-12-09 19:07:26,064 2614 INFO Table added: public.pgbench_accounts 2014-12-09 19:07:26,161 2614 INFO Table added: public.pgbench_branches 2014-12-09 19:07:26,238 2614 INFO Table added: public.pgbench_history 2014-12-09 19:07:26,287 2614 INFO Table added: public.pgbench_tellers Similarly, add all the tables to the replication on the leaf node: [postgres@localhost bin]$ ./londiste3 slave.ini add-table –all We will now generate some traffic on the node1 source database: pgbench -T 10 -c 5 node1 We will now use the compare utility available with the londiste3 command to check the tables in both the nodes; that is, both the source database (node1) and destination database (node2) have the same amount of data: [postgres@localhost bin]$ ./londiste3 slave.ini compare   2014-12-09 19:26:16,421 2982 INFO Checking if node1 can be used for copy 2014-12-09 19:26:16,424 2982 INFO Node node1 seems good source, using it 2014-12-09 19:26:16,425 2982 INFO public.pgbench_accounts: Using node node1 as provider 2014-12-09 19:26:16,441 2982 INFO Provider: node1 (root) 2014-12-09 19:26:16,446 2982 INFO Locking public.pgbench_accounts 2014-12-09 19:26:16,447 2982 INFO Syncing public.pgbench_accounts 2014-12-09 19:26:18,975 2982 INFO Counting public.pgbench_accounts 2014-12-09 19:26:19,401 2982 INFO srcdb: 200000 rows, checksum=167607238449 2014-12-09 19:26:19,706 2982 INFO dstdb: 200000 rows, checksum=167607238449 2014-12-09 19:26:19,715 2982 INFO Checking if node1 can be used for copy 2014-12-09 19:26:19,716 2982 INFO Node node1 seems good source, using it 2014-12-09 19:26:19,716 2982 INFO public.pgbench_branches: Using node node1 as provider 2014-12-09 19:26:19,730 2982 INFO Provider: node1 (root) 2014-12-09 19:26:19,734 2982 INFO Locking public.pgbench_branches 2014-12-09 19:26:19,734 2982 INFO Syncing public.pgbench_branches 2014-12-09 19:26:22,772 2982 INFO Counting public.pgbench_branches 2014-12-09 19:26:22,804 2982 INFO srcdb: 2 rows, checksum=-3078609798 2014-12-09 19:26:22,812 2982 INFO dstdb: 2 rows, checksum=-3078609798 2014-12-09 19:26:22,866 2982 INFO Checking if node1 can be used for copy 2014-12-09 19:26:22,877 2982 INFO Node node1 seems good source, using it 2014-12-09 19:26:22,878 2982 INFO public.pgbench_history: Using node node1 as provider 2014-12-09 19:26:22,919 2982 INFO Provider: node1 (root) 2014-12-09 19:26:22,931 2982 INFO Locking public.pgbench_history 2014-12-09 19:26:22,932 2982 INFO Syncing public.pgbench_history 2014-12-09 19:26:25,963 2982 INFO Counting public.pgbench_history 2014-12-09 19:26:26,008 2982 INFO srcdb: 715 rows, checksum=9467587272 2014-12-09 19:26:26,020 2982 INFO dstdb: 715 rows, checksum=9467587272 2014-12-09 19:26:26,056 2982 INFO Checking if node1 can be used for copy 2014-12-09 19:26:26,063 2982 INFO Node node1 seems good source, using it 2014-12-09 19:26:26,064 2982 INFO public.pgbench_tellers: Using node node1 as provider 2014-12-09 19:26:26,100 2982 INFO Provider: node1 (root) 2014-12-09 19:26:26,108 2982 INFO Locking public.pgbench_tellers 2014-12-09 19:26:26,109 2982 INFO Syncing public.pgbench_tellers 2014-12-09 19:26:29,144 2982 INFO Counting public.pgbench_tellers 2014-12-09 19:26:29,176 2982 INFO srcdb: 20 rows, checksum=4814381032 2014-12-09 19:26:29,182 2982 INFO dstdb: 20 rows, checksum=4814381032 How it works... The following is an explanation of the steps performed in the preceding section: Initially, in step 1, we create two databases, that is node1 and node2, that are used as the source and target databases, respectively, from a replication perspective. In step 2, we populate the node1 database using the pgbench utility. In step 3 of the preceding section, we add and define the respective primary key and foreign key relationships on different tables and put these DDL commands in a .sql file. In step 4, we execute these DDL commands stated in step 3 on the node1 database; thus, in this way, we force the primary key and foreign key definitions on the tables in the pgbench schema in the node1 database. In step 5, we extract the table definitions from the tables in the pgbench schema in the node1 database and load these definitions in the node2 database. We will now discuss steps 6 to 8 of the preceding section. In step 6, we create the configuration file, which is then used in step 7 to create the root node for the node1 source database. In step 8, we will launch the worker daemon for the root node. Regarding the entries mentioned in the configuration file in step 6, we first define a job that must have a name, so that distinguished processes can be easily identified. Then, we define a connect string with information to connect to the source database, that is node1, and then we define the name of the replication queue involved. Finally, we define the location of the log and pid files. We will now discuss steps 9 to 11 of the preceding section. In step 9, we define the configuration file, which is then used in step 10 to create the leaf node for the target database, that is node2. In step 11, we launch the worker daemon for the leaf node. The entries in the configuration file in step 9 contain the job_name connect string in order to connect to the target database, that is node2, the name of the replication queue involved, and the location of log and pid involved. The key part in step 11 is played by the slave, that is the target database—to find the master or provider, that is source database node1. We will now talk about steps 12 and 13 of the preceding section. In step 12, we define the ticker configuration, with the help of which we launch the ticker process mentioned in step 13. Once the ticker daemon has started successfully, we have all the components and processes setup and needed for replication; however, we have not yet defined what the system needs to replicate. In step 14 and 15, we define the tables to the replication that is set on both the source and target databases, that is node1 and node2, respectively. Finally, we will talk about steps 16 and 17 of the preceding section. Here, at this stage, we are testing the replication that was set up between the node1 source database and the node2 target database. In step 16, we generate some traffic on the node1 source database by running pgbench with five parallel database connections and generating traffic for 10 seconds. In step 17, we check whether the tables on both the source and target databases have the same data. For this purpose, we use the compare command on the provider and subscriber nodes and then count and checksum the rows on both sides. A partial output from the preceding section tells you that the data has been successfully replicated between all the tables that are part of the replication set up between the node1 source database and the node2 destination database, as the count and checksum of rows for all the tables on the source and target destination databases are matching: 2014-12-09 19:26:18,975 2982 INFO Counting public.pgbench_accounts 2014-12-09 19:26:19,401 2982 INFO srcdb: 200000 rows, checksum=167607238449 2014-12-09 19:26:19,706 2982 INFO dstdb: 200000 rows, checksum=167607238449   2014-12-09 19:26:22,772 2982 INFO Counting public.pgbench_branches 2014-12-09 19:26:22,804 2982 INFO srcdb: 2 rows, checksum=-3078609798 2014-12-09 19:26:22,812 2982 INFO dstdb: 2 rows, checksum=-3078609798   2014-12-09 19:26:25,963 2982 INFO Counting public.pgbench_history 2014-12-09 19:26:26,008 2982 INFO srcdb: 715 rows, checksum=9467587272 2014-12-09 19:26:26,020 2982 INFO dstdb: 715 rows, checksum=9467587272   2014-12-09 19:26:29,144 2982 INFO Counting public.pgbench_tellers 2014-12-09 19:26:29,176 2982 INFO srcdb: 20 rows, checksum=4814381032 2014-12-09 19:26:29,182 2982 INFO dstdb: 20 rows, checksum=4814381032 Summary This article demonstrates the high availability and replication concepts in PostgreSQL. After reading this chapter, you will be able to implement high availability and replication options using different techniques including streaming replication, Slony-I replication and replication using Longdiste. Resources for Article: Further resources on this subject: Running a PostgreSQL Database Server [article] Securing the WAL Stream [article] Recursive queries [article]

In this article, by the authors, Fabrizio Volpe, Alessio Giombini, Lasse Nordvik Wedø, and António Vargas of the book, Lync Server Cookbook, we will cover the following recipes: Introducing Lync Online Administering with the Lync Admin Center Using Lync Online Remote PowerShell Using Lync Online cmdlets Introducing Lync in a hybrid scenario Planning and configuring a hybrid deployment Moving users to the cloud Moving users back on-premises Debugging Lync Online issues (For more resources related to this topic, see here.) Introducing Lync Online Lync Online is part of the Office 365 offer and provides online users with the same Instant Messaging (IM), presence, and conferencing features that we would expect from an on-premises deployment of Lync Server 2013. Enterprise Voice, however, is not available on Office 365 tenants (or at least, it is available only with limitations regarding both specific Office 365 plans and geographical locations). There is no doubt that forthcoming versions of Lync and Office 365 will add what is needed to also support all the Enterprise Voice features in the cloud. Right now, the best that we are able to achieve is to move workloads, homing a part of our Lync users (the ones with no telephony requirements) in Office 365, while the remaining Lync users are homed on-premises. These solutions might be interesting for several reasons, including the fact that we can avoid the costs of expanding our existing on-premises resources by moving a part of our Lync-enabled users to Office 365. The previously mentioned configuration, which involves different kinds of Lync tenants, is called a hybrid deployment of Lync, and we will see how to configure it and move our users from online to on-premises and vice versa. In this Article, every time we talk about Lync Online and Office 365, we will assume that we have already configured an Office tenant. Administering with the Lync Admin Center Lync Online provides the Lync Admin Center (LAC), a dedicated control panel, to manage Lync settings. To open it, access the Office 365 portal and select Service settings, Lync, and Manage settings in the Lync admin center, as shown in the following screenshot: LAC, if you compare it with the on-premises Lync Control Panel (or with the Lync Management Shell), offers few options. For example, it is not possible to create or delete users directly inside Lync. We will see some of the tasks we are able to perform in LAC, and then, we will move to the (more powerful) Remote PowerShell. There is an alternative path to open LAC. From the Office 365 portal, navigate to Users & Groups | Active Users. Select a user, after which you will see a Quick Steps area with an Edit Lync Properties link that will open the user-editable part of LAC. How to do it... LAC is divided into five areas: users, organization, dial-in conferencing, meeting invitation, and tools, as you can see in the following screenshot: The Users panel will show us the configuration of the Lync Online enabled users. It is possible to modify the settings with the Edit option (the small pencil icon on the right): I have tried to summarize all the available options (inside the general, external communications, and dial-in conferencing tabs) in the following screenshot: Some of the user's settings are worth a mention; in the General tab, we have the following:    The Record Conversations and meetings option enables the Start recording option in the Lync client    The Allow anonymous attendees to dial-out option controls whether the anonymous users that are dialing-in to a conference are required to call the conferencing service directly or are authorized for callback    The For compliance, turn off non-archived features option disables Lync features that are not recorded by In-Place Hold for Exchange When you place an Exchange 2013 mailbox on In-Place Hold or Litigation Hold, the Microsoft Lync 2013 content (instant messaging conversations and files shared in an online meeting) is archived in the mailbox. In the dial-in conferencing tab, we have the configuration required for dial-in conferencing. The provider's drop-down menu shows a list of third parties that are able to deliver this kind of feature. The Organization tab manages privacy for presence information, push services, and external access (the equivalent of the Lync federation on-premises). If you enable external access, we will have the option to turn on Skype federation, as we can see in the following screenshot: The Dial-In Conferencing option is dedicated to the configuration of the external providers. The Meeting Invitation option allows the user to customize the Lync Meeting invitation. The Tools options offer a collection of troubleshooting resources. See also For details about Exchange In-Place Hold, see the TechNet post In-Place Hold and Litigation Hold at http://technet.microsoft.com/en-us/library/ff637980(v=exchg.150).aspx. Using Lync Online Remote PowerShell The possibility to manage Lync using Remote PowerShell on a distant deployment has been available since Lync 2010. This feature has always required a direct connection from the management station to the Remote Lync, and a series of steps that is not always simple to set up. Lync Online supports Remote PowerShell using a dedicated (64-bit only) PowerShell module, the Lync Online Connector. It is used to manage online users, and it is interesting because there are many settings and automation options that are available only through PowerShell. Getting ready Lync Online Connector requires one of the following operating systems: Windows 7 (with Service Pack 1), Windows Server 2008 R2, Windows Server 2012, Windows Server 2012 R2, Windows 8, or Windows 8.1. At least PowerShell 3.0 is needed. To check it, we can use the $PSVersionTable variable. The result will be like the one in the following screenshot (taken on Windows 8.1, which uses PowerShell 4.0): How to do it... Download Windows PowerShell Module for Lync Online from the Microsoft site at http://www.microsoft.com/en-us/download/details.aspx?id=39366 and install it. It is useful to store our Office 365 credentials in an object (it is possible to launch the cmdlets at step 3 anyway, and we will be required with the Office 365 administrator credentials, but using this method, we will have to insert the authentication information again every time it is required). We can use the $credential = Get-Credential cmdlet in a PowerShell session. We will be prompted for our username and password for Lync Online, as shown in the following screenshot: To use the Online Connector, open a PowerShell session and use the New-CsOnlineSession cmdlet. One of the ways to start a remote PowerShell session is $session = New-CsOnlineSession -Credential $credential. Now, we need to import the session that we have created with Lync Online inside PowerShell, with the Import-PSSession $session cmdlet. A temporary Windows PowerShell module will be created, which contains all the Lync Online cmdlets. The name of the temporary module will be similar to the one we can see in the following screenshot: Now, we will have the cmdlets of the Lync Online module loaded in memory, in addition to any command that we already have available in PowerShell. How it works... The feature is based on a PowerShell module, the LyncOnlineConnector, shown in the following screenshot: It contains only two cmdlets, the Set-WinRMNetworkDelayMS and New-CsOnlineSession cmdlets. The latter will load the required cmdlets in memory. As we have seen in the previous steps, the Online Connector adds the Lync Online PowerShell cmdlets to the ones already available. This is something we will use when talking about hybrid deployments, where we will start from the Lync Management Shell and then import the module for Lync Online. It is a good habit to verify (and close) your previous remote sessions. This can be done by selecting a specific session (using Get-PSSession and then pointing to a specific session with the Remove-PSSession statement) or closing all the existing ones with the Get-PSSession | Remove-PSSession cmdlet. In the previous versions of the module, Microsoft Online Services Sign-In Assistant was required. This prerequisite was removed from the latest version. There's more... There are some checks that we are able to perform when using the PowerShell module for Lync Online. By launching the New-CsOnlineSession cmdlet with the –verbose switch, we will see all the messages related to the opening of the session. The result should be similar to the one shown in the following screenshot: Another verification comes from the Get-Command -Module tmp_gffrkflr.ufz command, where the module name (in this example, tmp_gffrkflr.ufz) is the temporary module we saw during the Import-PSSession step. The output of the command will show all the Lync Online cmdlets that we have loaded in memory. The Import-PSSession cmdlet imports all commands except the ones that have the same name of a cmdlet that already exists in the current PowerShell session. To overwrite the existing cmdlets, we can use the -AllowClobber parameter. See also During the introduction of this section, we also discussed the possibility to administer on-premises, remote Lync Server 2013 deployment with a remote PowerShell session. John Weber has written a great post about it in his blog Lync 2013 Remote Admin with PowerShell at http://tsoorad.blogspot.it/2013/10/lync-2013-remote-admin-with-powershell.html, which is helpful if you want to use the previously mentioned feature. Using Lync Online cmdlets In the previous recipe, we outlined the steps required to establish a remote PowerShell session with Lync Online. We have less than 50 cmdlets, as shown in the result of the Get-Command -Module command in the following screenshot: Some of them are specific for Lync Online, such as the following: Get-CsAudioConferencingProvider Get-CsOnlineUser Get-CsTenant Get-CsTenantFederationConfiguration Get-CsTenantHybridConfiguration Get-CsTenantLicensingConfiguration Get-CsTenantPublicProvider New-CsEdgeAllowAllKnownDomains New-CsEdgeAllowList New-CsEdgeDomainPattern Set-CsTenantFederationConfiguration Set-CsTenantHybridConfiguration Set-CsTenantPublicProvider Update-CsTenantMeetingUrl All the remaining cmdlets can be used either with Lync Online or with the on-premises version of Lync Server 2013. We will see the use of some of the previously mentioned cmdlets. How to do it... The Get-CsTenant cmdlet will list Lync Online tenants configured for use in our organization. The output of the command includes information such as the preferred language, registrar pool, domains, and assigned plan. The Get-CsTenantHybridConfiguration cmdlet gathers information about the hybrid configuration of Lync. Management of the federation capability for Lync Online (the feature that enables Instant Messaging and Presence information exchange with users of other domains) is based on the allowed domain and blocked domain lists, as we can see in the organization and external communications screen of LAC, shown in the following screenshot: There are similar ways to manage federation from the Lync Online PowerShell, but it required to put together different statements as follows:     We can use an accept all domains excluding the ones in the exceptions list approach. To do this, we have put the New-CsEdgeAllowAllKnownDomains cmdlet inside a variable. Then, we can use the Set-CsTenantFederationConfiguration cmdlet to allow all the domains (except the ones in the block list) for one of our domains on a tenant. We can use the example on TechNet (http://technet.microsoft.com/en-us/library/jj994088.aspx) and integrate it with Get-CsTenant.     If we prefer, we can use a block all domains but permit the ones in the allow list approach. It is required to define a domain name (pattern) for every domain to allow the New-CsEdgeDomainPattern cmdlet, and each one of them will be saved in a variable. Then, the New-CsEdgeAllowList cmdlet will create a list of allowed domains from the variables. Finally, the Set-CsTenantFederationConfiguration cmdlet will be used. The domain we will work on will be (again) cc3b6a4e-3b6b-4ad4-90be-6faa45d05642. The example on Technet (http://technet.microsoft.com/en-us/library/jj994023.aspx) will be used: $x = New-CsEdgeDomainPattern -Domain "contoso.com" $y = New-CsEdgeDomainPattern -Domain "fabrikam.com" $newAllowList = New-CsEdgeAllowList -AllowedDomain $x,$y Set-CsTenantFederationConfiguration -Tenant " cc3b6a4e-3b6b-4ad4-90be-6faa45d05642" -AllowedDomains $newAllowList The Get-CsOnlineUser cmdlet provides information about users enabled on Office 365. The result will show both users synced with Active Directory and users homed in the cloud. The command supports filters to limit the output; for example, the Get-CsOnlineUser -identity fab will gather information about the user that has alias = fab. This is an account synced from the on-premises Directory Services, so the value of the DirSyncEnabled parameter will be True. See also All the cmdlets of the Remote PowerShell for Lync Online are listed in the TechNet post Lync Online cmdlets at http://technet.microsoft.com/en-us/library/jj994021.aspx. This is the main source of details on the single statement. Introducing Lync in a hybrid scenario In a Lync hybrid deployment, we have the following: User accounts and related information homed in the on-premises Directory Services and replicated to Office 365. A part of our Lync users that consume on-premises resources and a part of them that use online (Office 365 / Lync Online) resources. The same (public) domain name used both online and on-premises (Lync-split DNS). Other Office 365 services and integration with other applications available to all our users, irrespective of where their Lync is provisioned. One way to define Lync hybrid configuration is by using an on-premises Lync deployment federated with an Office 365 / Lync Online tenant subscription. While it is not a perfect explanation, it gives us an idea of the scenario we are talking about. Not all the features of Lync Server 2013 (especially the ones related to Enterprise Voice) are available to Lync Online users. The previously mentioned motivations, along with others (due to company policies, compliance requirements, and so on), might recommend a hybrid deployment of Lync as the best available solution. What we have to clarify now is how to make those users on different deployments talk to each other, see each other's presence status, and so on. What we will see in this section is a high-level overview of the required steps. The Planning and configuring a hybrid deployment recipe will provide more details about the individual steps. The list of steps here is the one required to configure a hybrid deployment, starting from Lync on-premises. In the following sections, we will also see the opposite scenario (with our initial deployment in the cloud). How to do it... It is required to have an available Office 365 tenant configuration. Our subscription has to include Lync Online. We have to configure an Active Directory Federation Services (AD FS) server in our domain and make it available to the Internet using a public FQDN and an SSL certificate released from a third-party certification authority. Office 365 must be enabled to synchronize with our company's Directory Services, using Active Directory Sync. Our Office 365 tenant must be federated. The last step is to configure Lync for a hybrid deployment. There's more... One of the requirements for a hybrid distribution of Lync is an on-premises deployment of Lync Server 2013 or Lync Server 2010. For Lync Server 2010, it is required to have the latest available updates installed, both on the Front Ends and on the Edge servers. It is also required to have the Lync Server 2013 administrative tools installed on a separate server. More details about supported configuration are available on the TechNet post Planning for Lync Server 2013 hybrid deployments at http://technet.microsoft.com/en-us/library/jj205403.aspx. DNS SRV records for hybrid deployments, _sipfederationtls._tcp.<domain> and _sip._tls.<domain>, should point to the on-premises deployment. The lyncdiscover. <domain> record will point to the FQDN of the on-premises reverse proxy server. The _sip._tls. <domain> SRV record will resolve to the public IP of the Access Edge service of Lync on-premises. Depending on the kind of service we are using for Lync, Exchange, and SharePoint, only a part of the features related to the integration with the additional services might be available. For example, skills search is available only if we are using Lync and SharePoint on-premises. The following TechNet post Supported Lync Server 2013 hybrid configurations at http://technet.microsoft.com/en-us/library/jj945633.aspx offers a matrix of features / service deployment combinations. See also Interesting information about Lync Hybrid configuration is presented in sessions available on Channel9 and coming from the Lync Conference 2014 (Lync Online Hybrid Deep Dive at http://channel9.msdn.com/Events/Lync-Conference/Lync-Conference-2014/ONLI302) and from TechEd North America 2014 (Microsoft Lync Online Hybrid Deep Dive at http://channel9.msdn.com/Events/TechEd/NorthAmerica/2014/OFC-B341#fbid=). Planning and configuring a hybrid deployment The planning phase for a hybrid deployment starts from a simple consideration: do we have an on-premises deployment of Lync Server? If the previously mentioned scenario is true, do we want to move users to the cloud or vice versa? Although the first situation is by far the most common, we have to also consider the case in which we have our first deployment in the cloud. How to do it... This step is all that is required for the scenario that starts from Lync Online. We have to completely deploy our Lync on-premises. Establish a remote PowerShell session with Office 365. Use the shared SIP address cmdlet Set-CsTenantFederationConfiguration -SharedSipAddressSpace $True to enable Office 365 to use a Shared Session Initiation Protocol (SIP) address space with our on-premises deployment. To verify this, we can use the Get-CsTenantFederationConfiguration command. The SharedSipAddressSpace value should be set to True. All the following steps are for the scenario that starts from the on-premises Lync deployment. After we have subscribed with a tenant, the first step is to add the public domain we use for our Lync users to Office 365 (so that we can split it on the two deployments). To access the Office 365 portal, select Domains. The next step is Specify a domain name and confirm ownership. We will be required to type a domain name. If our domain is hosted on some specific providers (such as GoDaddy), the verification process can be automated, or we have to proceed manually. The process requires to add one DNS record (TXT or MX), like the ones shown in the following screenshot: If we need to check our Office 365 and on-premises deployments before continuing with the hybrid deployment, we can use the Setup Assistant for Office 365. The tool is available inside the Office 365 portal, but we have to launch it from a domain-joined computer (the login must be performed with the domain administrative credentials). In the Setup menu, we have a Quick Start and an Extend Your Setup option (we have to select the second one). The process can continue installing an app or without software installation, as shown in the following screenshot: The app (which makes the assessment of the existing deployment easier) is installed by selecting Next in the previous screen (it requires at least Windows 7 with Service Pack 1, .NET Framework 3.5, and PowerShell 2.0). Synchronization with the on-premises Active Directory is required. This last step federates Lync Server 2013 with Lync Online to allow communication between our users. The first cmdlet to use is Set-CSAccessEdgeConfiguration -AllowOutsideUsers 1 -AllowFederatedUsers 1 -UseDnsSrvRouting -EnablePartnerDiscovery 1. Note that the -EnablePartnerDiscovery parameter is required. Setting it to 1 enables automatic discovery of federated partner domains. It is possible to set it to 0. The second required cmdlet is New-CSHostingProvider -Identity LyncOnline -ProxyFqdn "sipfed.online.lync.com" -Enabled $true -EnabledSharedAddressSpace $true -HostsOCSUsers $true –VerificationLevel UseSourceVerification -IsLocal $false -AutodiscoverUrl https://webdir.online.lync.com/Autodiscover/AutodiscoverService.svc/root. The result of the commands is shown in the following screenshot: If Lync Online is already defined, we have to use the Set- CSHostingProvider cmdlet, or we can remove it (Remove-CsHostingProvider -Identity LyncOnline) and then create it using the previously mentioned cmdlet. There's more... In the Lync hybrid scenario, users created in the on-premises directory are replicated to the cloud, while users generated in the cloud will not be replicated on-premises. Lync Online users are managed using the Office 365 portal, while the users on-premises are managed using the usual tools (Lync Control Panel and Lync Management Shell). Moving users to the cloud By moving users from Lync on-premises to the cloud, we will lose some of the parameters. The operation requires the Lync administrative tools and the PowerShell module for Lync Online to be installed on the same computer. If we install the module for Lync Online before the administrative tools for Lync 2013 Server, the OCSCore.msi file overwrites the LyncOnlineConnector.ps1 file, and New-CsOnlineSession will require a -TargetServer parameter. In this situation, we have to reinstall the Lync Online module (see the following post on the Microsoft support site at http://support.microsoft.com/kb/2955287). Getting ready Remember that to move the user to Lync Online, they must be enabled for both Lync Server on-premises and Lync Online (so we have to assign the user a license for Lync Online by using the Office 365 portal). Users with no assigned licenses will show the error Move-CsUser : HostedMigration fault: Error=(507), Description=(User must has an assigned license to use Lync Online. For more details, refer to the Microsoft support site at http://support.microsoft.com/kb/2829501. How to do it... Open a new Lync Management Shell session and launch the remote session on Office 365 with the cmdlets' sequence we saw earlier. We have to add the –AllowClobber parameter so that the Lync Online module's cmdlets are able to overwrite the corresponding Lync Management Shell cmdlets: $credential = Get-Credential $session = New-CsOnlineSession -Credential $credential Import-PSSession $session -AllowClobber Open the Lync Admin Center (as we have seen in the dedicated section) by going to Service settings | Lync | Manage settings in the Lync Admin Center, and copy the first part of the URL, for example, https://admin0e.online.lync.com. Add the following string to the previous URL /HostedMigration/hostedmigrationservice.svc (in our example, the result will be https://admin0a.online.lync.com/HostedMigration/hostedmigrationservice.svc). The following cmdlet will move users from Lync on-premises to Lync Online. The required parameters are the identity of the Lync user and the URL that we prepared in step 2. The user identity is fabrizio.volpe@absoluteuc.biz: Move-CsUser -Identity fabrizio.volpe@absoluteuc.biz –Target sipfed.online.lync.com -Credential $creds -HostedMigrationOverrideUrl https://admin0e.online.lync.com/HostedMigration/hostedmigrationservice.sVc Usually, we are required to insert (again) the Office 365 administrative credentials, after which we will receive a warning about the fact that we are moving our user to a different version of the service, like the one in the following screenshot: See the There's more... section of this recipe for details about user information that is migrated to Lync Online. We are able to quickly verify whether the user has moved to Lync Online by using the Get-CsUser | fl DisplayName,HostingProvider,RegistrarPool,SipAddress command. On-premises HostingProvider is equal to SRV: and RegistrarPool is madhatter.wonderland.lab (the name of the internal Lync Front End). Lync Online values are HostingProvider : sipfed.online.lync.com, and leave RegistrarPool empty, as shown in the following screenshot (the user Fabrizio is homed on-premises, while the user Fabrizio volpe is homed on the cloud): There's more... If we plan to move more than one user, we have to add a selection and pipe it before the cmdlet we have already used, removing the –identity parameter. For example, to move all users from an Organizational Unit (OU), (for example, the LyncUsers in the Wonderland.Lab domain) to Lync Online, we can use Get-CsUser -OU "OU=LyncUsers,DC=wonderland,DC=lab"| Move-CsUser -Target sipfed.online.lync.com -Credential $creds -HostedMigrationOverrideUrl https://admin0e.online.lync.com/HostedMigration/hostedmigrationservice.sVc. We are also able to move users based on a parameter to match using the Get-CsUser –Filter cmdlet. As we mentioned earlier, not all the user information is migrated to Lync Online. Migration contact list, groups, and access control lists are migrated, while meetings, contents, and schedules are lost. We can use the Lync Meeting Update Tool to update the meeting links (which have changed when our user's home server has changed) and automatically send updated meeting invitations to participants. There is a 64-bit version (http://www.microsoft.com/en-us/download/details.aspx?id=41656) and a 32-bit version (http://www.microsoft.com/en-us/download/details.aspx?id=41657) of the previously mentioned tool. Moving users back on-premises It is possible to move back users that have been moved from the on-premises Lync deployment to the cloud, and it is also possible to move on-premises users that have been defined and enabled directly in Office 365. In the latter scenario, it is important to create the user also in the on-premises domain (Directory Service). How to do it… The Lync Online user must be created in the Active Directory (for example, I will define the BornOnCloud user that already exists in Office 365). The user must be enabled in the on-premises Lync deployment, for example, using the Lync Management Shell with the following cmdlet: Enable-CsUser -Identity "BornOnCloud" -SipAddress "SIP:BornOnCloud@absoluteuc.biz" -HostingProviderProxyFqdn "sipfed.online.lync.com" Sync the Directory Services. Now, we have to save our Office 365 administrative credentials in a $cred = Get-Credential variable and then move the user from Lync Online to the on-premises Front End using the Lync Management Shell (the -HostedMigrationOverrideURL parameter has the same value that we used in the previous section): Move-CsUser -Identity BornOnCloud@absoluteuc.biz -Target madhatter.wonderland.lab -Credential $cred -HostedMigrationOverrideURL https://admin0e.online.lync.com/HostedMigration/hostedmigrationservice.svc The Get-CsUser | fl DisplayName,HostingProvider,RegistrarPool,SipAddress cmdlet is used to verify whether the user has moved as expected. See also Guy Bachar has published an interesting post on his blog Moving Users back to Lync on-premises from Lync Online (http://guybachar.wordpress.com/2014/03/31/moving-users-back-to-lync-on-premises-from-lync-online/), where he shows how he solved some errors related to the user motion by modifying the HostedMigrationOverrideUrl parameter. Debugging Lync Online issues Getting ready When moving from an on-premises solution to a cloud tenant, the first aspect we have to accept is that we will not have the same level of control on the deployment we had before. The tools we will list are helpful in resolving issues related to Lync Online, but the level of understanding on an issue they give to a system administrator is not the same we have with tools such as Snooper or OCSLogger. Knowing this, the more users we will move to the cloud, the more we will have to use the online instruments. How to do it… The Set up Lync Online external communications site on Microsoft Support (http://support.microsoft.com/common/survey.aspx?scid=sw;en;3592&showpage=1) is a guided walk-through that helps in setting up communication between our Lync Online users and external domains. The tool provides guidelines to assist in the setup of Lync Online for small to enterprise businesses. As you can see in the following screenshot, every single task is well explained: The Remote Connectivity Analyzer (RCA) (https://testconnectivity.microsoft.com/) is an outstanding tool to troubleshoot both Lync on-premises and Lync Online. The web page includes tests to analyze common errors and misconfigurations related to Microsoft services such as Exchange, Lync, and Office 365. To test different scenarios, it is necessary to use various network protocols and ports. If we are working on a firewall-protected network, using the RCA, we are also able to test services that are not directly available to us. For Lync Online, there are some tests that are especially interesting; in the Office 365 tab, the Office 365 General Tests section includes the Office 365 Lync Domain Name Server (DNS) Connectivity Test and the Office 365 Single Sign-On Test, as shown in the following screenshot: The Single Sign-On test is really useful in a scenario. The test requires our domain username and password, both synced with the on-premises Directory Services. The steps include searching the FQDN of our AD FS server on an Internet DNS, verifying the certificate and connectivity, and then validating the token that contains the credentials. The Client tab offers to download the Microsoft Connectivity Analyzer Tool and the Microsoft Lync Connectivity Analyzer Tool, which we will see in the following two dedicated steps: The Microsoft Connectivity Analyzer Tool makes many of the tests we see in the RCA available on our desktop. The list of prerequisites is provided in the article Microsoft Connectivity Analyzer Tool (http://technet.microsoft.com/library/jj851141(v=exchg.80).aspx), and includes Windows Vista/Windows 2008 or later versions of the operating system, .NET Framework 4.5, and an Internet browser, such as Internet Explorer, Chrome, or Firefox. For the Lync tests, a 64-bit operating system is mandatory, and the UCMA runtime 4.0 is also required (it is part of Lync Server 2013 setup, and is also available for download at http://www.microsoft.com/en-us/download/details.aspx?id=34992). The tools propose ways to solve different issues, and then, they run the same tests available on the RCA site. We are able to save the results in an HTML file. The Microsoft Lync Connectivity Analyzer Tool is dedicated to troubleshooting the clients for mobile devices (the Lync Windows Store app and Lync apps). It tests all the required configurations, including autodiscover and webticket services. The 32-bit version is available at http://www.microsoft.com/en-us/download/details.aspx?id=36536, while the 64-bit version can be downloaded from http://www.microsoft.com/en-us/download/details.aspx?id=36535. .NET Framework 4.5 is required. The tool itself requires a few configuration parameters; we have to insert the user information that we usually add in the Lync app, and we have to use a couple of drop-down menus to describe the scenario we are testing (on-premises or Internet, and the kind of client we are going to test). The Show drop-down menu enables us to look not only at a summary of the test results but also at the detailed information. The detailed view includes all the information and requests sent and received during the test, with the FQDN included in the answer ticket from our services, and so on, as shown in the following screenshot: The Troubleshooting Lync Online sign-in post is a support page, available in two different versions (admins and users), and is a walk-through to help admins (or users) to troubleshoot login issues. The admin version is available at http://support.microsoft.com/common/survey.aspx?scid=sw;en;3695&showpage=1, while the user version is available at http://support.microsoft.com/common/survey.aspx?scid=sw;en;3719&showpage=1. Based on our answers to the different scenario questions, the site will propose to information or solution steps. The following screenshot is part of the resolution for the log-I issues of a company that has an enterprise subscription with a custom domain: The Office 365 portal includes some information to help us monitor our Lync subscription. In the Service Health menu, navigate to Service Health; we have a list of all the incidents and service issues of the past days. In the Reports menu, we have statistics about our Office 365 consumption, including Lync. In the following screenshot, we can see the previously mentioned pages: There's more... One interesting aspect of the Microsoft Lync Connectivity Analyzer Tool that we have seen is that it enables testing for on-premises or Office 365 accounts (both testing from inside our network and from the Internet). The previously mentioned capability makes it a great tool to troubleshoot the configuration for Lync on the mobile devices that we have deployed in our internal network. This setup is usually complex, including hair-pinning and split DNS, so the diagnostic is important to quickly find misconfigured services. See also The Troubleshooting Lync Sign-in Errors (Administrators) page on Office.com at http://office.microsoft.com/en-001/communicator-help/troubleshooting-lync-sign-in-errors-administrators-HA102759022.aspx contains a list of messages related to sign-in errors with a suggested solution or a link to additional external resources. Summary In this article, we have learned about managing Lync 2013 and Lync Online and using Lync Online Remote PowerShell and Lync Online cmdlets. Resources for Article: Further resources on this subject: Adding Dialogs [article] Innovation of Communication and Information Technologies [article] Choosing Lync 2013 Clients [article]

In this article by Peter De Tender, the author of Mastering Hyper-V, we will talk about the building blocks for creating your virtualization platform through Hyper-V. We need to clearly define a detailed list of required server hardware, storage hardware, physical and virtual machine operating systems, and anything else we need to be able to build our future virtualization platform. These components are known as the Hyper-V building blocks, and we describe each one of them in the following sections. (For more resources related to this topic, see here.) Physical server hardware One of the first important components when building a virtualization platform is the physical server hardware. One of the key elements to check is the Microsoft certified hardware and software supportability and compatibility list. This list gives a detailed overview of all tested and certified server brands, server types, and their corresponding configuration components. While it is not a requirement to use this kind of machine, we can only recommend it, based on our own experience. Imagine you have a performance issue with one of your applications running inside a VM, being hosted on non-supported hardware, using non-supported physical NICs, and you're not getting decent support from your IT partner or Microsoft on that specific platform, as the hardware is not supported. The landing page for this compatibility list is http://www.windowsservercatalog.com. After checking the compatibility of the server hardware and software, you need to find out which system resources are available for Hyper-V. The following table shows the maximum scaling possibilities for different components of the Hyper-V platform (the original source is Microsoft TechNet Library article at http://technet.microsoft.com/en-us/library/jj680093.aspx.) System Resource Maximum number   Windows 2008 R2 Windows Server 2012 (R2) Host Logical processors on hardware 64 320 Physical memory 1 TB 4 TB Virtual processors per host 512 1,024 Virtual machine Virtual processors per virtual machine 4 64 Memory per virtual machine 64 GB 1 TB Active virtual machines 384 1,024 Virtual disk size 2 TB 64 TB Cluster Nodes 16 64 Virtual machines 1,000 4,000 Physical storage hardware Next to the physical server component, another vital part of the virtualization environment is the storage hardware. In the Hyper-V platform, multiple kinds of storage are supported, that is DAS, NAS, and/or SAN: Direct Attached Storage (DAS): This is directly connected to the server (think of disk which is located inside the server chassis). Network Attached Storage (NAS): This is the storage provided via the network and presented to the Hyper-V server or virtual machines as file shares. This disk type is file-based access. Server 2012 and 2012 R2 make use of SMB 3.0 as file-sharing protocol, which allows us to use plain file shares as virtual machine storage location Storage Area Network (SAN): This is also network-based storage, but relies on block-based access. The volumes are presented as local disks to the host. Popular protocols within SAN environments are iSCSI and Fibre Channel. The key point of consideration when sizing your disk infrastructure is providing enough storage, at the best performance available, and preferably high availability as well. Depending on the virtual machine's required resources, the disk subsystem can be based on high-performant / expensive SSD disks (solid-state drives), performant / medium-priced SAS disks (serial attached SCSI), or slower but cheaper SATA (serial ATA) disks. Or it could even be a combination of all these types. Although a bit outside of Hyper-V as such, one technology that is configured and used a lot in combination with Hyper-V Server 2012 R2, is Storage Spaces. Storage Spaces is new as of Server 2012, and can be considered as a storage virtualization subsystem. Storage Spaces are disk volumes built on top of physical storage pools, which is in fact just a bunch of physical disks (JBOD). A very important point to note is that the aforementioned network-based SAN and NAS storage solutions cannot be a part of Storage Spaces, as it is only configurable for DAS storage. The following schema diagram provides a good overview of the Storage Spaces topology, possibilities, and features: Physical network devices It's easy to understand that your virtual platform is dependent on your physical network devices such as physical (core) switches and physical NICs in the Hyper-V hosts. When configuring Hyper-V, there are a few configurations to keep into consideration. NIC Teaming NIC Teaming is the configuration of multiple physical network interface cards into a single team, mainly used for high availability or higher bandwidth purposes. NIC Teaming as such is no technology of Hyper-V, but Hyper-V can make good use of this operating system feature. When configuring a NIC team, the physical network cards are bundled and presented to the host OS as one or more virtual network adapter(s). Within Hyper-V, two basic sets of algorithms exist where you can choose from during the configuration of Hyper-V networking: Switch-independent mode: In this configuration, the teaming is configured regardless of the switches to which the host is connected. The main advantage in this configuration is the fact the teaming can be configured to use multiple switches (for example, two NICs in the host are connected to switch 1 and 2 NICs are configured to use switch 2). Switch-dependent mode: In this configuration, the underlying switch is part of the teaming configuration; this automatically requires all NICs in the team to be connected to the same switch. NIC Teaming is managed through the Server Manager / NIC Teaming interface or by using PowerShell cmdlets. Depending on your server hardware and brand, the vendor might provide you with specific configuration software to achieve the same. For example, the HP Proliant series of servers allows for HP Team configuration, which is managed by using a specific HP Team tool. Network virtualization Within Hyper-V 2012 R2, network virtualization not only refers to the virtual networking connections that are used by the virtual machines but also refers to the technology that allows for true network isolation to the different networks in which virtual machines operate. This feature set is very important for hosting providers, who run different virtual machines for their customers in an isolated network. You have to make sure that there is no connection possible between the virtual machines from customer A and the virtual machines from customer B. That's exactly the main purpose of network virtualization. Another possible way of configuring network segmentation is by using VLANs. However, this also requires VLAN configuration to be done on the physical switches, where the described network virtualization completely runs inside the virtual network switch of Hyper-V. Server editions and licensing The last component that comprises the Hyper-V building blocks is the server editions and licensing of the physical and virtual machines operating system. Summary In this article, we looked at the various building blocks for building a virtualization platform using Hyper-V.

In this article by Ken Cherven, author of the book, Mastering Gephi Network Visualization, we will learn how to select the most appropriate types based on the characteristics of your network data. (For more resources related to this topic, see here.) Assessing your graphing needs Now that you have seen the broad array of available layout options and a bit of their respective capabilities, it is time to step back and reconsider what story you want to tell through the data. As you have just seen, there are many directions you can take within Gephi, and there is no absolute standard for right or wrong in your layout selection. However, there are some simple guidelines that can be followed to help narrow the choices. If you are experienced with Gephi or another network analysis tool, you might wish to dive directly into the next section and begin assessing each layout type using your very own dataset; I will not attempt to convince you otherwise. This is a great way to quickly learn the basics of every layout offering and can be a great experience. On the other hand, if you wish to take a more focused approach, I will offer you a brief checklist of considerations that might help to narrow your pool of layout candidates, allowing you to spend more time with those likely to provide the best results. Think of this as akin to shopping for clothes —you could try on every type of clothing on the rack, or you can quickly narrow your choices based on certain criteria—body type, complementary colors, preferred styles, and so on. So let's have a look at some of the basic points to consider while shopping for an appropriate layout: What is the goal of your analysis? Are you attempting to show complementarity within the network, as in the relationships between nodes or sets of nodes, or is the goal to display divisions within the data? Does geography play a critical role in the network? Perhaps you are seeking to sort or rank networks based on some attribute within the data. Each of these factors can play a determining role in which layout algorithm is best for your specific network. Is the dataset small, medium, or large? Admittedly, this is a subjective criteria, but we can put some general bounds around these definitions. In my mind, if the number of nodes is measured in tens or dozens, then this is likely a small dataset that can be easily displayed in a conventional space—the Gephi workspace window or a simple letter-sized paper for a printed version. If, however, the nodes run into the hundreds, we are now moving away from a very simple network and potentially reducing the number of practical layout options. When the number of nodes in a network moves into the thousands and beyond, we have what can practically be considered a large network, at least for display considerations. With datasets of this scope, additional display considerations come into play, such as judicious use of filters, layers, and interactivity. How densely connected is the network? In our previous example using the power grid data, we had a fairly large dataset numbering in the thousands, but one that was not highly connected, at least as compared to social networks. In that case, we might have an easier time selecting and applying an effective layout, while the highly connected nature of social networks presents an additional challenge. Does the network exhibit certain measurable behaviors such as clustering and homophily? In some cases, we might not know this until the network has been visually and programmatically analyzed, but in others we might already know that the data is likely to cluster based on certain attributes that influence the network structure, including geographic proximity, alumni networks, professional associations, and a host of other possibilities. Knowing some of these in advance might help guide us either toward or away from specific layout types. Will the network be displayed on a single level, or will it be bipartite or multipartite? In this case some networks might be hierarchical, with individuals (for example) linking only to an organization, and not to other individuals in the network. There are many instances where we will wish to present hierarchical networks in this fashion. This could be used to display corporate structures, academic hierarchies, player to team relationships, and so on, and requires some different considerations than networks without this structure. Does the data have a temporal element? In simple terms, will the story be told more effectively by viewing network changes over time? This can be very effective in showing diffusion/contagion patterns, random growth, and simple shifts in behavior within a network, for example—were Thomas and James friends at T1, but no longer so at T3 (where T equals time)? If our data has a specific time element, this leads to identify layouts that will best display these changes and tell an effective story. Will the network be interactive on the user end, or will it be static? This can ultimately lead to a different layout selection when users have the ability to navigate a network via the Web. You might have additional considerations, including the speed of the layout algorithm, but the preceding list should help you to narrow the list of practical layouts, allowing you to test the remaining candidates. Actual example – the Miles Davis network Let's walk through a process following the preceding guidelines, and applying them to a project previously created by me. This will help us migrate from the theoretical constructs above to a practical application of many of these principles. The project I'll use as our example traces the studio albums recorded by the legendary jazz trumpeter, Miles Davis—48 in all. Here are the details for this project, following the above progression. Analysis goal The goal of the analysis was to inform viewers, who might or might not be jazz fans, about the remarkable, far reaching recording legacy of Miles Davis. Since the career of Davis moved through many stages, he crossed paths with and employed an incredible number of artists across a diverse range of instruments that ranged far beyond the normal jazz instrumentation. Therefore, part of the goal of the analysis was to expose viewers to this great diversity, and give them the ability to see changes and patterns within the scope of his career. Dataset parameters The dataset in this case is not insignificant—while 48 albums would represent a small network if left on its own, we know from the data that there are typically at least four musicians per recording, and often far more, numbering into the 20s in some cases. Many of the musicians are represented on multiple recordings, but there is still a multiplicative impact on the size of the network, which turns out to have about 350 nodes. While this certainly doesn't rival the enormous datasets often seen in social networks, it is large enough that we need to be thoughtful about the layout and how users will interact with the project. Here is a look at some of the underlying data for the nodes: Miles Davis nodes Notice that the nodes are a combination of an individual musician and a specific instrument, since so many of these musicians play a second (or even third) instrument. The data is then grouped by instrument, which allows you to partition and custom color the data. Now, the following figure illustrates a partial view of the edge's data: Miles Davis data edges In the preceding screenshot, we see only album level connections, with Miles Davis as the source and each album as the target, although the edges are left undirected. If we move further into the edge's data, we can see how the network is structured a bit more clearly: Miles Davis data edge details This data shows some of the musician level connections to specific recordings, as well as the instrument played on that album. This completes the basic structure of the network, as each musician will have an edge connecting them to any and all albums they played on. So this gives us a basic understanding of how the data will be represented in the network—Miles at the core, all albums at a second level, followed by every contributing musician at a tertiary level. Network density We have all seen many highly connected networks with edges crossing between nodes or groups within a graph that become virtually impenetrable for the viewer. Fortunately, this was not a major concern with this network, given its relatively modest size, but it could still play a role in the final layout selection. As always, the goal is to provide clarity and understanding, regardless of the relative size of the network, so minimizing visual clutter is always a priority. Network behaviors Examining the network behaviors can be an interesting exercise, as it often leads us to findings that were not necessarily anticipated. In the case of this project, we know from viewing the data that Miles played with certain musicians on a frequent basis, but would then often play with an entirely new group during his next phase, before switching yet again to a completely unrelated group of musicians. In other words, there were multiple aggregations of musicians who only occasionally intersected with one another. This is very nearly a proxy for homophily, with distinct clusters connected to each other through a single node (Miles Davis in this case) or perhaps a small subset of network members who act as bridges between various clusters. Based on this knowledge, we would anticipate a highly clustered network with a significant level of connectedness within a given cluster, and a limited set of connections between clusters. The next decision to make was how best to display this network. Network display We just saw the underlying data structure, which had a bipartite nature to it, with each musician connecting to one or more albums, rather than to other musicians. Given this type of network, we want to select a layout that eases our ability to see not only the connections between Miles Davis and each recording, but also from each album to all of the participating musicians. This will require a layout that provides enough empty space to make for clear viewing, but also one that manages to combine this with a minimal number of edge crossings. Remember that many of these musicians played on multiple recordings, so they must be positioned in proximity to several albums at the same time, without adding to a cluttered look. After testing several layouts, some of which simply didn't work effectively with the above two needs, I settled on the ARF algorithm for its visual clarity to display this particular network. The ability to see patterns within the network, even prior to adding interactivity, is a plus; if the network passes that test, it should be very effective once users interact with the information. Temporal elements Another interesting aspect of the network that could have been utilized was the timeline for the recordings. With more than four decades of recordings, this could have provided a wealth of information about changes over time in the musicians' network and instrumentation on each album. This element was not highlighted, but it does make its presence felt in the final network, with albums from one period with a consistent cast of musicians occupying one sector of the graph, while other types of albums with many infrequently used musicians land in another area. Interactivity The final decision was whether to make the network interactive, giving users the ability to learn more through self-navigation of the graph. This was considered important from the very start, so that the viewers could see not only the body of work represented by the 48 recordings, but also the evolution of which musicians were involved, as well as shining a light on the wide array of instruments used as Miles' career evolved. After each of these considerations was evaluated, and through a period of testing the network using multiple layouts, I settled on the ARF force-directed layout coupled with the Sigma.js plugin for interactivity. Here's a look at the final output, which includes options using the Sigma.js plugin: The Miles Davis network graph The link to the project can be found at http://visual-baseball.com/gephi/jazz/miles_davis/. I hope this example helps to generate some ideas or at least opens up the possibilities for what Gephi is capable of creating, and that the process illustrated earlier helps to provide at least a foundation for your own work. Summary In this article, you learned how to select the most appropriate types based on the characteristics of your network data. Resources for Article: Further resources on this subject: Data visualization [article] Visualization as a Tool to Understand Data [article] Creating Network Graphs with Gephi [article]

In this article by Mark Vasilkov, author of the book, Kivy Blueprints, we will emulate the Modern UI by using the grid structure and scalable vector icons and develop a sound recorder for the Android platform using Android Java classes. (For more resources related to this topic, see here.) Kivy apps usually end up being cross-platform, mainly because the Kivy framework itself supports a wide range of target platforms. In this write-up, however, we're building an app that will be single-platform. This gives us an opportunity to rely on platform-specific bindings that provide extended functionality. The need for such bindings arises from the fact that the input/output capabilities of a pure Kivy program are limited to those that are present on all platforms. This amounts to a tiny fraction of what a common computer system, such as a smartphone or a laptop, can actually do. Comparison of features Let's take a look at the API surface of a modern mobile device (let's assume it's running Android). We'll split everything in two parts: things that are supported directly by Python and/or Kivy and things that aren't. The following are features that are directly available in Python or Kivy: Hardware-accelerated graphics Touchscreen input with optional multitouch Sound playback (at the time of writing, this feature is available only from the file on the disk) Networking, given the Internet connectivity is present The following are the features that aren't supported or require an external library: Modem, support for voice calls, and SMS Use of built-in cameras for filming videos and taking pictures Use of a built-in microphone to record sound Cloud storage for application data associated with a user account Bluetooth and other near-field networking features Location services and GPS Fingerprinting and other biometric security Motion sensors, that is, accelerometer and gyroscope Screen brightness control Vibration and other forms of haptic feedback Battery charge level For most entries in the "not supported" list, different Python libraries are already present to fill the gap, such as audiostream for a low-level sound recording, and Plyer that handles many platform-specific tasks. So, it's not like these features are completely unavailable to your application; realistically, the challenge is that these bits of functionality are insanely fragmented across different platforms (or even consecutive versions of the same platform, for example, Android); thus, you end up writing platform-specific, not portable code anyway. As you can see from the preceding comparison, a lot of functionality is available on Android and only partially covered by an existing Python or Kivy API. There is a huge untamed potential in using platform-specific features in your applications. This is not a limitation, but an opportunity. Shortly, you will learn how to utilize any Android API from Python code, allowing your Kivy application to do practically anything. Another advantage of narrowing the scope of your app to only a small selection of systems is that there are whole new classes of programs that can function (or even make sense) only on a mobile device with fitting hardware specifications. These include augmented reality apps, gyroscope-controlled games, panoramic cameras, and so on. Introducing Pyjnius To harness the full power of our chosen platform, we're going to use a platform-specific API, which happens to be in Java and is thus primarily Java oriented. We are going to build a sound recorder app, similar to the apps commonly found in Android and iOS, albeit more simplistic. Unlike pure Kivy apps, the underlying Android API certainly provides us with ways of recording sound programmatically. The rest of the article will cover this little recorder program throughout its development to illustrate the Python-Java interoperability using the excellent Pyjnius library, another great project made by Kivy developers. The concept we chose—sound recording and playback—is deliberately simple so as to outline the features of such interoperation without too much distraction caused by the sheer complexity of a subject and abundant implementation details. The source code of Pyjnius, together with the reference manual and some examples, can be found in the official repository at https://github.com/kivy/pyjnius. Modern UI While we're at it, let's build a user interface that resembles the Windows Phone home screen. This concept, basically a grid of colored rectangles (tiles) of various sizes, was known as Metro UI at some point in time but was later renamed to Modern UI due to trademark issues. Irrespective of the name, this is how it looks. This will give you an idea of what we'll be aiming at during the course of this app's development: Design inspiration – a Windows Phone home screen with tiles Obviously, we aren't going to replicate it as is; we will make something that resembles the depicted user interface. The following list pretty much summarizes the distinctive features we're after: Everything is aligned to a rectangular grid UI elements are styled using the streamlined, flat design—tiles use bright, solid colors and there are no shadows or rounded corners Tiles that are considered more useful (for an arbitrary definition of "useful") are larger and thus easier to hit If this sounds easy to you, then you're absolutely right. As you will see shortly, the Kivy implementation of such a UI is rather straightforward. The buttons To start off, we are going to tweak the Button class in Kivy language (let's name the file recorder.kv): #:import C kivy.utils.get_color_from_hex <Button>:background_normal: 'button_normal.png'background_down: 'button_down.png'background_color: C('#95A5A6')font_size: 40 The texture we set as the background is solid white, exploiting the same trick that was used while creating the color palette. The background_color property acts as tint color, and assigning a plain white texture equals to painting the button in background_color. We don't want borders this time. The second (pressed background_down) texture is 25 percent transparent white. Combined with the pitch-black background color of the app, we're getting a slightly darker shade of the same background color the button was assigned: Normal (left) and pressed (right) states of a button – the background color is set to #0080FF The grid structure The layout is a bit more complex to build. In the absence of readily available Modern UI-like tiled layout, we are going to emulate it with the built-in GridLayout widget. One such widget could have fulfilled all our needs, if not for the last requirement: we want to have bigger and smaller buttons. Presently, GridLayout doesn't allow the merging of cells to create bigger ones (a functionality similar to the rowspan and colspan attributes in HTML would be nice to have). So, we will go in the opposite direction: start with the root GridLayout with big cells and add another GridLayout inside a cell to subdivide it. Thanks to nested layouts working great in Kivy, we arrive at the following Kivy language structure (in recorder.kv): #:import C kivy.utils.get_color_from_hex GridLayout:    padding: 15    Button:        background_color: C('#3498DB')        text: 'aaa'    GridLayout:        Button:            background_color: C('#2ECC71')            text: 'bbb1 '        Button:            background_color: C('#1ABC9C')            text: 'bbb2'        Button:            background_color: C('#27AE60')            text: 'bbb3'        Button:            background_color: C('#16A085')            text: 'bbb4'    Button:        background_color: C('#E74C3C')        text: 'ccc'    Button:        background_color: C('#95A5A6')        text: 'ddd' Note how the nested GridLayout sits on the same level as that of outer, large buttons. This should make perfect sense if you look at the previous screenshot of the Windows Phone home screen: a pack of four smaller buttons takes up the same space (one outer grid cell) as a large button. The nested GridLayout is a container for those smaller buttons. Visual attributes On the outer grid, padding is provided to create some distance from the edges of the screen. Other visual attributes are shared between GridLayout instances and moved to a class. The following code is present inside recorder.kv: <GridLayout>:    cols: 2    spacing: 10    row_default_height:        (0.5 * (self.width - self.spacing[0]) -        self.padding[0])    row_force_default: True It's worth mentioning that both padding and spacing are effectively lists, not scalars. spacing[0] refers to a horizontal spacing, followed by a vertical one. However, we can initialize spacing with a single value, as shown in the preceding code; this value will then be used for everything. Each grid consists of two columns with some spacing in between. The row_default_height property is trickier: we can't just say, "Let the row height be equal to the row width." Instead, we compute the desired height manually, where the value 0.5 is used because we have two columns: If we don't apply this tweak, the buttons inside the grid will fill all the available vertical space, which is undesirable, especially when there aren't that many buttons (every one of them ends up being too large). Instead, we want all the buttons nice and square, with empty space at the bottom left, well, empty. The following is the screenshot of our app's "Modern UI" tiles, which we obtained as result from the preceding code: The UI so far – clickable tiles of variable size not too dissimilar from our design inspiration Scalable vector icons One of the nice finishing touches we can apply to the application UI is the use of icons, and not just text, on buttons. We could, of course, just throw in a bunch of images, but let's borrow another useful technique from modern web development and use an icon font instead—as you will see shortly, these provide great flexibility at no cost. Icon fonts Icon fonts are essentially just like regular ones, except their glyphs are unrelated to the letters of a language. For example, you type P and the Python logo is rendered instead of the letter; every font invents its own mnemonic on how to assign letters to icons. There are also fonts that don't use English letters, instead they map icons to Unicode's "private use area" character code. This is a technically correct way to build such a font, but application support for this Unicode feature varies—not every platform behaves the same in this regard, especially the mobile platform. The font that we will use for our app does not assign private use characters and uses ASCII (plain English letters) instead. Rationale to use icon fonts On the Web, icon fonts solve a number of problems that are commonly associated with (raster) images: First and foremost, raster images don't scale well and may become blurry when resized—there are certain algorithms that produce better results than others, but as of today, the "state of the art" is still not perfect. In contrast, a vector picture is infinitely scalable by definition. Raster image files containing schematic graphics (such as icons and UI elements) tend to be larger than vector formats. This does not apply to photos encoded as JPEG obviously. With an icon font, color changes literally take seconds—you can do just that by adding color: red (for example) to your CSS file. The same is true for size, rotation, and other properties that don't involve changing the geometry of an image. Effectively, this means that making trivial adjustments to an icon does not require an image editor, like it normally would when dealing with bitmaps. Some of these points do not apply to Kivy apps that much, but overall, the use of icon fonts is considered a good practice in contemporary web development, especially since there are many free high-quality fonts to choose from—that's hundreds of icons readily available for inclusion in your project. Using the icon font in Kivy In our application, we are going to use the Modern Pictograms (Version 1) free font, designed by John Caserta. To load the font into our Kivy program, we'll use the following code (in main.py): from kivy.app import Appfrom kivy.core.text import LabelBaseclass RecorderApp(App):    passif __name__ == '__main__':    LabelBase.register(name='Modern Pictograms',                       fn_regular='modernpics.ttf')    RecorderApp().run() The actual use of the font happens inside recorder.kv. First, we want to update the Button class once again to allow us to change the font in the middle of a text using markup tags. This is shown in the following snippet: <Button>:    background_normal: 'button_normal.png'    background_down: 'button_down.png'    font_size: 24    halign: 'center'    markup: True The halign: 'center' attribute means that we want every line of text centered inside the button. The markup: True attribute is self-evident and required because the next step in customization of buttons will rely heavily on markup. Now we can update button definitions. Here's an example of this: Button:    background_color: C('#3498DB')    text:        ('[font=Modern Pictograms][size=120]'        'e[/size][/font]nNew recording') Notice the character 'e' inside the [font][size] tags. That's the icon code. Every button in our app will use a different icon, and changing an icon amounts to replacing a single letter in the recorder.kv file. Complete mapping of these code for the Modern Pictograms font can be found on its official website at http://modernpictograms.com/. Long story short, this is how the UI of our application looks after the addition of icons to buttons: The sound recorder app interface – a modern UI with vector icons from the Modern Pictograms font This is already pretty close to the original Modern UI look. Using the native API Having completed the user interface part of the app, we will now turn to a native API and implement the sound recording and playback logic using the suitable Android Java classes, MediaRecorder and MediaPlayer. Thankfully, the task at hand is relatively simple. To record a sound using the Android API, we only need the following five Java classes: The class android.os.Environment provides access to many useful environment variables. We are going to use it to determine the path where the SD card is mounted so we can save the recorded audio file. It's tempting to just hardcode '/sdcard/' or a similar constant, but in practice, every other Android device has a different filesystem layout. So let's not do this even for the purposes of the tutorial. The class android.media.MediaRecorder is our main workhorse. It facilitates capturing audio and video and saving it to the filesystem. The classes android.media.MediaRecorder$AudioSource, android.media.MediaRecorder$AudioEncoder, and android.media.MediaRecorder$OutputFormat are enumerations that hold the values we need to pass as arguments to the various methods of MediaRecorder. Loading Java classes The code to load the aforementioned Java classes into your Python application is as follows: from jnius import autoclassEnvironment = autoclass('android.os.Environment')MediaRecorder = autoclass('android.media.MediaRecorder')AudioSource = autoclass('android.media.MediaRecorder$AudioSource')OutputFormat = autoclass('android.media.MediaRecorder$OutputFormat')AudioEncoder = autoclass('android.media.MediaRecorder$AudioEncoder') If you try to run the program at this point, you'll receive an error, something along the lines of: ImportError: No module named jnius: You'll encounter this error if you don't have Pyjnius installed on your machine jnius.JavaException: Class not found 'android/os/Environment': You'll encounter this error if Pyjnius is installed, but the Android classes we're trying to load are missing (for example, when running on a desktop) This is one of the rare cases when receiving an error means we did everything right. From now on, we should do all of the testing on Android device or inside an emulator because the code isn't cross-platform anymore. It relies unequivocally on Android-specific Java features. Now we can use Java classes seamlessly in our Python code. Looking up the storage path Let's illustrate the practical cross-language API use with a simple example. In Java, we will do something like this in order to find out where an SD card is mounted: import android.os.Environment;String path = Environment.getExternalStorageDirectory().getAbsolutePath(); When translated to Python, the code is as follows: Environment = autoclass('android.os.Environment')path = Environment.getExternalStorageDirectory().getAbsolutePath() This is the exact same thing as shown in the previous code, only written in Python instead of Java. While we're at it, let's also log this value so that we can see which exact path in the Kivy log the getAbsolutePath method returned to our code: from kivy.logger import LoggerLogger.info('App: storage path == "%s"' % path) On my testing device, this produces the following line in the Kivy log: [INFO] App: storage path == "/storage/sdcard0" Recording sound Now, let's dive deeper into the rabbit hole of the Android API and actually record a sound from the microphone. The following code is again basically a translation of Android API documents into Python. If you're interested in the original Java version of this code, you may find it at http://developer.android.com/guide/topics/media/audio-capture.html —it's way too lengthy to include here. The following preparation code initializes a MediaRecorder object: storage_path = (Environment.getExternalStorageDirectory()                .getAbsolutePath() + '/kivy_recording.3gp')recorder = MediaRecorder()def init_recorder():    recorder.setAudioSource(AudioSource.MIC)    recorder.setOutputFormat(OutputFormat.THREE_GPP)    recorder.setAudioEncoder(AudioEncoder.AMR_NB)    recorder.setOutputFile(storage_path)    recorder.prepare() This is the typical, straightforward, verbose, Java way of initializing things, which is rewritten in Python word for word. Now for the fun part, the Begin recording/End recording button: class RecorderApp(App):    is_recording = False    def begin_end_recording(self):        if (self.is_recording):            recorder.stop()            recorder.reset()            self.is_recording = False            self.root.ids.begin_end_recording.text =                 ('[font=Modern Pictograms][size=120]'                 'e[/size][/font]nBegin recording')            return        init_recorder()        recorder.start()        self.is_recording = True        self.root.ids.begin_end_recording.text =             ('[font=Modern Pictograms][size=120]'             '%[/size][/font]nEnd recording') As you can see, no rocket science was applied here either. We just stored the current state, is_recording, and then took the action depending on it, namely: Start or stop the MediaRecorder object (the highlighted part). Flip the is_recording flag. Update the button text so that it reflects the current state (see the next screenshot). The last part of the application that needs updating is the recorder.kv file. We need to tweak the Begin recording/End recording button so that it calls our begin_end_recording() function: Button:        id: begin_end_recording        background_color: C('#3498DB')        text:            ('[font=Modern Pictograms][size=120]'            'e[/size][/font]nBegin recording')        on_press: app.begin_end_recording() That's it! If you run the application now, chances are that you'll be able to actually record a sound file that is going to be stored on the SD card. However, please see the next section before you do this. The button that you created will look something like this: Begin recording and End recording – this one button summarizes our app's functionality so far. Major caveat – permissions The default Kivy Launcher app at the time of writing this doesn't have the necessary permission to record sound, android.permission.RECORD_AUDIO. This results in a crash as soon as the MediaRecorder instance is initialized. There are many ways to mitigate this problem. For the sake of this tutorial, we provide a modified Kivy Launcher that has the necessary permission enabled. The latest version of the package is also available for download at https://github.com/mvasilkov/kivy_launcher_hack. Before you install the provided .apk file, please delete the existing version of the app, if any, from your device. Alternatively, if you're willing to fiddle with the gory details of bundling Kivy apps for Google Play, you can build Kivy Launcher yourself from the source code. Everything you need to do this can be found in the official Kivy GitHub account, https://github.com/kivy. Playing sound Getting sound playback to work is easier; there is no permission for this and the API is somewhat more concise too. We need to load just one more class, MediaPlayer: MediaPlayer = autoclass('android.media.MediaPlayer')player = MediaPlayer() The following code will run when the user presses the Play button. We'll also use the reset_player() function in the Deleting files section discussed later in this article; otherwise, there could have been one slightly longer function: def reset_player():    if (player.isPlaying()):        player.stop()    player.reset()def restart_player():    reset_player()    try:        player.setDataSource(storage_path)        player.prepare()        player.start()    except:        player.reset() The intricate details of each API call can be found in the official documents, but overall, this listing is pretty self-evident: reset the player to its initial state, load the sound file, and press the Play button. The file format is determined automatically, making our task at hand a wee bit easier. Deleting files This last feature will use the java.io.File class, which is not strictly related to Android. One great thing about the official Android documentation is that it contains reference to these core Java classes too, despite the fact they predate the Android operating system by more than a decade. The actual code needed to implement file removal is exactly one line; it's highlighted in the following listing: File = autoclass('java.io.File')class RecorderApp(App):    def delete_file(self):        reset_player()        File(storage_path).delete() First, we stop the playback (if any) by calling the reset_player() function and then remove the file—short and sweet. Interestingly, the File.delete() method in Java won't throw an exception in the event of a catastrophic failure, so there is no need to perform try ... catch in this case. Consistency, consistency everywhere. An attentive reader will notice that we could also delete the file using Python's own os.remove() function. Doing this using Java achieves nothing special compared to a pure Python implementation; it's also slower. On the other hand, as a demonstration of Pyjnius, java.io.File works as good as any other Java class. At this point, with the UI and all three major functions done, our application is complete for the purposes of this tutorial. Summary Writing nonportable code has its strengths and weaknesses, just like any other global architectural decision. This particular choice, however, is especially hard because the switch to native API typically happens early in the project and may be completely impractical to undo at a later stage. The major advantage of the approach was discussed at the beginning of this article: with platform-specific code, you can do virtually anything that your platform is capable of. There are no artificial limits; your Python code has unrestricted access to the same underlying API as the native code. On the downside, depending on a single-platform is risky for a number of reasons: The market of Android alone is provably smaller than that of Android plus iOS (this holds true for about every combination of operating systems). Porting the program over to a new system becomes harder with every platform-specific feature you use. If the project runs on just one platform, exactly one political decision may be sufficient to kill it. The chances of getting banned by Google is higher than that of getting the boot from both App Store and Google Play simultaneously. (Again, this holds true for practically every set of application marketplaces.) Now that you're well aware of the options, it's up to you to make an educated choice regarding every app you develop. Resources for Article: Further resources on this subject: Reversing Android Applications [Article] Creating a Direct2D game window class [Article] Images, colors, and backgrounds [Article]