Loop functions are some of the most powerful functions in the R language and they make it kind of very easy to use, especially in an interactive setting. The idea behind a loop function is you want to execute a loop over an object or a set of objects in a way that's kind of that does a lot of work in, in a very small amount of space. That way, you don't have to type as much on the command line. Of course, we already learned about loops. We know about for loops and while loops, things like that, and those are all, work very well; however, they are com, less compact in a certain way. So, there are a couple of loop functions in R and they usually have the word apply in them somewhere. So some of the key ones are lapply, sapply, apply, tapply, mapply and the real workhorse function that I, that I'd like to talk about here is lapply. And the idea behind lapply is that you have a list of objects and you want to loop over the list of objects and apply a function to every element of that list. And so it's a very general concept. And it can be used very powerfully to do a lot of computations in a few, in just a little bit of typing. Sapply is a variant of lapply that simplifies the results. Apply is a function that operates over the margins of an array. So, this is very useful if you want to take summaries of matrices or other or, higher dimensional arrays. Tapply is short for table apply. And, it applies a function over subsets of a vector. And mapply is a multivariate version of real of lapply. So I'll go into details about how these work in a, in a, in a minute. There's also another function called split which doesn't actually apply anything to objects. But it's often useful in conjunction with functions like lapply or sapply because it splits objects into sub-pieces. So, lapply. Lapply takes three arguments. Basically the first argument is a list which is called X. The second argument is a function or the name of a function and then there are other arguments that are, can be passed to the dot dot dot argument. And the dot, dot, dot argument is used to pass arguments that go with the function that you're being, that's being applied to each of the elements in the list. If X is not a list, then you will be coerced to a list if possible. If it's not possible to coerce the object to a list, then you will get an error. So the lapply function, you can see, is very simple. The code for it is right here. Basically the func we look for the function if it's, if the object is not a list then it's coerced to a list using as.list and then the, the rest of the Lapply function is, is,is implemented internally in C code to make it a little bit faster. So the idea with Lapply is that you're going to take this list of things. And remember a list can contain any, any number of different types of objects. So they could be vectors, or matrices, or data frames, or whatever it may be and you want to apply a function to each one of these elements of the list. And that function is going to return something. It may not be the same thing that it originally was on the list. So, for example, it may take as an input, as a vector, but then it may return a scalar as a result. So, the function's going to return something for every single object in that list, and the return values are going to be assembled in a new list. And that's what lapply is going to return. So lapply, it's key to remember, it always returns a list. What goes in may or may not a list but it will be coerced to a list. And what comes out will always be a list. So here's a simple example. I'm creating a list of two elements, the first one's called A, and it's a sequence from one to five, the second one is called B, and it's it's ten or more random variables. So what I, and then, what I want to do is I want to loop over this lists of two elements and apply the mean function to each of those elements. So you can see that when I call Lapply on x and I apply the mean function I get another list back, w-, and notice the list has the same names as the original list, a and b. But now I've got the mean of the first element and the mean of the second element. And so that's how lapply works. Here I've got a slightly more complicated list. I've got four elements and I've got, I'm calling lapply to each of those elements and I'm getting the mean of each of those elements. So, now I've got a list with four elements. The names are preserved and notice, of course, you know, each of the elements of the original list was a vector of some, of a numeric vector of some sort. But what I'm getting back is a vector with just a single number in it, for each element of the list. So, here's another way I way to call, lapply. Here I'm creating a sequence one, of x, 1 to 4, and I'm calling runif, so, which generates a uniform random variables, to, using a random number generator. Now, the first argument to runif, is the number of uniform random variables that you want to generate. So if I say runif 1 it's going to generate a single random variable. If I say runif 2, it's going to generate a vector of two random variables. So, here I'm applying l, the runif function to sequence 1, 2, 3, 4. So, what I'm going to get is a list where the first element is a single random number random uniform. The second element's going to be a vector of two random uniforms. The third element's going to be a vector of three. And the fourth element's going to be a vector for random uniforms. And so ret, you'll note, if you know the runif function, you'll know that it has other arguments to it beyond the, the number of uniforms to generate. But those other arguments have default values so I don't need to specify them. Now, suppose I want to call the runif function on each one of these elements of X but I didn't want to just generate a uniform between zero and one which is default. Suppose I want to generate a uniform between zero and ten so now I need to pass some arguments to the runif function which are not the default values. In particular I need to change the max value. So I can do that through with lapply by passing these arguments through the dot dot dot argument. So here I'm calling lapply on X, I'm calling the run, I'm passing the runif function, but that I'm specifying that I want the min to be zero and the max to be ten. So now when I the, the list that I get out of this has random uniforms that are between zero and ten. So lapply and the associated functions make heavy use of what, of what are called anonymous functions. Anonymous functions are functions that don't have names, so you don't assign them a name of some sort but you can kind of generate them on the fly. So here is a just a quick example, I'm going to create a list that contains two matrices in it. The first is a mat, a two by two matrix and the second is a three by two matrix. So you can see the list here. There's two elements. They are named A and B. And suppose I want to, I want to extract the first column from each one of these matrices. So what I can do is I can call lapply so, there's no function that, out there that already extracts the first column of a matrix but this is easy to do. You can just write a function that just takes the first element, the first column of that matrix. So here I'm going to call lapply on x. And I'm, I'm going to write, I'm going to write the function right here, so I'm going to say function, and then I'm, I'm going to give it an argument, and then given that argument, I extract the first column. So here, when I call Lapply with this function I get the first column from A, and the first column from B. So this function doesn't exist except within the context of Lapply, and after the Lapply function is finished, the function basically goes away. So that's an anonymous function, because it doesn't have a name and lapply and a lot of these other types of functions use anonymous functions very heavily. Because unless there already exists a function that does the operation that you want to do, you're going to have to write the function kind of on the spot. So sapply is just a variant of lapply and all it does is it tries to simplify the result of lapply if possible. So recall that lapply always returns a list but sometimes you don't want a list, sometimes you just want something different. So for example, if the, if the result is a list where every element is a length 1 then what sapply will do is it'll return a vector of all,of all, of all those elements. Usually you don't want an ele, a list where every, where every element is a single number, for example, and so sapply will simplify that into just a vector. if, if the result is a list where every element is a vector of the same length. For example, if the, if the list comes back and every element has a length five, for example. Then what sapply will do it'll, it'll put those elements in a matrix that's, that's five by however long the matrix, the, the list is. So that, that's often what you want to happen. But if it, if it can't figure out how to simplify the object when it comes back, for example, if the object has many different types of things that comes back then it's just going to then it won't do anything. It will just return a list. So here in this, in this example when I called lapply and I applied the mean to everything what happens is that I got a list back that's of length four and every element of the list is a single number. So it would make, it would be a lot nicer if I just got my list back that was just, I'm sorry, if I just got a vector back with all these numbers in it. And that's exactly what sapply does. So sapply called on x with the mean function gives me a vector with four numbers in it. Of course, if I called mean on the, on the list by itself, that's not really going to work because mean is not meant to be applied to lists. And so you'll get a warning message of n a back.
