R Classes and Methods (Part 1)

Loading...

En provenance du cours de Johns Hopkins University

Developing Data Products

1652 notes

Johns Hopkins University

Developing Data Products

1652 notes

Cours 9 sur 10 dans Specialization Data Science

A data product is the production output from a statistical analysis. Data products automate complex analysis tasks or use technology to expand the utility of a data informed model, algorithm or inference. This course covers the basics of creating data products using Shiny, R packages, and interactive graphics. The course will focus on the statistical fundamentals of creating a data product that can be used to tell a story about data to a mass audience.

À partir de la leçon

R Packages

In this module, we'll dive into the world of creating R packages and practice developing an R Markdown presentation that includes a data visualization built using Plotly.

R Packages (Part 1)7:11

R Packages (Part 2)14:59

Building R Packages Demo18:00

R Classes and Methods (Part 1)13:49

R Classes and Methods (Part 2)11:19

Rencontrer les enseignants

Brian Caffo, PhD
Professor, Biostatistics
Bloomberg School of Public Health
Jeff Leek, PhD
Associate Professor, Biostatistics
Bloomberg School of Public Health
Roger D. Peng, PhD
Associate Professor, Biostatistics
Bloomberg School of Public Health

Hi everyone, this lecture is about writing classes and methods in R.

The basic idea of writing classes and methods in R, is that

you want to represent new kinds of data, or new data types.

That R doesn't kind of you know, by default support.

So for example you can think of a basic data type, like a list.

This is this, this is kind of inherent to R.

There's all kinds of functions and kind of processes that you can do

to lists that gets all kind of built in that kind of comes with R.

But if you can imagine thinking of a new data type.

For example, a new field emerges and they come up with their own data.

This is not something that the original designers of R could have foreseen.

And so, you need to develop a new structure to support that data,

a new set of functions to kind of apply to that kind of data type.

And so, that's what the idea behind writing classes and methods in R is.

So any time something new comes up, the R system can adapt.

And kind of support new types of data and new data structures.

So that's what we're going to talk about in this lecture.

The classes and methods system in R

is basically a system for doing object-oriented programming.

And years ago when R was first coming out it was quite an interesting language.

Because it was both an interactive system that

supported a kind of formal system for object orientation.

And so, many other languages at the time and things

like, things that had kind of

well known object orientated programming infrastructure.

Like C++ and Java were generally not considered interactive languages.

Those are the kind of languages where you write programs,

you compile them and then you could run them.

And not the kind of thing where you type things in the command

line and execute things one at a time like you do in R.

So a lot of the R code, a lot of the code in R

for supporting devel, the development of classes

and methods was written by John Chambers himself.

So he was the creator of S and has written a lot of code in the R system.

And so the basic ideas are documented in his book Programming with Data.

And so, the idea behind the class and methods system in R is to

kind of allow people to cross over from being a user to being a programmer.

And you recall from the basic overview of the R

language that we talked about in the R programming class.

John Chambers original idea for the, for the S language was that people would

start off as users and slowly as they became more familiar with the language.

They would cross over and become programmers as

they tried to customize the system for themselves.

So R has a var-, the sa-, the basic same, the same basic philosophy.

Whereas that people could start off, users can have executing functions and then

as their needs grow a-, and as their needs kind of outgrow the R system.

They could become programmers and develop new things in R.

And that's the basis of classes and methods.

So object oriented programming is a little bit different

in R than it is most of the languages.

So if, if you're familiar with the language, like C++ or Java.

There's document that is helpful, but some of the

ideas in R are a little bit different so.

It's used.

And you might want to pay, pay attention to some of the details here.

So in R there are basically two styles of classes and methods.

And this is a function of kind of historical development there's kind of

an older style, and the, the new, there's a newer style.

So sometimes we'll see the older styles referred to as S3 classes and methods.

So that corresponds with version three of the S

language and these classes and methods are, are informal.

They'are a little kludgy and they're kind of not very formally rigorous.

But they get the job done and they work quite well.

And a lot of the basic functions that you use

in R, are kind of built on S3 classes and methods.

So we know that they are really quite useful.

And so, and they're not going away anytime soon.

But the more formal and rigorous system for writing classes

and methods is called, usually referred to as the S4 system.

And the became, because it came with the version 4 of the S language.

It was introduced into R at version 1.4.0, which was in December 2001.

And these are sometimes called new styled classes and methods.

And for new projects, it's generally encouraged, for for

people to kind of use the S4 classes or methods.

If they're going to be developing new types of classes and methods.

So, these two systems for the foreseeable future will be living side, by side.

Just because of the sort of history of R and a

lot of the kind of built in functionality that's written in S3.

But the S3 and S4 systems are separate in the sense that

you don't typically mix then together as you are programming new things.

So the systems are fairly independent of each other.

And so developers are encouraged to use the S4 style classes

in methods that use extensively for example in the bio-conductor project.

But the S3 classes and methods are sometimes

useful because they are very quick and dirty.

They are a little bit easier to program.

And for short term types of projects you may, it may be tempting

to kind of build up some S3 classes and methods in the first go.

But for the most part we'll focus on S4.

I'll talk a little bit about the S3 system just so you're familiar with it.

But I'll talk mostly about developing in the S4 system.

So one thing that's important to know is that all of the

implementation of the object orient method in the S4 classes and methods.

Is implemented in the methods package.

The methods packaged is usually loaded by default.

So if you just start up R from a

clean installation the methods package will usually be loaded.

But if it's not, you can always load it

with the library function by calling the library methods.

So the basic idea this, in object oriented programming R is

that you have classes and methods as you may have guessed already.

So a class is basically a description of some thing, right?

And the thing will usually be some new thing like a new data type or a new idea.

And a class can be defined using the set class method in the methods package.

And then you have objects which are instances of that classes.

And objects can be created using the new function or

are often there can be created using other types of functions.

So a method is a function that operates

on certain class on a certain class of objects.

And so, methods are functions that are specific to certain classes.

And then finally you have a generic function,

which is an R function that dispatches methods.

So generic function is usually it tries to encapsulate a generic kind of

concept, and we'll see a few of these a little bit later.

For example, the plot function is probably one of the most used generic functions.

And the plot functions, plotting is a fairly generic concept.

And what, how you make a plot may depend on the kind of data that you're plotting.

And so there are many different plotting methods

that can be applied to different data types.

And the plots function jobs it to, to look at the

data types and, and dispatch specific methods on those different data types.

So generic functions do not actually do anything.

They don't do any computation, they don't do an sort of work.

Their main job is to dispatch methods, and we'll

talk about what that means more specifically in a second.

And so, a method is the implementation of a

generic function for an object of a particular class.

So the basic concepts are class, object, method, and generic function here.

So, as you're working your way through the, learning this system, there are

a number of help files in the methods package that may come in handy.

so, the primary documentation are in the help files.

So you can look at the classes and methods help pages.

You can also look at help pages for ?setClass, ?setMethod, and ?setGeneric.

Some of this gets very technical so you may want to skim through it right now.

But, as you learn the system, it will make more sense in the future.

So every object and R has a class, and the

class of that object can be determined with the class function.

I just take the number one here.

You may recall an R the numbers by

default are kind of real numbers or double precision numbers.

And so, there considered a numeric class, and so that's what you see here.

Illogicals have there own class there true and false are logicals.

If you generate some random numbers, those

are numeric that again is a numeric vector.

The missing value NA is by default a logical, but you

can have missing kind of of a numeric unit missing integer etcetera.

And then this a character string will have in the class of characters.

So that's all fairly straightforward.

These are kind of atomic classes.

But you can go beyond the atomic classes.

For example, I could fit a linear model.

And the output from the lm function is an lm class.

So if you think about it, the output from fitting a linear

model is not what you would think of as a standard data type.

Like numeric or integer or character but it is a data type.

There are certain ideas that are kind of encapsulated

in, from the output of a linear model.

If you are familiar with statistics.

And so and so it may deserve its own data type.

And so therefore we call it the lm data type, or data, or class.

So that, that's the idea behind classes in R.

So generics and methods are the, these types of functions that

you'll, you'll write to implement

functionality for certain types of classes.

An S4 and S3 generic functions will

look different but they are conceptually the same.

They play the same role.

And so when you can when you program you

can write meth-, new methods for an existing generics.

So there may be a generic like plot that's already out there.

And then you'll write a new method for a new type of data.

But there may be new generics that you create.

So you may create a new generic function for doing a new operation.

And then, you'll have to write the generic and the method for a given data type.

Here are a few generic functions.

These are S3 generic functions.

And so these are probably functions that you've seen already.

The mean is a generic function.

Print is a generic function.

And so you see when I print out these

functions, there's not there's almost no code associated with them.

The only code that comes out of the mean function is this use method mean.

And similarly for the print function, the generic

function, the all you see is use method print.

And so, the basic idea is that, these generic functions, there job is

to find an appropriate method for whatever data type is being passed to them.

So you can see what methods are available, for

the mean generic function, by calling the methods function.

And you can see there's a mean function for dates.

There's a, [UNKNOWN] there's a default method.

And there's one for div time, Posixct and Posixlt.

Now as you add as you kind of load packages into

the R environment, those packages may add methods for the means.

So, as you kind of load packages throughout your work in R, this set

of methods that are available to the mean function may change over time.

And so, this list can grow as you add more packages.

So, here's an example of an S4 generic

function so this is from the methods package.

It's called show, and it's the equivalent of print, but it's the S4 version.

And you see there's a lot more stuff that gets printed out when

you, when you, when you print out the body of the show function.

The base, the basic idea's the same.

The, the code for show is the standard generic.

And then show and based, and it just tries to

identify the class of the object that's being passed to it.

And then it will dispatch the appropriate method.

And so, most of the time in R you

don't call the Show function or the Print function.

Because most, because objects will be auto printed at the command line.

So here you can see what methods

are associated with the show, generic function.

And you can see that this these are, there are number of methods here

one is for these are all in the methods package, that you can see.

There is one for the in ref class, is

one for functions, is one for generic functions etc,.

So all these different types of objects

have a special print method or show method.

So, the basic mechanism by which generic functions and

methods interact with each other is, goes as follows.

So, the first argument for any generic

function is an object of a particular class.

All right?

There may be other arguments but the first object is important.

Sorry, the first argument is important.

And so the way the process works, is that

the generic function will check the class of that object.

So it will try to identify the class.

It will do, the generic function will essentially do a search

to see if there's a method that is specific for that class.

If there is a method for that class, then it will call

the method of that object, and then the generic functions job is done.

The method will do the rest of the work.

If there is no method for that class Then the

generic will search, will see if there's a default method.

So a kind of catch all method that always gets applied

if there's no other kind of more specific type of method.

If there'a default method, then the default method

is called and then it does all the work.

If there's no default method, then, there's an error.

Because the generic function doesn't know what to call, and

there's no default method so, it can't do anything by itself.

so, a lot of times you may want to look at the code for methods.

But it's important to note that it's generally not

possible to, kind of, print the code of a method function.

Just like you might print the code of any other function.

So some functions that are not generics or methods.

You can just type the name of the function in the command line and

you, you'll see the body of the function when you look at the code.

That's not always the case with S3 or S4 methods.

And so, the appropriate way to look at the code or method

is to use either the get S3 method or the get method function.

So get S3 method is for gen,ge,gen S3 generics, I'm sorry methods.

And the get method function is for S4.

And the basic idea, is you pass this function.

The first argument is the name of the generic.

And the second argument is the is the class of the object.

Explorer notre catalogue

Rejoignez-nous gratuitement et obtenez des recommendations, des mises à jour et des offres personnalisées.

Coursera propose un accès universel à la meilleure formation au monde, en partenariat avec des universités et des organisations du plus haut niveau, pour proposer des cours en ligne.

© 2019 Coursera Inc. Tous droits réservés.

Télécharger dans l'App Store

Disponible sur Google Play