When it comes to choosing software for statistical computing, it's tough to argue against R. Who could dislike a high quality, cross-platform, open source, statistical software product? It has an interactive console for exploratory work. It can run as a scripting language to replicate processes. It has a lot of statistical models built in, so you don't have to reinvent the wheel, but when the base toolset is not enough, you have access to a rich ecosystem of external packages. And, it's free! No wonder R has become a favorite in the age of data.

R was inspired by the S statistical language developed by John Chambers at AT&T. The name S is an allusion to another one-letter-name programming language also developed at AT&T, the famous C language. R was created by Ross Ihaka and Robert Gentleman in the Department of Statistics at the University of Auckland in 1991.

The general S philosophy sets the stage for the design of the R language itself, which many programmers coming from other programming languages find somewhat odd and confusing. In particular, it's important to realize that S was developed to make data analysis as easy as possible.

The key part here is the transition from analyst to developer. They wanted to build a language that could easily service both types of users. They wanted to build language that would be suitable for interactive data analysis through a command line but which could also be used to program complex systems, like traditional programming languages.

It's no coincidence that this book is structured that way. We will start doing data analysis first, and we will gradually move toward developing a full and complex system for information retrieval with a web application on top.

R is comparable, and often superior, to commercial products when it comes to programming capabilities, complex systems development, graphic production, and community ecosystems. Researchers in statistics and machine learning, as well as many other data-related disciplines, will often publish R packages to accompany their publications. This translates into immediate public access to the very latest statistical techniques and implementations. Whatever model or graphic you're trying to develop, chances are that someone has already tried it, and if not, you can at least learn from their efforts.

As we have seen, in addition to providing statistical tools, R is a general-purpose programming language. You can use R to extend its own functionality, automate processes that make use of complex systems, and many other things. It incorporates features from other object-oriented programming languages and has strong foundations for functional programming, which is well suited for solving many of the challenges of data analysis. R allows the user to write powerful, concise, and descriptive code.

In many ways, a language is successful inasmuch as it creates a platform with which many people can create new things, and R has proven to be very successful in this regard. One key limitation of the S language was that it was only available in a commercial package, but R is free software. Free as in freedom, and free as in free beer.

The copyright for the primary source code for R is held by the R Foundation and is published under General Public License (GPL). According to the Free Software Foundation (http://www.fsf.org/), with free software (free as in freedom) you are granted the following four freedoms:

• Freedom 0: Run the program for any purpose
• Freedom 1: Study how the program works and adapt it to your needs
• Freedom 2: Redistribute copies so you can help your neighbor
• Freedom 3: Improve the program and release your improvements to the public

These freedoms have allowed R to develop strong prolific communities that include world-class statisticians and programmers as well as many volunteers, who help improve and extend the language. They also allow for R to be developed and maintained for all popular operating systems, and to be easily used by individuals and organizations who wish to do so, possibly sharing their findings in a way that others can replicate their results. Such is the power of free software.

No programming language or system is perfect. R certainly has a number of drawbacks, the most common being that it can be painfully slow (when not used correctly). Keep in mind that R is essentially based on 40-year-old technology, going back to the original S system developed at Bell Labs. Therefore, several of its imperfections come from the fact that it was not built in anticipation for the data age we live in now. When R was born, disk and RAM were very expensive and the internet was just getting started. Notions of large-scale data analysis and high-performance computing were rare.

Fast-forward to the present, hardware cost is just a fraction of what it used to be, computing power is available online for pennies, and everyone is interested in collecting and analyzing data at large scale. This surge in data analysis has brought to the forefront two of R's fundamental limitations, the fact that it's single-threaded and memory-bound. These two characteristics drastically slow it down. Furthermore, R is an interpreted dynamically typed language, which can make it even slower. And finally, R has object immutability and various ways to implement object-oriented programming, both of which can make it hard for people, specially those coming from other languages, to write high-quality code if they don't know how to deal with them. You should know that all of the characteristics mentioned in this paragraph are addressed in Chapter 9, Implementing an Efficient Simple Moving Average.

A double-edged sword in R, is that most of its users do not think of themselves as programmers, and are more concerned with results than with process (which is not necessarily a bad thing). This means that much of the R code you can find online is written without regard for elegance, speed, or readability, since most R users do not revise their code to address these shortcomings. This permeates into code that is patchy and not rigorously tested, which in turn produces many edge cases that you must take into account when using low-quality packages. You will do well to keep this in mind.