Want to share your content on R-bloggers? click here if you have a blog, or here if you don't.

In programming, sometimes it’s useful to write a macro rather than a function. (Don’t worry if you’ve never heard the term before.) In this post, I’ll give an example of use of macros in R. using the gtools package on CRAN.

I wanted to write some utility code to help me reuse my earlier R commands during an interactive R session. Most (though not all) of what I wanted is already provided in the excellent R user interface systems such as ESS, RStudio and vim-r, but for various reasons I generally use the command line directly, especially for short sessions. (I also have developed my own vim editing mappings for R.) Specifically, I wanted to develop utilities to perform the following tasks:

• Display all my recent commands, possibly restricting to those matching or excluding certain character strings.
• Choose one of the recent commands for re-execution.
• Re-execute a recent command by number or matching string.

One nice thing about R is that one can easily form some command programmatically in a character string, say using paste(), and then execute the string as a command, using eval(). Thus it would be easy to code up the above-listed tasks into functions that I can call when needed, except for one problem: Within the body of a function, one has a different environment than at the caller’s level.

Take for instance (a slightly modified version of) the first example in the online help for defmacro() in gtools, in which to goal is to write a function to recode as NAs all entries in a data frame column with a certain value. The following will NOT work:

setNA <- function(df,var,val)
df$var[df$var == val] <- NA


The problem is that within setNA(), df will be a data frame that is only a copy of the one in your call. The recoding to NAs will be made only to the copy. The defmacro() function in gtools would do what you really want:

setNA <- defmacro(df,var,val,
expr={df$var[df$var == val] <- NA}

With this, df really will be the desired data frame. I urge you to try a little test with both of the above code snippets.

The code I wrote for my command-history utilities is rather short, but too long to place in a blog post; instead, you can access it here. There is a short sample session at the end of the file, which again I urge you to execute by hand to fully understand it. Take note of the tasks I coded as macros instead of functions, and think about why I needed to do so.

How is all this magic accomplished? The defmacro() function still makes use of eval(), substitute() etc., but it already does that work for you, so that you can write your macro while thinking of it as function. Your code is a function — defmacro() builds the function and returns it, and by the way note that that means it is debugable — but again, the point is that you don’t have to deal with all the calls to eval() etc.

If you are a C/C++ programmer, note that this differs from macros in that language, which are straight substitutions made by the compiler preprocessor.