16

I want to view the source code of R's match.call function. As it is an internal function, I downloaded the R source code, went to ./src/main/names.c and looked for match.call there.

Thus, I found out that the corresponding .c function is called do_matchcall. Ok, but how can I find out which of the dozens of .c files in ./src/main/ contains the function do_matchcall?

Btw I'm using a Windows machine, in case that makes a difference.

vrwim
  • 13,020
  • 13
  • 63
  • 118
Wolfgang Pößnecker
  • 641
  • 1
  • 6
  • 13

4 Answers4

10

As a Windows user, here are a couple of options. The first one is preferable, but the second one is OK for occasional use:

  • Download grepwin, which will allow you to search Windows directories using the powerful grep command that both Joshua and Gavin have mentioned. It (or some equivalent) is indispensable if you'll be doing much poking around in program source directories.

  • Use the search bar at this site to search the R source directory for the definition of do_matchcall. Clicking on the result it returns will tell you that do_matchcall is "[defined] at line 1193 of file unique.c", and will provide a hyperlink to the code in unique.c.

Like I said, though, you'll ultimately be much happier if you equip your Windows box with some implementation of grep.

Josh O'Brien
  • 159,210
  • 26
  • 366
  • 455
  • Wow, thanks, the site you linked is awesome. Perfect for someone who only occasionally wants to look up a single R internal. – Wolfgang Pößnecker Nov 07 '12 at 23:02
  • @WolfgangPößnecker -- Glad that helps. Do be aware that the link will break when R is next updated. To make it work then, just edit the "version part" of http://fossies.org/dox/R-2.15.2/index.html from "R-2.15.2" to whatever the current version is. – Josh O'Brien Nov 07 '12 at 23:04
  • @WolfgangPößnecker: [searching for `do_matchcall` on rseek.org](http://www.rseek.org/?cx=010923144343702598753%3Aboaz1reyxd4&q=do_matchcall&newwindow=1&cof=FORID%3A11) is another alternative, and [rseek.org](http://www.rseek.org/) is handy in general. – Joshua Ulrich Nov 07 '12 at 23:36
  • Google magic as an option too `site:https://svn.r-project.org/R/trunk/src/ do_matchcall` – thelatemail Nov 08 '12 at 00:47
10

I know this has been asked a long time ago, but since it's still relevant, I thought I'd add some resources people can use to find the proper R source files.

  1. First, with pryr you can use the show_c_source function which will search on GitHub the relevant piece of code in the C source files.

    body(match.call)

# .Internal(match.call(definition, call, expand.dots))

pryr::show_c_source(.Internal(match.call(definition, call, expand.dots)))

    Which takes you to this page, showing that unique.c contains the function do_matchcall.

  2. I've put together this tab delimited file, building on the names.c file and using find-in-files to determine the location of the source code. There are some functions that have platform-specific files, and a handful of others for which there is more than one file with relevant source code. But for the rest the mapping is pretty well established, at least for the current version (3.1.2).

Dominic Comtois
  • 10,230
  • 1
  • 39
  • 61
6

Uwe Ligges wrote an "R Help Desk" article on this very topic in R News (2006, 6(4):43-45).

Once you've identified the actual C function that is used, then use your filesystem search tools to search for the function name in the relevant source folder; in this case ./src/main/, e.g. on Linux

$ grep -r -H "do_matchcall" ./src/main/
./src/main/.svn/text-base/names.c.svn-base:{"match.call",   do_matchcall,   0,  11, 3,  {PP_FUNCALL, PREC_FN,   0}},
./src/main/.svn/text-base/unique.c.svn-base:SEXP attribute_hidden do_matchcall(SEXP call, SEXP op, SEXP args, SEXP env)
./src/main/unique.c:SEXP attribute_hidden do_matchcall(SEXP call, SEXP op, SEXP args, SEXP env)
./src/main/names.c:{"match.call",   do_matchcall,   0,  11, 3,  {PP_FUNCALL, PREC_FN,   0}},

indicating that unique.c is the place to look in this case.

To the best of my knowledge, there is no way (other than invoking a system call to the terminal) to identify from within R which source file contains the C code for a given function in R - well, not without rewriting grep or find or similar using R code :-)

Gavin Simpson
  • 170,508
  • 25
  • 396
  • 453
  • I'll second your recommendation of Uwe Ligges' article. It's essential reading for R users. – Josh O'Brien Nov 07 '12 at 22:46
  • 1
    Thanks for your helpfull answer. However, I already knew the referenced article. But the lines "If the filename is not obvious, then it can be found by simply ‘grep’ping for the string in R’s $R_HOME/src/" never made sense to me as I never used Linux. The answer by Josh O'Brien also indirectly shows that 'grep'ing is not thaaaat 'simple' or 'natural' for a windows user. – Wolfgang Pößnecker Nov 07 '12 at 22:58
  • As a non-Windows user, I was assuming that you would know how to search for strings in files on your OS. Really that is your question and you might get better help if you wrote that as a Q (but not on [so] as it is not a programming issue - SuperUser might be OK) instead of directing this at the [tag:r] tag people. – Gavin Simpson Nov 07 '12 at 23:00
  • 1
    Well, while you are obviously right, I simply couldn't know that my question was referring to a Windows-specific utility when I posted it. I was thinking that some R-specific function or utility or trick existed to solve my problem. Before your answers, I didn't realize that my question was ill-posed. – Wolfgang Pößnecker Nov 07 '12 at 23:36
  • @WolfgangPößnecker I don't think your question is necessarily ill-posed. I was just pointing out that I had made certain assumptions when answering. The real answer to your question is that there isn't an R utility of similar that will give you what you want. I perhaps should have said that too... And this often happens - the real question *evolves* from the "discussion" or answers you are provided with here on [so]. – Gavin Simpson Nov 08 '12 at 09:04
1

Don't download a utility for a one time file content search. You can use findstr. It isn't suitable for a full file system search, but I think it's perfect for what you are trying to do.

  1. run the command shell

cmd

  1. go to the directory (if you are not sure what the path is, find it in explorer than view the full path by clicking in the address bar) You can copy/paste that directory.
    If the path is C:\Users\jsmith\documents\projects\happy\src\main

cd C:\Users\jsmith\documents\projects\happy\src\main

  1. use findstr

findstr do_matchcall

if you want to search sub directories

findstr /s do_matchcall

if only want to know the filename

findstr /m do_matchcall

if you want to send the results to another file

findstr /s /m do__matchcall >searchresult.txt

if you want the line numbers

findstr /n do_matchcall

Findstr also supports regular expressions. To see all the options

findstr /?

Chris McCowan
  • 467
  • 5
  • 10