Stack Exchange

Building documents with rubber

2011-12-04 by . 12 comments

Post to Twitter

It’s common to compile our documents a couple of times to ensure, amongst other things, correct cross-referencing and indices. Sometimes, we also rely on a Makefile in order to make things easier. Now enters rubber, a powerful tool to help us on our building adventures.

Our first example

According to the man pages, rubber  “is  a wrapper for LaTeX and companion programs. Its purpose is, given a LaTeX source to process, to compile it enough times to  resolve all  references,  possibly  running  satellite programs such as BibTeX, makeindex, Metapost, etc.  to produce appropriate data files.”

All the major Linux distros have rubber in their repositories. A simple yum install rubber does the job for me. Let’s see an example, with mydoc.tex:

We should now run pdflatex, makeindex, pdflatex again. Hm let’s run it again three more times just to be sure (I know I do). Or we could go with rubber --pdf mydoc . Done. Of course, don’t mention Macbeth. :)

There we go! rubber calls pdflatex and makeindex as many times as need. The --pdf flag tells rubber to run the PDF tools instead the DVI ones. Do you think the current folder is messy, full of temporary files? Go with rubber --clean mydoc and voilà, now you’ll have only the resulting PDF file and your tex source file.

If you want to get rid of all generated files, except of course the sources, just add the --clean --pdf flags:

Thankfully, rubber is smart enough to check if the files are updated:

High hopes part II

Yes, there goes another rubber tree plant… again. Time for more advanced stuff with rubber.

I had the opportunity of answering a nice question on about rubber. It makes use of directives, the hidden gems.

A directive is a comment in the form of

% rubber: cmd args

which adds information for rubber. It’s very simple. Let’s say I want to use a custom style for my index. A quick look into the rubber manual gives me the following entry:

  • <style>: specifies the index style to be used. It’s important to note that each of these directives may be used with an optional first argument of the form (foo,bar,quux) in order to specify that the directive only applies to the indexes named foo, bar and quux. By default, directives are applied to all indices.

Now, I just need to add a line to mydoc.tex:

Well, it didn’t work. Why? I was wondering what happened. Then the manual tells me why:

  • When using the package makeidx instead of index, the directives must of course be prefixed by makeidx. instead of index., and the optional first argument is not accepted.

Since I’m using makeidx, a quick fix

does the trick.

There are several directives for you to play with. Have fun!

Mac users, don’t worry!

Of course, there’s an app for that… I mean, rubber can be used with Mac too! Go to the official rubber repository and download the current compressed file, e.g, rubber-1.1.tar.gz. Extract it to a folder and use ./configure and sudo make install.

Update: XeLaTeX, where are you?

Unfortunatelly, rubber offers no XeLaTeX support out of the box. But there a way to fix this.

Wouter Bolsterlee provided an elegant solution. Go here, download the attached file and copy it to rubber‘s rules/latex directory. In my Fedora system, the full path is /usr/lib/python2.7/site-packages/rubber/rules/latex. Now, just run

rubber --module xelatex mydoc

and XeLaTeX is recognized! I’m pretty sure a similar file might be used for LuaLaTeX.

Final thoughts

rubber has a lot more features than the ones I described in these examples. Take a look on the manual by running man rubber and info rubber for a complete reference. This tool is certainly a great addition to our TeX utility belt. :)

Filed under LaTeX


Subscribe to comments with RSS.

  • Me says:

    I should mention an alternative, I’ve been using for a while: CMake with UseLaTeX

    It supports the same items that you mentioned in your post, additionally it does out of source builds!

  • Karl Berry says:

    Hi — I have nothing against rubber, but for the record, there are a number of programs around to do this job. Off the top of my head, three are latexmk, mkjobtexmf, and texi2dvi/texi2pdf. I’m sure that perusing CTAN would turn up more. (For myself, I tend to write the rules I need directly in make instead of involving another tool.)


    • Paulo Cereda says:

      Hi Karl, thanks a lot for the feedback! Indeed, we have lots of tools around to do this job. :) I decided to write about rubber mostly because there was a question on and I actually put some effort on answering it. Apart from the man pages, I could not find a nice rubber introduction, so I thought of writing one. :) IMHO rubber has its merit, unfortunately I miss some important features, like a built-in support for both emerging XeTeX and LuaTeX engines. Anyway, for simple documents, it’s quite suitable. For complex structures, I agree 100% with you, I also tend to write the rules in a Makefile.

  • «If you want to get rid of all generated files, except of course the sources, just add the –pdf flag:»

    It should be: «just add the –clean flag.»

  • Raphael says:

    Thanks for presenting rubber! I did not know it and have in fact written a less-complete script with similar features.

    Note that --clean seems to ignore some files (hardcoded?). If you observe such behaviour, you can add a directive to your LaTeX file. Just put % rubber: clean file1 file2 ... before \documentclass.

    --into does not work well for me. I have a document which includes files from several subfolders via input. rubber complains that it does not find files via relative paths. Does somebody have a fix?

    Any recommendations for runnings rubbe continuously? It seems suited for that kind of usage (--cache). I was thinking of just putting a call to rubber into a (timed) loop.

    • Paulo Cereda says:

      Thanks for the feedback, Raphael! :)

      The clean directive is a lifesaver. In one of my initial attempts with rubber, two of the auxiliary files weren’t removed at all. I have no idea why some files are ignored, maybe they are indeed hardcoded. Anyway, at least we have a workaround. :P

      I tried to reproduce the error you mentioned, but my code worked. I’ll take a deeper look, your code is probably more sophisticated than mine. :)

      In other news, during my tests, I got surprised. It seems rubber requires a certain parameters order, with the project name appearing last. I tried rubber --pdf mydoc --into testfolder, but the pdf was generated in the current folder; rubber --pdf --into testfolder mydoc works. Anyway. :P

      I really like the timed loop approach. :) I was talking to a friend and he told me about incron (inotify cron). It’s similar to cron, but it handles filesystem events rather than time periods. He suggested to watch mydoc.tex for changes and call rubber. I’m quite reticent to try this approach. :)

  • [...] just stumbled on this blog article about rubber and was thinking: “Isn’t that especially what latexmk does?”. So, now I wonder: [...]

  • copain says:

    Hello there.

    I also had the clean directive not cleaning two file (.bbl and .blg). I noticed that using rubber –clean –pdf file or rubber –clean –pdf file.tex instead of rubber –clean –pdf file.pdf

    will remove the two files.

  • Hello there,

    I should mention an alternative: AutoLaTeX.

    It is also available on CTAN.

    It is similar to Rubber. Additionally, it run tools (named translators) to generate PDF files from several figure sources files (svg, dia…) In this way you do not need any more to export manually your figures into PDF/PNG files.


  • Leave a comment

    Log in
    with Stack Exchange