Workflow with Subversion and LaTeX

In the previous post I showed how easy it is to install your own Subversion nowadays. But that brings you only that far because you need to integrate SVN into your daily working routine. Here I show how I do that with my LaTeX projects.

The basic workflow with Subversion is as follows:

  • Update the working copy (right-click and ‘SVN Update’). This updates the locally stored file with the one from the repository. You need to do this when working on the project with others.
  • Make some changes to the document.
  • Commit the changes to the repository (right-click and ‘SVN Commit’). Make sure to add some notes what you changed to make version tracking easier.

When working on large documents it is really important for me to visualise changes. MikTeX and TeX Live come with the great tool latexdiff that highlights differences between two documents. The syntax is simply latexdiff old.tex new.tex > diff.tex from the command-line. Using SVN appears perfect for the job as we can access old revisions easily. The syntax is then, for example, latexdiff-vc --svn -r HEAD file.tex, which compares the file.tex with the HEAD revision (i.e. the latest revision in the repository).

I generally work with multi-file LaTeX projects, i.e. the main file is called master.tex where all sections of the project are included via \include or \input (introduction, chapter 1, chapter 2 etc.). For this, latexdiff offers the -flatten option that replaces the \include and \input commands with the actual content. Unfortunately, latexdiff currently does not support -flatten and -vc simultaneously, so we have to do that manually.

The idea is as follows:

  • Flatten the document, i.e. merge the contents of all .tex files into a single document. This includes the bibliography.
  • Compare the flattened document with a previous version with latexdiff and generate a PDF automatically.
  • Commit the flattened document and the PDF.

I also want to create two more versions of the flattened document:

  • One were each sentence starts in a separate line. This is useful if you check differences on the source rather than the compiled document latexdiff changes the source by including \DIFdelbegin and \DIFdel tags to highlight changes, so if you want to copy & paste some of the text later this is more convinient.
  • One were all comments and to-dos (using the todonotes package \todo{}) are removed. This file could be sent to publishers (or just people who want to have a look), but are not supposed to read your comments.

That sounds like a handful, but there are fortunately scripts around for every task I mentioned above. As my programming skills are just at cavemen level I am only able to combine these scripts into a single batch file that does all these steps in one go.

What you need:

  • flatex flattens the document. It is quite old but works perfectly. It also inputs the bibliography so that you truly need only one .tex file for the whole project. Get the single flatten.c source here and compile or download the executable  flatex.exe (1143 downloads) .
  • by . A perl script that removes comments and to-dos (see post). In principle it flattens the document as well, but for some reason it messes up the bibliography and tables in my files. So we don’t use that feature and use flatex for flattening.
  • by Andrew Stacey. Another perl script that moves each sentence to a new line.

For the last two scripts to work you need to install Perl.

Now it’s time for my batch  cleandiff.bat (1205 downloads)  that combines all scripts. To run it you need to copy it were your master/main tex-file is (in my script it is called ‘master.tex’). You also need to specify were you save the three scripts above (I just copied them to C:).

Here’s what it does:

Create new working folder. Cleandiff contains all new files, temp will be deleted

mkdir cleandiff
mkdir cleandiff\temp

Call the ancient flatex. It does its job perfectly, but for some reason it does not create a new file as it is supposed to do, but instead the new file is called.flt. So we just rename it to clean-comments.tex and move it to a new folder.

C:\flatex.exe master.tex
rename master.flt clean-comments.tex
move clean-comments.tex cleandiff\clean-comments.tex

cd cleandiff

We add and commit this file which is then the HEAD Revision.

svn add clean-comments.tex
svn commit -m "" clean-comments.tex

Remove comments and \todo{} remarks. On its own the script does the same as flatex, but it does not work well with my documents (tables and the bibliography get messed up). So we just call it, if necessary, after flatex.

perl C:\ clean-comments.tex > clean-nocomments.tex

Put each sentence into a new line. Useful if using svn diff as it makes it easier to track changes. Should be called upon clean-comments.tex as it is useful to keep tack on your comments and todos as well.

perl C:\ clean-comments.tex > clean-linebreak.tex

Compare HEAD revision with the PREVIOUS revision. First checkout, the run latexdiff. You can specify the revisions you want to compare, e.g. -r 28 uses the 28th revision.

svn cat -r PREV clean-comments.tex > temp\PREV.tex
svn cat -r HEAD clean-comments.tex > temp\HEAD.tex

latexdiff -t UNDERLINE temp\PREV.tex temp\HEAD.tex > rev_head_prev.tex

Compile the document. It is necessary to run pdflatex 3 times in that order to get the bibliography and cross references right. nonstopmode is used to ignore errors.

pdflatex rev_head_prev.tex -interaction=nonstopmode
bibtex rev_head_prev.tex
pdflatex rev_head_prev.tex -interaction=nonstopmode
pdflatex rev_head_prev.tex -interaction=nonstopmode

We delete some files that are not required to track changes. The last command deletes empty 0 byte files that may be generated.

del *.lof
del *.lot
del *.toc
del *.log
del *.dvi
del *.aux
del *.bbl
del *.blg
del *.brf
del *.out
rmdir /Q /S temp
for /r %%F in (*) do if %%~zF==0 del "%%F"

Finally we add and commit the remaining files

svn add *
svn commit -m ""
svn commit cleandiff

If all goes well you should have five committed files in the cleandiff folder:

  • clean-comments.tex: the flattened document that is used for latexdiff
  • clean-linebreak.tex: flattened document including comments where all sentences start on a new line.
  • clean-nocomments.tex: flattened document where comments and to-dos are removed
  • rev_head_prev.pdf and rev_head_prev.tex: PDF and .tex highlighting the changes to the previous version.

After a long working day where I made lot’s of improvements on my papers I run the script so that I can always keep track of my changes.