4

I want to run a postprocessing script on my pdf. For integration with my editor it should be executed at the end of the pdflatex run, automatically -- ideally without having to change the compilation command in order not to break the behaviour for other documents.

I thought that I could do this with the atveryend package, using 

%% minimal example file where \write18 is used for cropping.
\documentclass[]{article}

\title{}
\author{}
\usepackage{lipsum}
\usepackage{atveryend}
\AtVeryVeryEnd{
    \message{^^J=== STARTING POSTPROCESSING ===^^J}
    \immediate\write18{pdfcrop \jobname.pdf \jobname.pdf}
}

\begin{document}

\maketitle

\begin{abstract}
    \lipsum[1]
\end{abstract}

\section{A section}

\lipsum[2-3]

\end{document}

But this fails. One can see from the log entries that even \AtVeryVeryEnd is executed a while before Output written on document.pdf is printed, i.e. the PDF file's handle is still open at that time. I can't say how this would behave on unix (where exclusive file locks are less common), but on my Windows machine it makes pdfcrop crash.

Is there some method to execute the command at the "VeryVeryVery" end of the pdflatex run, after the PDF file has been closed?

kdb
  • 1,889
  • 1
    No. There's no macro-level access to processes at or after the end of the PDF creation process. – Joseph Wright Jul 16 '15 at 09:32
  • 4
    The output PDF file is finalized only after (the primitive) \end has been executed, so nothing can be read from the input at that time. What about pdflatex file && pdfcrop file? – egreg Jul 16 '15 at 09:32
  • 2
    only thing you could do is launch a shell command that waits for the pdf to appear before editing it, that seems rather fragile though, compared to just configuring the editor to run the commands in sequence, all editor configs will allow that, so they can do latex bibtex latex, etc. – David Carlisle Jul 16 '15 at 09:35

1 Answers1

5

You can do the following:

\newcommand{\postProcessDocument}[1]{%
\ifx\popCallStop\undefined%
  \immediate\write18{pdfLaTeX
    -synctex=1
    -interaction=nonstopmode
    -jobname=\jobname\space 
    \gdef\string\popCallStop{x}%
    \string\input{\string\"\jobname\string\"}}%
  #1%
  \expandafter\stop%
\fi}
\postProcessDocument{%  post procesing commands
  \message{^^J=== STARTING POSTPROCESSING ===^^J}
  \immediate\write18{pdfcrop \jobname.pdf \jobname.pdf}
}

\documentclass[]{article}

\title{}
\author{}
\usepackage{lipsum}

\begin{document}

\maketitle

\begin{abstract}
    \lipsum[1]
\end{abstract}

\section{A section}

\lipsum[2-3]

\end{document}

This will compile your document twice (with one call). First, \popCallStop is undefinded. Therefore, your main document is compiled. Afterwards, \popCallStop is defined with some dummy value. So, the second part of the compilation is running your post-processing. But as a user, you only do a single compilation.

This is some variation of a trick I saw once on stackexchange.

kdb
  • 1,889
Thomas F. Sturm
  • 25,210
  • I submitted a change suggestion to your code to also support documents, where the root file has spaces in its name. I also added -interaction=nonstopmode to prevent the process from hanging, which is problematic when running pdflatex from an editor interface. – kdb Jul 16 '15 at 11:44
  • I just noticed that this breaks synctex in weird ways (different behaviour for different files) on my machine, regardless whether the file has spaces in the name or not. I assume that pdfcrop is what breaks it though. – kdb Jul 16 '15 at 12:01
  • @kdb: Normally you should post a comment to the answer, asking about adding some suggestions instead of editing answers directly –  Jul 16 '15 at 12:38
  • @kdb I never use spaces in file names (old habit from DOS days), so I have'nt experienced this kind of problem before. But your edit seems to solve it. For synctex: yes, I can image that this whole process breaks it... – Thomas F. Sturm Jul 16 '15 at 13:36
  • @Christian Hupfer: Isn't that what the "peer review" layer when editing other people's post is for? It would have been difficult and to describe in a comment in an easy-to-understand way the suggested change. – kdb Jul 17 '15 at 09:08
  • @kdb: I disagree with that. Posts should be edited only if they contain wrong points or are formatted badly. I reject such editings which could be proposed in comments generally when it comes to my Suggested Edit queue –  Jul 17 '15 at 10:59