Take the 2-minute tour ×
TeX - LaTeX Stack Exchange is a question and answer site for users of TeX, LaTeX, ConTeXt, and related typesetting systems. It's 100% free, no registration required.

Is there a script to automate externalizing TikZ graphics?

(As suggested by Scott here)

share|improve this question
possible duplicate of Automagically externalising pdftricks and TikZ pictures –  Sharpie Aug 13 '10 at 23:19
Yes, but the answers here are better. –  Neil G Aug 14 '10 at 3:46

5 Answers 5

up vote 36 down vote accepted

The TikZ 2.10 library, external, addresses this issue. Externalizing all TikZ graphics in a LaTeX document is as easy as:


If your document is named report.tex, this will dump out a series of images: report-figure0.pdf, report-figure1.pdf, etc, etc. To get a more manageable output, you can specify a directory for in which the files are to be placed and names that are more descriptive than report-figure0.pdf:


% Rest of preamble
% Begin document, write stuff

% Set a filename for the next tikzpicture.
\begin{tikzpicture}  % This will be output to figures/importantFigure.pdf
  % Picture code

To compile the document, you will need to ensure the figures directory exists and run pdflatex with shell execution enabled:

mkdir figures
pdflatex -shell-escape <tex file>

A whole pile of options is available to customize the way the figures are output. See section 31.4 (this numbering is subject to change!) of the development version documentation for complete info and options. The manual also covers how to obtain EPS output- but this option looks less polished than PDF output.

share|improve this answer
+1 for "ramming a flaming stick into your eyeball" ;-) –  Giel Aug 10 '10 at 13:19
Typo: you use a dollar for a comment after the \begin{tikzpicture}. –  Loop Space Sep 19 '10 at 19:48
Checking out with CVS is more fun than trying to contribute back with CVS, I think. Though I've been shying away from it lately as I can't bring myself to use a non-distributed VCS directly anymore :-), and it's unlikely that there is a DVCS that can pull from CVS do to it's lack of atomicity :-(. –  SamB Dec 19 '10 at 5:36
Wow, this produces … “interesting” crashes (spurious, random “file ended prematurely” messages etc.) when used together with e.g. todonotes, which also uses TikZ under the hood. I’ve tried conditionally enabling that package and disabling externalization for the to do entries – to no avail. –  Konrad Rudolph Jan 29 '11 at 19:34
This external library is now available as part of PGF 2.10. –  Faheem Mitha Feb 26 '11 at 21:20

I've made a shell script, which makes pdf's from all my tikz-pictures, but only if necessary, i.e. if the source is newer than an existing pdf.

As far as I see, my solution has the advantage to others, that the pdfs are compiled in parallel, but you get noticed if one or multiple pdfs fail to compile, and then the rest of the compilation stops.

So, I have got an .sty-file for the script-call like the following. There is a definition for the images-Folder, two example images and the script call and return value handling. \usepackage{imgs} is the first line after \documentclass in my main document, so it's made sure all pdfs are always up to date.

% images dir and tex/pdf file names
\newcommand*\blank{ }

% script exection, need shell-escape option enabled!
\immediate\write18{\imgs img.sh \imgs\blank
    \imgb\blank % insert additional images like this...
    ; echo $? > pdfscreated.res % script return value stored, and later read to ScriptResult

\read\myscriptresult to \ScriptResult


The image is included like this in a (very short) main document:

\usepackage{imgs} % see code above!
\usepackage{header} % may be the same as for images, or different, as you like

An example for an image would be

\usepackage{../header} % own header for images, may be the same as for main document or not...
    %draw sthg...
    \draw (0,0) -- (1,1);

And the called shell script (which could easily be replaced by perhaps nicer solutions like the python script by pmav99) looks like the following:


# first argument contains folder of .tex-files
cd $1

# counting the number of pdfs, e.g. for the laiter waiting for pids

# relevant files and commands
texbin="lualatex --shell-escape "
if [ ! -e $header ] ; then
    echo "$header doesn't exists. This is bad."
    exit 3

while [[ $# > 0 ]] ; do
    names[$cnt]="$1" # store name for later error messages and pdfcrop

    if [ ! -e $tex ] ; then
        echo "$tex doesn't exists. Maybe forgot to create it?"
        exit 2
    if [ ! -e $pdf ] ; then
        # start latex in background, and wait later for all pids, so it takes only as long as the longest compilation takes
        $texbin $tex &

    # do nothing, if pdf is newer then all source files
    if [ $pdf -nt $tex ] ; then
        if [ $pdf -nt $header ] ; then
            echo "$pdf is up to date."
    # now rebuild pdf as it's necessary
    $texbin $tex &

#store occured errors during pdf compilation!

while [[ $cnt>0 ]]; do
    if [ ${pids[$cnt]} ]; then
        wait ${pids[$cnt]}
        if [ 0 -ne $? ] ; then
            # error occured!
            logs="${names[$cnt]}.log $logs"
            errorstring="${names[$cnt]} $errorstring"
        $pdfcrop $pdf $pdf &

if [ 0 -ne $errorflag ]; then
    cat $logs
    echo --------------------------------
    echo errors occured in: $errorstring
    exit 1
    echo --------------------
    echo - all pdfs created -
    echo --------------------
    # only exit code 0 -> no error!
    exit 0
share|improve this answer

When I was writing my MSc thesis, I had created all of my figures and diagrams with tikz and pgfplots (using matplotlib2tikz). Keeping all of them within the source was not acceptable, since the compilation was getting really slow, so I was searching for something that would allow me to automatically compile all those figures to pdf files that I could include in my text.

For various reasons, I was not entirely satisfied with any of the other solutions so I wrote a Python script that suited my purpose.

The things you must keep in mind are:

  1. You put the script on the same level as your main.tex file
  2. You write your tikz code. You place each figure within a file with extension *.tikz. Each file contains only \begin{tikzpicture} ... \end{tikzpicture}.
  3. You write the preamble that you want to use with the *.tikz files. This doesn't necessarily contains the same packages as the preamble of your main.tex file. The reason you keep a common preamble for all the figures is to make it easy to make changes in things like fonts etc.
  4. You run the script!

The script searches recursively within a folder structure for *.tikz files and compiles them using the specified preamble. Each time the scripts runs, it stores the modification time of each *.tikz file and on subsequent runs it compiles only the newly created files of the files that have been modified since the last run. If you make changes in your preamble, you just pass a command-line argument and it compiles all the *.tikz files.

On linux it works great. I haven't tested it on windows, but I don't see why it shouldn't run. Anyway, even if there is any problem, the fix is going to be really trivial

share|improve this answer

Here's my alternative; Konrad's answer is probably just as good, and certainly better documented.

for job in `grep -o beginpgfgraphicnamed\{[0-9A-Za-z-]*\} < $1 | sed -e 's/beginpgfgraphicnamed{//' | sed -e 's/}//'`; do
    pdflatex --jobname $job $1;
    rm $job.log;
    rm $job.out;
    rm $job.aux;
share|improve this answer
thank you! –  pmav99 May 12 '11 at 21:36

I’ve just come up with the following. It seems to work quite well (on Unix/Linux and Mac):

for name in $( \
    grep '^[:space:]*\\beginpgfgraphicnamed' $SOURCE | \
    sed s'/[:space:]*\\beginpgfgraphicnamed{\(.*\)}/\1/' \
); do
    pdflatex --jobname=$name $SOURCE

An explanation:

It uses grep to filter out all lines that contain \beginpgfgraphicsnamed, but only if there’s only whitespace before the command in the line. This prevents that commented-out commands are also found.

Next, it invokes sed to filter from those lines only the relevant part, i.e. the command argument that’s to be used as the pdflatex jobname. It then uses for to iterate over all these jobnames and calls pdflatex in turn for each of them.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.