I just had to prepare a poster based on roughly 30 publications, and for several of them I didn't have the original figures but only the manuscript as a pdf file. Using okular and a magnification of 800%, I've got screenshots of these figures as comparatively highly resolved bitmaps, but the price I had to pay was that the editing of the poster in LibreOffice (which I've used as the least common denominator) was getting almost unbearably slow.
I couldn't silence the thought that it should be possible to take a 'vector screenshot' from a pdf file. I had the vague idea that pdftocairo could be useful in this respect, since it can output arbitrary parts of a pdf file as pdf or svg. And it turned out that Peter Williams, a young radio astronomer from Harvard, had the same idea and came up with a script which does exactly what I wanted.
I've fixed a small error (pageh should also be an integer) and ensured Arch and Fedora compatibility (python2), but otherwise its Peter's script:
|
#! /bin/bash
|
|
|
|
# original: <https://gist.github.com/pkgw/3892706>
|
|
# see <http://newton.cx/~peter/2012/10/extracting-pdf-figures-as-pdfs-in-linux/>
|
|
|
|
margin=1
|
|
|
|
# XPDF gives its y coordinates in terms of the standard PDF coordinate
|
|
# system, where (0,0) is the bottom left corner and y increases going
|
|
# up. But pdftocairo uses Cairo coordinates, in which (0,0) is the top
|
|
# left corner and y increases going down. We can use pdfinfo to get
|
|
# the page size to translate between these conventions.
|
|
|
|
file="$1"
|
|
page="$2"
|
|
pageh=$(pdfinfo -f $page -l $page "$file" |grep '^Page.*size' \
|
|
|sed -e 's/.* x ' -e 's/pts.*$')
|
|
|
|
# Our variables end up in Cairo convention, so the box height is ybr -
|
|
# ytl.
|
|
|
|
xtl=$(python2 -c "import math; print int (math.floor ($3))")
|
|
ytl=$(python2 -c "import math; print int ($pageh) - int (math.ceil ($4))")
|
|
xbr=$(python2 -c "import math; print int (math.ceil ($5))")
|
|
ybr=$(python2 -c "import math; print int ($pageh) - int (math.floor ($6))")
|
|
w=$(python2 -c "print $xbr - $xtl")
|
|
h=$(python2 -c "print $ybr - $ytl")
|
|
|
|
# Lamebrained uniqifying of output filename.
|
|
|
|
n=1
|
|
|
|
while [ -f fig$n.pdf ] ; do
|
|
n=$((n + 1))
|
|
done
|
|
|
|
# OK to go.
|
|
|
|
echo pdftocairo -pdf -f $page -l $page -x $xtl -y $ytl -W $w -H $h \
|
|
-paperw $w -paperh $h "$file" '|' pdfcrop --margin $margin fig$n.pdf
|
|
exec pdftocairo -pdf -f $page -l $page -x $xtl -y $ytl -W $w -H $h \
|
|
-paperw $w -paperh $h "$file" - | pdfcrop --margin $margin - fig$n.pdf
|
Unlike Peter (and thanks to piet and haui), I can show an actual vector screenshot made by this script:
The size of this shot is 2.7 kB. A bitmap of this size showing the same section is so terribly ugly that I've decided not to present it here.
