Yes, I considered using my build system as well (in my case, a makefile) to run ImageMagick commands before running the XSLT processor. This is definitely the easiest way I could run external programs.
However, the locations of the images (as given by the fileref attributes in the XML document) are not known until processing time. So, if I want to run another command on these images before XSLT time, I would need to first parse the XML document for the image locations with something like XMLStarlet (or even just grep and sed), run the commands, then finally run the XSLT processor. That would work, but not as nicely as doing everything at XSLT processing time.
Perhaps Ant might be a better build solution than make, however. I will look into it.
Colin
On 8/17/07, David Cramer <
> wrote:
We do what you propose, but use our build system (an ant
script) to call batik http://xmlgraphics.apache.org/batik/ and
rasterize the images when going to html. The way we have it set up, when
you insert the svg image in XMetaL, a macro sees it's an svg and pops up a
dialog asking if you want it rasterized. If so, it converts it to a png and
displays the png (since XMetaL can't display svg !?!). The system also runs an
xslt on svgs that come from Visio to workaround a Visio bug that causes
arrowheads not to appear in pngs rasterized from Visio svg
source.
I understand that Inkscape
http://www.inkscape.org/ can also convert svgs to other formats from the command
line, but I haven't tried it.
David
From: Colin Shapiro
[mailto:]
Sent: Friday, August 17, 2007 10:24
AM
To:
Subject:
[docbook-apps] Image manipulation via XSLT extensions
Hello,
"Unfortunately, there is no single graphics file
format that meets all needs." - Bob
Bob is right. One of
the biggest annoyances I have when writing documents is that, when I want to
insert a graphic, I must reference and maintain multiple versions of the image
file in order to accommodate the various types of output I am producing:
<mediaobject>
<imageobject
role="fo">
<imagedata format="SVG"
fileref="figure.svg"/>
</imageobject>
<imageobject role="html">
<imagedata
format="PNG" fileref="figure.png"/>
</imageobject>
</mediaobject>
Let's say I created a
figure in SVG, as in the above example. This SVG is my original image
source, and in a perfect world, it would be all I'd need to carry
around. But, since we can't trust SVGs with our web browsers, I must
convert it to PNG and carry around another version of the file for HTML
output.
What if I actually could only keep the SVG, and have the
stylesheet somehow convert the image to alternate formats as needed? If
I run the HTML stylesheet, it sees an SVG, and calls an external program--say,
ImageMagick--to convert the file to a PNG.
This leads me to my
question. I know that with processors like Saxon and Xalan, you can
write extensions in Java/JavaScript. It seems to me that one would be
able to write an extension to call ImageMagick as needed. However, I
have absolutely no experience with XSLT extensions, and don't really know
where to start.
I do know where to find resources on the
subject. However, I thought I would first post here to see if anyone has
done anything similar, or thought about anything like this before.
-
Colin
