I’ve been having a good time taking advantage of the Transformiix XSLT processing engine that is part of the newer versions of Mozilla. However, I think I finally ran into something that it breaks on. Here’s what’s going on. I’m dynamically generating some graphics and need to output an imagemap with them. The initial plan was to plunk the imagemap that I generate from neato or dot as a chunk of CDATA. The problem is that transformiix ALWAYS escapes the output. It seems to completely ignore the disable-output-escaping attribute. So needless to say, this annoys me as it works fine with xsltproc.
My options were either have the server do the XSLT transformation or hack a way around it. I chose the latter. The first problem was that the imagemaps spit out by dot and neato aren’t valid as they look like this:
<area shape="rect" href="vendor.py?id=317" title="OpenSSH" alt="OpenSSH" coords="441,436,569,495">
Now to fix that I need to add the “/” to the end of it. So it was back to hacking the code on Graphviz (which I already had to do for it to compile on Fedora). Then I fixed that. Now the problem was I needed to make new nodes out of all these area tags. The solution was to create another XML document in memory and transfer the node over. This is probably not the most efficient, but it seems to work:
cmap = graphviz.renderGraph(buildGraph(vendorInfo,vulns), format="cmap", engine="neato") cmap = """<?xml version="1.0"?>\n<imagemap>\n%s</imagemap>""" % cmap imageMapDoc = libxml2.parseDoc(cmap) mapNode = imageMapDoc.children mapNode.setNs(myNS) imageNode.addChild(mapNode)
Probably not the most useful out of context but I’ll sum it briefly. So the first line makes the call that returns all of
the area tags as a string. The next line creates a dummy XML document while the third line parses
it. The fourth selects the