An SVG-esque to imagemap converter
Common Lisp Ruby Emacs Lisp
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Not Production Ready

System Goals

runtime performance (for a change)

It's not ready, but my intention is to expose this as a web-app at some point. That implies that I'll want a short parse/write time for a given map. It hasn't happened yet (I get the feeling it's mainly because of my liberal regex use, but I haven't whipped out the profiler yet. We'll see). It'll be kind of hard to optimize output since that involves mapping over the entire list of tags, so this basically translates to "very fast read time".

end user flexibility

This is one reason I don't use an XML-parser like others who have attempted an svg to imagemap converter (actually, he seems to be the only one)

  • Firstly because deserializing XML seems like it would be slower than parsing for relevant tags
  • Second beccause I can get away with it (all but one of the tags relevant to conversion are self-terminating, and I only need the opening tag of that one)
  • Third because it seems like it would be more difficult to get these tags out through a xpath expressions rather than through straight search (tags can be arbitrarily grouped, which means I can't just get all children of svg/)
  • Finally because XML parsers tend to barf when some jackass from marketing gives you an RTF file instead of an SVG (or a partial copy of an SVG). I would prefer this not barf if it can get any data at all out of the input.

These are notes for the author rather than usage notes

The relevant tags are

  • polygon and polyline are basically the same. They're a string of x, y pairs in the points attribute. Just get them, round them (they can be floats in SVG, but seem to be restricted to ints for area tags) and output as a comma separated list (In SVG, they're points="x,y x,y x,y...", in area tags they're coords="x,y,x,y,x,y..."
  • rect is slightly more complicated. Four relevant attributes: x, y, width, height in SVG. The area tags require two x,y pairs (top left and bottom right). In other words, x,y (x+width),(y+height)
  • circle has the same three attributes in SVG and area tags. The coordinates of the center and the radius. They're stored differently though; SVG stores it as three attributes (cx, cy, r) area tags represent it as a comma separated list in the coords property
  • path is basically a stripped down implementation of PS in a tag. Right now, it gets parsed and treated the same way as polygon and polyline, but that'll mishandle the H and V directives (because they each provide single numbers instead of x,y pairs). Other than that, it actually works fine.
  • g is the SVG group tag. I really don't have to care about them, except that they can have transform attributes which have to be applied to all child elements. The good part is that I can probably do that at read time if I'm sneaky enough. The bad part is that I'm not sure whether that's such a good idea...
  • svg the only reason I might care about this is that it contains the document width/height settings (which might come in handy if I want to let the user specify dimensions for this imagemap)


  • implement real path support. Really, this just means checking if a given directive is marked H or V and generating an appropriate point for it. Generating the second coordinate could concievably be done at write-time, but I'd still need to track input to make sure that they got stored as single numbers rather than paired randomly.

  • improve parser performance. I'm using regexes all over the place, which isn't the best idea speed-wise, but it was good to get it running. This is actually enough for my personal purposes, but it takes about 4 seconds just to parse this, which I get the feeling I can improve.

  • implement transform support. By the look of it, this'll be either 1, 4 or 5 functions called on each point under certain conditions, depending on how I want to implement it. As much as I'd like to abstract the entire thing into matrix-transform and then implement translate, scale, rotate, skew in terms of them, nested matrix transformations look like they could get scary fast (again, depending on how I implement them).

Other notes

  • The readers.lisp file is currently unused, and I've yet to put together a system definition