I've patched up some bugs to MapThing 1.2 (which I never got around to releasing anyway) and have added some performance-related enhancements.
In particular, MapThing objects can now be serialised -- this allows sketches using large numbers of files, or files containing complex geometries, to be loaded much more quickly from Java serialisation (.ser) files. I'll eventually add an example, but you'll currently need to write the code to do this yourself. That said, it's only a few lines and is quite straightforward.
However, serialisation does one very BAD THING: the connection to the underlying shape file is lost because the GT classes that underpin shape file reading and writing do not themselves support serialisation. This means that the data cannot be reprojected on to the screen. So if you (re)launch the sketch at a different resolution or size from the one in use when the data was first serialised then it will most likely be mis-projected.
In effect, serialisation is most useful if you want to use the shape files in a 'kiosk' context where you have full control of how the sketch is presented to the user. The start-up time can fall to as little as 10% (!!!) of the un-serialised start-up time; so in a kiosk context this can be a big plus but you pay for it with flexibility.
Things are also slowly getting smarter around parsing the column names in a file for x and y coordinates and also for making it easier for the developer to project multiple value columns from the same shape file without a lot of faffing. I suppose that a clone() function would be useful, but for now you have to follow the example PDE file for copying data.
Anyway, always happy to have feature requests -- I can't always implement them right away (or even at all) but I do listen and do try. Currently top of the 1.4 'to do' list is arrowheads on lines (beginning and/or end).
Oh, and I finally remembered to put this all in SVN so future releases should be a little more precise about what has changed.