I always knew this would be a long project, although I'm heartened as to how far it has gone. It appears to have been useful to some people. It is a feature of Open Source that the only time I can be really sure it is being used is when I get bug reports - which could be rather off-putting to anyone without the skin of a rhino.
This latest release, version 0.05, fixes quite a few bugs, and adds support for things like trap (signal handling). In my defence, most of the bugs are missing support for certain features, and it is still in beta. I still have not got pipelines sorted out (that is a huge job), or $?, but they will come eventually. I only get to work on this in odd moments, evenings and so on, and it is difficult to find the time for development stints of more that two hours. I could really do with spending a couple of solid days on it.
One problem I had was in inserting a marker into a string which would indicate a token delimiter. Everything I tried clashed in someway with a real data item. I solved it eventually by using a gash reference, but I can't help thinking there must be a more elegant solution.
I have also started using OO techniques more - this product really does benefit from that.
I am now using the profiler and Coverage for my tests, but have not sorted out a good way to integrate my command-line tests. The "Perl Testing" book has ideas, but they require CPAN modules and I would rather not increase the number of dependencies. If you download this program and are dismayed at the tests (I basically check the thing will load) then be assured that I have a test suite of over 90 scripts, just one of which is my regression test script which currently has 55 test paragraphs within it. Each time I find a bug or add a feature I add a test, just like I should.