The assorted finds of Artefact Publishing
On a conlang discussion forum, someone mentioned the potential usefulness of a reverse sound change applier. Such a thing would take a word and a set of sound change rules, and return a list of all the words which, if they underwent the specified sound changes, would end up as the specified word. This is a good deal more tricky than going the other way!
I immediately figured that I should add such a thing to IPA Zounds. This has necessitated a refactoring of the existing code, which I am still engaged in. In the course of this, I have realised two things:
Unit tests are critical. I can’t imagine trying to make these invasive changes without them, since I would have no idea whether I had broken anything or not.
I don’t understand why I put in particular restrictions. Why can’t I specify a suprasegmental in the result component of a rule, or a homorganic variable in the source component? Maybe there was a good reason, but if so I didn’t write it down, and it doesn’t make sense to me now.
Once this refactoring is done, I shall add in some form of reverse engine, and then update the GUI. Aside from getting a whizzier looking About page automatically, I can’t see this last part being much fun. I am currently stalled on the dictionary project because of the GUI.
Posted by jamie on October 9, 2005 13:17+13:00