The assorted finds of Artefact Publishing
Having done the bulk of the major refactoring work I wanted to do on IPA Zounds, I can now say that it is slower and has (temporarily) less functionality than it did before I started. Oops. The functionality will be restored, in a much nicer and useful form, soon, and with the new code I should be able to build the reverse applier without much hassle. The speed issue is proving intractable — I haven’t managed to figure out what exactly is causing the 30% slowdown, and I suspect that in part I just have to live with it. The code now does a bunch more work (yes, yes, none of it actually changes how the program behaves now, but it allows for future goodness, as well as being easier to maintain), and that takes time.
That said, I shall now see whether I can get anything more useful out of the Python profiler.
Posted by jamie on
October 12, 2005 10:25+13:00
To repeat my brilliance from last night:
So you've worked on it and it now does more stuff none of which has any visible benefit. In doing so you've made it slower and removed functionality. This is all intended to allow you to add functionality that no-one wants and you're not sure you can build. Wanna work for an IT consultancy firm? :)
Posted by: Anita
on October 12, 2005 14:24+13:00
Not quite that bad! There isn’t a great clamour for a reverse applier, but the new codebase will allow me to integrate automatic transliteration between the IPA and other alphabets based on user-supplied correspondences. This is a requested feature (yay for my solitary user!), and it wasn’t possible before to have this work in all of the desirable circumstances.
Posted by: Jamie
on October 12, 2005 17:32+13:00
And lo, there is a great clamour for a reverse applier. Plus, it’s no longer slower (or if it is, not by much at all) and the transliteration feature is mostly done in the back end.
Posted by: Jamie
on October 17, 2005 12:07+13:00