kolibri-discuss team mailing list archive

[Anders Steinlein <anders@xxxxxxx>]

I've done some profiling and performance testing, and things are looking good. Although pure performance isn't improved significantly, memory consumption drops about ~10% which is pretty cool. Speed seems quite hard to improve much though, since we really are dependant on traversing and evaluating all data to make our decisions. This is what holds us back, not creating the DOM document in itself.

However, after some experimentation, a couple of small improvements can be made. As isXml() is as trivial as it currently is and only has one caller, inlining it in buildScalar() gives a small bonus. We can also short-circuit empty array values in buildComplex() in the same way we short-circuit null and empty strings (using a static variable for the array to compare with gave me the most consistent results). These two changes resulted in small, although consistently improved results for me.
-- 
https://code.launchpad.net/~frode-danielsen/kolibri/xml-generator-rewrite/+merge/5284
Your team Kolibri Discuss is subscribed to branch lp:kolibri.