That branch has now been merged to trunk (in r1067). By no means is the optimisation phase over but we just fixed a few obvious areas that could be improved. Thanks to the work (mostly) done by Antti Kaiholawe're seeing a 50% reduction in encoding times for AMF0 and up to 10x (!) for AMF3 (depending on size of data set).
When we hit stable, the plan is to develop a c extension module for PyAMF that will be used by default instead of the pure python implementation (degrading where necessary). This has been achieved in other similar projects (specifically AMFPHP) with some dramatic results.
On Mon, 2008-02-18 at 04:58 +0100, Thijs Triemstra | Collab wrote: