On Fri, 2004-12-31 at 05:55 -0500, Alan Cox wrote: > On Thu, Dec 30, 2004 at 10:50:26PM -0800, Jamie Zawinski wrote: > > It takes *nearly a minute* to do that! I'm on a 2GHz machine. > > If it's not hitting the net, what's it doing, raytracing? > > Demonstrating the myth that interpreted languages are fast enough. > > It is actually doing a fair bit - its computing the dependancies, checking it > won't clash, checking the files fit etc To be clear. The part I _think_ Jamie is complaining about is just importing the metadata. And I agree with Daniel, it's all tied up in the python interning of the strings. There are a few ways to 'solve' the problem: 1. use a python xml library that does it all internal to python which should remove the translation issue. 2. thin out the metadata (reduce the number of nodes). 2a. alternatively, reorganize some of the metadata so it's faster to get to specific information. 3. write the xml importing layer in C. Since this would require that I learn how to adequately program in C before going on I think this part is unlikely. 4. ignore it for long enough and machines will just speed up. :) -sv
