just starting bzr 10 times slower than mercurial

[SoloTurn]

just issueing "bzr" which displays the options seems to be 10 times slower than issueing "hg" for displaying the options (v 0.10). we get the worst performance difference with this using cygwin.

[SoloTurn (soloturn)]

$ time bzr
...
real 0m0.516s
user 0m0.440s
sys 0m0.050s

$ time hg
...
real 0m0.082s
user 0m0.050s
sys 0m0.030s

[John A Meinel (jameinel)]

I'm not sure how you do your math, 516/82 = 6.2, not 10.
Nonetheless, this is a known issue. 'hg' gets around this by using 'demanload', which delays loading python modules until they are actually used.
We plan on using a similar scheme. And I have the basic groundwork set up in the associated branch. (I'm writing it in a test-driven fashion, so I'm getting the "demandload" stuff working first, and then we'll start using it in modules)
I am hoping to get at least the lazy import stuff into 0.11, though we may not use it fully until the next release.

[John A Meinel (jameinel)]

the lazy load was merged into 0.11, but using it won't be until 0.12

[John A Meinel (jameinel)]

The initial stages of this has been merged, which helps a lot.

Specifically on my Mac:
time bzr-0.11
1.175
time hg
0.214
time bzr (.dev > 2071)
0.349

So we aren't quite as fast as hg yet, but I haven't fully tweaked all the things that I want to change. I *can* say that we are doing a lot better than we used to. (Also, bzr is slightly penalized because I have some plugins installed, but none for hg).

[John A Meinel (jameinel)]

closing this since the first work has been done.