Optimize dirstate walking
This generally cuts the time for hg status/diff in half, from 2s down to 1s.
The main parts I'm trying to optimize are:
1) os.walk stats every file. dirstate.changes then stats every file again.
2) os.walk yields every file and subdir to dirstate.traverse who yields every
file and everything in the dirstate map. dirstate.walk then
filters this mass and yields every file to the caller. There should be
fewer steps in here, and fewer duplicate strings yielded.
3) dirstate.walk runs util.unique on the results from dirstate.traverse,
even though it is also passing things through dirstate.seen to look for
duplicates.
I've turned os.walk into something hg specific that takes all the dirstate
ignore and matching rules into account. The new function also takes an
function arg (statmatch()) the caller supplies to help filter out
files it doesn't care about. dirstate.changes uses this to update state
for each file, avoiding the second stat call.
dirstate.walk is changed to turn the match function it is passed into
a statmatch function. The only real difference is that a statmatch
function takes the stat data as a second parameter. It now calls
dirstate.walkhelper, who requires a statmatch function to be passed.
This fails test-walk, but right now I think this is from a sorting error
fixed by this patch.
Index: crew/mercurial/dirstate.py
===================================================================
Mercurial git BK (*)
storage revlog delta compressed revisions SCCS weave
storage naming by filename by revision hash by filename
merge file DAGs changeset DAG file DAGs?
consistency SHA1 SHA1 CRC
signable? yes yes no
retrieve file tip O(1) O(1) O(revs)
add rev O(1) O(1) O(revs)
find prev file rev O(1) O(changesets) O(revs)
annotate file O(revs) O(changesets) O(revs)
find file changeset O(1) O(changesets) ?
checkout O(files) O(files) O(revs)?
commit O(changes) O(changes) ?
6 patches/s 6 patches/s slow
diff working dir O(changes) O(changes) ?
< 1s < 1s ?
tree diff revs O(changes) O(changes) ?
< 1s < 1s ?
hardlink clone O(files) O(revisions) O(files)
find remote csets O(log new) rsync: O(revisions) ?
git-http: O(changesets)
pull remote csets O(patch) O(modified files) O(patch)
repo growth O(patch) O(revisions) O(patch)
kernel history 300M 3.5G? 250M?
lines of code 2500 6500 (+ cogito) ??
* I've never used BK so this is just guesses