##// END OF EJS Templates
bdiff: improve worst case behavior by 100x....
bdiff: improve worst case behavior by 100x. on 5.8MB (244.000 lines) text file with similar lines, hash before this change made diff against empty file take 75 seconds. this change improves performance to 0.6 seconds. result is that clone of smallish repo (137MB) with some files like this takes 1 minute instead of 10 minutes. common case of diff is 10% slower now, probably because of worse cache locality. but diff does not affect overall performance in common case (less than 1% of runtime is in diff when it is working ok), so this tradeoff looks good.

File last commit:

r2470:fe168927 default
r2577:fa76c5d6 default
Show More
filelog.py
107 lines | 3.4 KiB | text/x-python | PythonLexer
# filelog.py - file history class for mercurial
#
# Copyright 2005 Matt Mackall <mpm@selenic.com>
#
# This software may be used and distributed according to the terms
# of the GNU General Public License, incorporated herein by reference.
from revlog import *
from demandload import *
demandload(globals(), "bdiff os")
class filelog(revlog):
def __init__(self, opener, path, defversion=REVLOG_DEFAULT_VERSION):
revlog.__init__(self, opener,
os.path.join("data", self.encodedir(path + ".i")),
os.path.join("data", self.encodedir(path + ".d")),
defversion)
# This avoids a collision between a file named foo and a dir named
# foo.i or foo.d
def encodedir(self, path):
return (path
.replace(".hg/", ".hg.hg/")
.replace(".i/", ".i.hg/")
.replace(".d/", ".d.hg/"))
def decodedir(self, path):
return (path
.replace(".d.hg/", ".d/")
.replace(".i.hg/", ".i/")
.replace(".hg.hg/", ".hg/"))
def read(self, node):
t = self.revision(node)
if not t.startswith('\1\n'):
return t
s = t.find('\1\n', 2)
return t[s+2:]
def readmeta(self, node):
t = self.revision(node)
if not t.startswith('\1\n'):
return {}
s = t.find('\1\n', 2)
mt = t[2:s]
m = {}
for l in mt.splitlines():
k, v = l.split(": ", 1)
m[k] = v
return m
def add(self, text, meta, transaction, link, p1=None, p2=None):
if meta or text.startswith('\1\n'):
mt = ""
if meta:
mt = [ "%s: %s\n" % (k, v) for k,v in meta.items() ]
text = "\1\n%s\1\n%s" % ("".join(mt), text)
return self.addrevision(text, transaction, link, p1, p2)
def renamed(self, node):
if self.parents(node)[0] != nullid:
return False
m = self.readmeta(node)
if m and m.has_key("copy"):
return (m["copy"], bin(m["copyrev"]))
return False
def annotate(self, node):
def decorate(text, rev):
return ([rev] * len(text.splitlines()), text)
def pair(parent, child):
for a1, a2, b1, b2 in bdiff.blocks(parent[1], child[1]):
child[0][b1:b2] = parent[0][a1:a2]
return child
# find all ancestors
needed = {node:1}
visit = [node]
while visit:
n = visit.pop(0)
for p in self.parents(n):
if p not in needed:
needed[p] = 1
visit.append(p)
else:
# count how many times we'll use this
needed[p] += 1
# sort by revision which is a topological order
visit = [ (self.rev(n), n) for n in needed.keys() ]
visit.sort()
hist = {}
for r,n in visit:
curr = decorate(self.read(n), self.linkrev(n))
for p in self.parents(n):
if p != nullid:
curr = pair(hist[p], curr)
# trim the history of unneeded revs
needed[p] -= 1
if not needed[p]:
del hist[p]
hist[n] = curr
return zip(hist[n][0], hist[n][1].splitlines(1))