xdiff: add a preprocessing step that trims files
xdiff has a `xdl_trim_ends` step that removes common lines, unmatchable
lines. That is in theory good, but happens too late - after splitting,
hashing, and adjusting the hash values so they are unique. Those splitting,
hashing and adjusting hash values steps could have noticeable overhead.
Diffing two large files with minor (one-line-ish) changes are not uncommon.
In that case, the raw performance of those preparation steps seriously
matter. Even allocating an O(N) array and storing line offsets to it is
expensive. Therefore my previous attempts [1] [2] cannot be good enough
since they do not remove the O(N) array assignment.
This patch adds a preprocessing step - `xdl_trim_files` that runs before
other preprocessing steps. It counts common prefix and suffix and lines in
them (needed for displaying line number), without doing anything else.
Testing with a crafted large (169MB) file, with minor change:
```
open('a','w').write(''.join('%s\n' % (i % 100000) for i in xrange(
30000000) if i != 6000000))
open('b','w').write(''.join('%s\n' % (i % 100000) for i in xrange(
30000000) if i != 6003000))
```
Running xdiff by a simple binary [3], this patch improves the xdiff perf by
more than 10x for the above case:
```
# xdiff before this patch
2.41s user 1.13s system 98% cpu 3.592 total
# xdiff after this patch
0.14s user 0.16s system 98% cpu 0.309 total
# gnu diffutils
0.12s user 0.15s system 98% cpu 0.272 total
# (best of 20 runs)
```
It's still slightly slower than GNU diffutils. But it's pretty close now.
Testing with real repo data:
For the whole repo, this patch makes xdiff 25% faster:
```
# hg perfbdiff --count 100 --alldata -c
d334afc585e2 --blocks [--xdiff]
# xdiff, after
! wall 0.058861 comb 0.050000 user 0.050000 sys 0.000000 (best of 100)
# xdiff, before
! wall 0.077816 comb 0.080000 user 0.080000 sys 0.000000 (best of 91)
# bdiff
! wall 0.117473 comb 0.120000 user 0.120000 sys 0.000000 (best of 67)
```
For files that are long (ex. commands.py), the speedup is more than 3x, very
significant:
```
# hg perfbdiff --count 3000 --blocks commands.py.i 1 [--xdiff]
# xdiff, after
! wall 0.690583 comb 0.690000 user 0.690000 sys 0.000000 (best of 12)
# xdiff, before
! wall 2.240361 comb 2.210000 user 2.210000 sys 0.000000 (best of 4)
# bdiff
! wall 2.469852 comb 2.440000 user 2.440000 sys 0.000000 (best of 4)
```
[1]: https://phab.mercurial-scm.org/D2631
[2]: https://phab.mercurial-scm.org/D2634
[3]:
```
// Code to run xdiff from command line. No proper error handling.
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include "mercurial/thirdparty/xdiff/xdiff.h"
#define ensure(x) if (!(x)) exit(255);
mmfile_t readfile(const char *path) {
struct stat st; int fd = open(path, O_RDONLY);
fstat(fd, &st); mmfile_t file = { malloc(st.st_size), st.st_size };
ensure(read(fd, file.ptr, st.st_size) == st.st_size); close(fd);
return file;
}
int main(int argc, char const *argv[]) {
mmfile_t a = readfile(argv[1]), b = readfile(argv[2]);
xpparam_t xpp = {0}; xdemitconf_t xecfg = {0}; xdemitcb_t ecb = {0};
xdl_diff(&a, &b, &xpp, &xecfg, &ecb);
return 0;
}
```
Differential Revision: https://phab.mercurial-scm.org/D2686
# darcs.py - darcs support for the convert extension
#
# Copyright 2007-2009 Matt Mackall <mpm@selenic.com> and others
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
from __future__ import absolute_import
import errno
import os
import re
import shutil
import tempfile
from mercurial.i18n import _
from mercurial import (
error,
util,
)
from mercurial.utils import dateutil
from . import common
NoRepo = common.NoRepo
# The naming drift of ElementTree is fun!
try:
import xml.etree.cElementTree.ElementTree as ElementTree
import xml.etree.cElementTree.XMLParser as XMLParser
except ImportError:
try:
import xml.etree.ElementTree.ElementTree as ElementTree
import xml.etree.ElementTree.XMLParser as XMLParser
except ImportError:
try:
import elementtree.cElementTree.ElementTree as ElementTree
import elementtree.cElementTree.XMLParser as XMLParser
except ImportError:
try:
import elementtree.ElementTree.ElementTree as ElementTree
import elementtree.ElementTree.XMLParser as XMLParser
except ImportError:
pass
class darcs_source(common.converter_source, common.commandline):
def __init__(self, ui, repotype, path, revs=None):
common.converter_source.__init__(self, ui, repotype, path, revs=revs)
common.commandline.__init__(self, ui, 'darcs')
# check for _darcs, ElementTree so that we can easily skip
# test-convert-darcs if ElementTree is not around
if not os.path.exists(os.path.join(path, '_darcs')):
raise NoRepo(_("%s does not look like a darcs repository") % path)
common.checktool('darcs')
version = self.run0('--version').splitlines()[0].strip()
if version < '2.1':
raise error.Abort(_('darcs version 2.1 or newer needed (found %r)')
% version)
if "ElementTree" not in globals():
raise error.Abort(_("Python ElementTree module is not available"))
self.path = os.path.realpath(path)
self.lastrev = None
self.changes = {}
self.parents = {}
self.tags = {}
# Check darcs repository format
format = self.format()
if format:
if format in ('darcs-1.0', 'hashed'):
raise NoRepo(_("%s repository format is unsupported, "
"please upgrade") % format)
else:
self.ui.warn(_('failed to detect repository format!'))
def before(self):
self.tmppath = tempfile.mkdtemp(
prefix='convert-' + os.path.basename(self.path) + '-')
output, status = self.run('init', repodir=self.tmppath)
self.checkexit(status)
tree = self.xml('changes', xml_output=True, summary=True,
repodir=self.path)
tagname = None
child = None
for elt in tree.findall('patch'):
node = elt.get('hash')
name = elt.findtext('name', '')
if name.startswith('TAG '):
tagname = name[4:].strip()
elif tagname is not None:
self.tags[tagname] = node
tagname = None
self.changes[node] = elt
self.parents[child] = [node]
child = node
self.parents[child] = []
def after(self):
self.ui.debug('cleaning up %s\n' % self.tmppath)
shutil.rmtree(self.tmppath, ignore_errors=True)
def recode(self, s, encoding=None):
if isinstance(s, unicode):
# XMLParser returns unicode objects for anything it can't
# encode into ASCII. We convert them back to str to get
# recode's normal conversion behavior.
s = s.encode('latin-1')
return super(darcs_source, self).recode(s, encoding)
def xml(self, cmd, **kwargs):
# NOTE: darcs is currently encoding agnostic and will print
# patch metadata byte-for-byte, even in the XML changelog.
etree = ElementTree()
# While we are decoding the XML as latin-1 to be as liberal as
# possible, etree will still raise an exception if any
# non-printable characters are in the XML changelog.
parser = XMLParser(encoding='latin-1')
p = self._run(cmd, **kwargs)
etree.parse(p.stdout, parser=parser)
p.wait()
self.checkexit(p.returncode)
return etree.getroot()
def format(self):
output, status = self.run('show', 'repo', no_files=True,
repodir=self.path)
self.checkexit(status)
m = re.search(r'^\s*Format:\s*(.*)$', output, re.MULTILINE)
if not m:
return None
return ','.join(sorted(f.strip() for f in m.group(1).split(',')))
def manifest(self):
man = []
output, status = self.run('show', 'files', no_directories=True,
repodir=self.tmppath)
self.checkexit(status)
for line in output.split('\n'):
path = line[2:]
if path:
man.append(path)
return man
def getheads(self):
return self.parents[None]
def getcommit(self, rev):
elt = self.changes[rev]
dateformat = '%a %b %d %H:%M:%S %Z %Y'
date = dateutil.strdate(elt.get('local_date'), dateformat)
desc = elt.findtext('name') + '\n' + elt.findtext('comment', '')
# etree can return unicode objects for name, comment, and author,
# so recode() is used to ensure str objects are emitted.
newdateformat = '%Y-%m-%d %H:%M:%S %1%2'
return common.commit(author=self.recode(elt.get('author')),
date=dateutil.datestr(date, newdateformat),
desc=self.recode(desc).strip(),
parents=self.parents[rev])
def pull(self, rev):
output, status = self.run('pull', self.path, all=True,
match='hash %s' % rev,
no_test=True, no_posthook=True,
external_merge='/bin/false',
repodir=self.tmppath)
if status:
if output.find('We have conflicts in') == -1:
self.checkexit(status, output)
output, status = self.run('revert', all=True, repodir=self.tmppath)
self.checkexit(status, output)
def getchanges(self, rev, full):
if full:
raise error.Abort(_("convert from darcs does not support --full"))
copies = {}
changes = []
man = None
for elt in self.changes[rev].find('summary').getchildren():
if elt.tag in ('add_directory', 'remove_directory'):
continue
if elt.tag == 'move':
if man is None:
man = self.manifest()
source, dest = elt.get('from'), elt.get('to')
if source in man:
# File move
changes.append((source, rev))
changes.append((dest, rev))
copies[dest] = source
else:
# Directory move, deduce file moves from manifest
source = source + '/'
for f in man:
if not f.startswith(source):
continue
fdest = dest + '/' + f[len(source):]
changes.append((f, rev))
changes.append((fdest, rev))
copies[fdest] = f
else:
changes.append((elt.text.strip(), rev))
self.pull(rev)
self.lastrev = rev
return sorted(changes), copies, set()
def getfile(self, name, rev):
if rev != self.lastrev:
raise error.Abort(_('internal calling inconsistency'))
path = os.path.join(self.tmppath, name)
try:
data = util.readfile(path)
mode = os.lstat(path).st_mode
except IOError as inst:
if inst.errno == errno.ENOENT:
return None, None
raise
mode = (mode & 0o111) and 'x' or ''
return data, mode
def gettags(self):
return self.tags