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
"""automatically manage newlines in repository files
This extension allows you to manage the type of line endings (CRLF or
LF) that are used in the repository and in the local working
directory. That way you can get CRLF line endings on Windows and LF on
Unix/Mac, thereby letting everybody use their OS native line endings.
The extension reads its configuration from a versioned ``.hgeol``
configuration file found in the root of the working directory. The
``.hgeol`` file use the same syntax as all other Mercurial
configuration files. It uses two sections, ``[patterns]`` and
``[repository]``.
The ``[patterns]`` section specifies how line endings should be
converted between the working directory and the repository. The format is
specified by a file pattern. The first match is used, so put more
specific patterns first. The available line endings are ``LF``,
``CRLF``, and ``BIN``.
Files with the declared format of ``CRLF`` or ``LF`` are always
checked out and stored in the repository in that format and files
declared to be binary (``BIN``) are left unchanged. Additionally,
``native`` is an alias for checking out in the platform's default line
ending: ``LF`` on Unix (including Mac OS X) and ``CRLF`` on
Windows. Note that ``BIN`` (do nothing to line endings) is Mercurial's
default behavior; it is only needed if you need to override a later,
more general pattern.
The optional ``[repository]`` section specifies the line endings to
use for files stored in the repository. It has a single setting,
``native``, which determines the storage line endings for files
declared as ``native`` in the ``[patterns]`` section. It can be set to
``LF`` or ``CRLF``. The default is ``LF``. For example, this means
that on Windows, files configured as ``native`` (``CRLF`` by default)
will be converted to ``LF`` when stored in the repository. Files
declared as ``LF``, ``CRLF``, or ``BIN`` in the ``[patterns]`` section
are always stored as-is in the repository.
Example versioned ``.hgeol`` file::
[patterns]
**.py = native
**.vcproj = CRLF
**.txt = native
Makefile = LF
**.jpg = BIN
[repository]
native = LF
.. note::
The rules will first apply when files are touched in the working
directory, e.g. by updating to null and back to tip to touch all files.
The extension uses an optional ``[eol]`` section read from both the
normal Mercurial configuration files and the ``.hgeol`` file, with the
latter overriding the former. You can use that section to control the
overall behavior. There are three settings:
- ``eol.native`` (default ``os.linesep``) can be set to ``LF`` or
``CRLF`` to override the default interpretation of ``native`` for
checkout. This can be used with :hg:`archive` on Unix, say, to
generate an archive where files have line endings for Windows.
- ``eol.only-consistent`` (default True) can be set to False to make
the extension convert files with inconsistent EOLs. Inconsistent
means that there is both ``CRLF`` and ``LF`` present in the file.
Such files are normally not touched under the assumption that they
have mixed EOLs on purpose.
- ``eol.fix-trailing-newline`` (default False) can be set to True to
ensure that converted files end with a EOL character (either ``\\n``
or ``\\r\\n`` as per the configured patterns).
The extension provides ``cleverencode:`` and ``cleverdecode:`` filters
like the deprecated win32text extension does. This means that you can
disable win32text and enable eol and your filters will still work. You
only need to these filters until you have prepared a ``.hgeol`` file.
The ``win32text.forbid*`` hooks provided by the win32text extension
have been unified into a single hook named ``eol.checkheadshook``. The
hook will lookup the expected line endings from the ``.hgeol`` file,
which means you must migrate to a ``.hgeol`` file first before using
the hook. ``eol.checkheadshook`` only checks heads, intermediate
invalid revisions will be pushed. To forbid them completely, use the
``eol.checkallhook`` hook. These hooks are best used as
``pretxnchangegroup`` hooks.
See :hg:`help patterns` for more information about the glob patterns
used.
"""
from __future__ import absolute_import
import os
import re
from mercurial.i18n import _
from mercurial import (
config,
error as errormod,
extensions,
match,
pycompat,
registrar,
util,
)
# Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for
# extensions which SHIP WITH MERCURIAL. Non-mainline extensions should
# be specifying the version(s) of Mercurial they are tested with, or
# leave the attribute unspecified.
testedwith = 'ships-with-hg-core'
configtable = {}
configitem = registrar.configitem(configtable)
configitem('eol', 'fix-trailing-newline',
default=False,
)
configitem('eol', 'native',
default=pycompat.oslinesep,
)
configitem('eol', 'only-consistent',
default=True,
)
# Matches a lone LF, i.e., one that is not part of CRLF.
singlelf = re.compile('(^|[^\r])\n')
def inconsistenteol(data):
return '\r\n' in data and singlelf.search(data)
def tolf(s, params, ui, **kwargs):
"""Filter to convert to LF EOLs."""
if util.binary(s):
return s
if ui.configbool('eol', 'only-consistent') and inconsistenteol(s):
return s
if (ui.configbool('eol', 'fix-trailing-newline')
and s and s[-1] != '\n'):
s = s + '\n'
return util.tolf(s)
def tocrlf(s, params, ui, **kwargs):
"""Filter to convert to CRLF EOLs."""
if util.binary(s):
return s
if ui.configbool('eol', 'only-consistent') and inconsistenteol(s):
return s
if (ui.configbool('eol', 'fix-trailing-newline')
and s and s[-1] != '\n'):
s = s + '\n'
return util.tocrlf(s)
def isbinary(s, params):
"""Filter to do nothing with the file."""
return s
filters = {
'to-lf': tolf,
'to-crlf': tocrlf,
'is-binary': isbinary,
# The following provide backwards compatibility with win32text
'cleverencode:': tolf,
'cleverdecode:': tocrlf
}
class eolfile(object):
def __init__(self, ui, root, data):
self._decode = {'LF': 'to-lf', 'CRLF': 'to-crlf', 'BIN': 'is-binary'}
self._encode = {'LF': 'to-lf', 'CRLF': 'to-crlf', 'BIN': 'is-binary'}
self.cfg = config.config()
# Our files should not be touched. The pattern must be
# inserted first override a '** = native' pattern.
self.cfg.set('patterns', '.hg*', 'BIN', 'eol')
# We can then parse the user's patterns.
self.cfg.parse('.hgeol', data)
isrepolf = self.cfg.get('repository', 'native') != 'CRLF'
self._encode['NATIVE'] = isrepolf and 'to-lf' or 'to-crlf'
iswdlf = ui.config('eol', 'native') in ('LF', '\n')
self._decode['NATIVE'] = iswdlf and 'to-lf' or 'to-crlf'
include = []
exclude = []
self.patterns = []
for pattern, style in self.cfg.items('patterns'):
key = style.upper()
if key == 'BIN':
exclude.append(pattern)
else:
include.append(pattern)
m = match.match(root, '', [pattern])
self.patterns.append((pattern, key, m))
# This will match the files for which we need to care
# about inconsistent newlines.
self.match = match.match(root, '', [], include, exclude)
def copytoui(self, ui):
for pattern, key, m in self.patterns:
try:
ui.setconfig('decode', pattern, self._decode[key], 'eol')
ui.setconfig('encode', pattern, self._encode[key], 'eol')
except KeyError:
ui.warn(_("ignoring unknown EOL style '%s' from %s\n")
% (key, self.cfg.source('patterns', pattern)))
# eol.only-consistent can be specified in ~/.hgrc or .hgeol
for k, v in self.cfg.items('eol'):
ui.setconfig('eol', k, v, 'eol')
def checkrev(self, repo, ctx, files):
failed = []
for f in (files or ctx.files()):
if f not in ctx:
continue
for pattern, key, m in self.patterns:
if not m(f):
continue
target = self._encode[key]
data = ctx[f].data()
if (target == "to-lf" and "\r\n" in data
or target == "to-crlf" and singlelf.search(data)):
failed.append((f, target, bytes(ctx)))
break
return failed
def parseeol(ui, repo, nodes):
try:
for node in nodes:
try:
if node is None:
# Cannot use workingctx.data() since it would load
# and cache the filters before we configure them.
data = repo.wvfs('.hgeol').read()
else:
data = repo[node]['.hgeol'].data()
return eolfile(ui, repo.root, data)
except (IOError, LookupError):
pass
except errormod.ParseError as inst:
ui.warn(_("warning: ignoring .hgeol file due to parse error "
"at %s: %s\n") % (inst.args[1], inst.args[0]))
return None
def ensureenabled(ui):
"""make sure the extension is enabled when used as hook
When eol is used through hooks, the extension is never formally loaded and
enabled. This has some side effect, for example the config declaration is
never loaded. This function ensure the extension is enabled when running
hooks.
"""
if 'eol' in ui._knownconfig:
return
ui.setconfig('extensions', 'eol', '', source='internal')
extensions.loadall(ui, ['eol'])
def _checkhook(ui, repo, node, headsonly):
# Get revisions to check and touched files at the same time
ensureenabled(ui)
files = set()
revs = set()
for rev in xrange(repo[node].rev(), len(repo)):
revs.add(rev)
if headsonly:
ctx = repo[rev]
files.update(ctx.files())
for pctx in ctx.parents():
revs.discard(pctx.rev())
failed = []
for rev in revs:
ctx = repo[rev]
eol = parseeol(ui, repo, [ctx.node()])
if eol:
failed.extend(eol.checkrev(repo, ctx, files))
if failed:
eols = {'to-lf': 'CRLF', 'to-crlf': 'LF'}
msgs = []
for f, target, node in sorted(failed):
msgs.append(_(" %s in %s should not have %s line endings") %
(f, node, eols[target]))
raise errormod.Abort(_("end-of-line check failed:\n") + "\n".join(msgs))
def checkallhook(ui, repo, node, hooktype, **kwargs):
"""verify that files have expected EOLs"""
_checkhook(ui, repo, node, False)
def checkheadshook(ui, repo, node, hooktype, **kwargs):
"""verify that files have expected EOLs"""
_checkhook(ui, repo, node, True)
# "checkheadshook" used to be called "hook"
hook = checkheadshook
def preupdate(ui, repo, hooktype, parent1, parent2):
repo.loadeol([parent1])
return False
def uisetup(ui):
ui.setconfig('hooks', 'preupdate.eol', preupdate, 'eol')
def extsetup(ui):
try:
extensions.find('win32text')
ui.warn(_("the eol extension is incompatible with the "
"win32text extension\n"))
except KeyError:
pass
def reposetup(ui, repo):
uisetup(repo.ui)
if not repo.local():
return
for name, fn in filters.iteritems():
repo.adddatafilter(name, fn)
ui.setconfig('patch', 'eol', 'auto', 'eol')
class eolrepo(repo.__class__):
def loadeol(self, nodes):
eol = parseeol(self.ui, self, nodes)
if eol is None:
return None
eol.copytoui(self.ui)
return eol.match
def _hgcleardirstate(self):
self._eolmatch = self.loadeol([None, 'tip'])
if not self._eolmatch:
self._eolmatch = util.never
return
oldeol = None
try:
cachemtime = os.path.getmtime(self.vfs.join("eol.cache"))
except OSError:
cachemtime = 0
else:
olddata = self.vfs.read("eol.cache")
if olddata:
oldeol = eolfile(self.ui, self.root, olddata)
try:
eolmtime = os.path.getmtime(self.wjoin(".hgeol"))
except OSError:
eolmtime = 0
if eolmtime > cachemtime:
self.ui.debug("eol: detected change in .hgeol\n")
hgeoldata = self.wvfs.read('.hgeol')
neweol = eolfile(self.ui, self.root, hgeoldata)
wlock = None
try:
wlock = self.wlock()
for f in self.dirstate:
if self.dirstate[f] != 'n':
continue
if oldeol is not None:
if not oldeol.match(f) and not neweol.match(f):
continue
oldkey = None
for pattern, key, m in oldeol.patterns:
if m(f):
oldkey = key
break
newkey = None
for pattern, key, m in neweol.patterns:
if m(f):
newkey = key
break
if oldkey == newkey:
continue
# all normal files need to be looked at again since
# the new .hgeol file specify a different filter
self.dirstate.normallookup(f)
# Write the cache to update mtime and cache .hgeol
with self.vfs("eol.cache", "w") as f:
f.write(hgeoldata)
except errormod.LockUnavailable:
# If we cannot lock the repository and clear the
# dirstate, then a commit might not see all files
# as modified. But if we cannot lock the
# repository, then we can also not make a commit,
# so ignore the error.
pass
finally:
if wlock is not None:
wlock.release()
def commitctx(self, ctx, error=False):
for f in sorted(ctx.added() + ctx.modified()):
if not self._eolmatch(f):
continue
fctx = ctx[f]
if fctx is None:
continue
data = fctx.data()
if util.binary(data):
# We should not abort here, since the user should
# be able to say "** = native" to automatically
# have all non-binary files taken care of.
continue
if inconsistenteol(data):
raise errormod.Abort(_("inconsistent newline style "
"in %s\n") % f)
return super(eolrepo, self).commitctx(ctx, error)
repo.__class__ = eolrepo
repo._hgcleardirstate()