Mercurial > hg
view mercurial/store.py @ 42044:bb271ec2fbfb
compression: introduce a `storage.revlog.zstd.level` configuration
This option control the zstd compression level used when compressing revlog
chunk. The usage of zstd for revlog compression has not graduated from
experimental yet, but we intend to fix that soon.
The option name for the compression level is more straight forward to pick, so
this changesets comes first. Having a dedicated option for each compression
engine is useful because they don't support the same range of values.
I ran the same measurement as for the zlib compression level (in the parent
changesets). The variation in repository size is stay mostly in the same (small)
range. The "read/write" performance see smallish variation, but are overall much
better than zlib. Write performance show the same tend of having better write
performance for when reaching high-end compression.
Again, we don't intend to change the default zstd compression level (currently:
3) in this series. However this is worth investigating in the future.
The Performance comparison of zlib vs zstd is quite impressive. The repository
size stay in the same range, but the performance are much better in all
situations.
Comparison summary
==================
We are looking at:
- performance range for zlib
- performance range for zstd
- comparison of default zstd (level-3) to default zlib (level 6)
- comparison of the slowest zstd time to the fastest zlib time
Read performance:
-----------------
| zlib | zstd | cmp | f2s
mercurial | 0.170159 - 0.189219 | 0.144127 - 0.149624 | 80% | 88%
pypy | 2.679217 - 2.768691 | 1.532317 - 1.705044 | 60% | 63%
netbeans | 122.477027 - 141.620281 | 72.996346 - 89.731560 | 58% | 73%
mozilla | 147.867662 - 170.572118 | 91.700995 - 105.853099 | 56% | 71%
Write performance:
------------------
| zlib | zstd | cmp | f2s
mercurial | 53.250304 - 56.2936129 | 40.877025 - 45.677286 | 75% | 86%
pypy | 460.721984 - 476.589918 | 270.545409 - 301.002219 | 63% | 65%
netbeans | 520.560316 - 715.930400 | 370.356311 - 428.329652 | 55% | 82%
mozilla | 739.803002 - 987.056093 | 505.152906 - 591.930683 | 57% | 80%
Raw data
--------
repo alg lvl .hg/store size 00manifest.d read write
mercurial zlib 1 49,402,813 5,963,475 0.170159 53.250304
mercurial zlib 6 47,197,397 5,875,730 0.182820 56.264320
mercurial zlib 9 47,121,596 5,849,781 0.189219 56.293612
mercurial zstd 1 49,737,084 5,966,355 0.144127 40.877025
mercurial zstd 3 48,961,867 5,895,208 0.146376 42.268142
mercurial zstd 5 48,200,592 5,938,676 0.149624 43.162875
mercurial zstd 10 47,833,520 5,913,353 0.145185 44.012489
mercurial zstd 15 47,314,604 5,728,679 0.147686 45.677286
mercurial zstd 20 47,330,502 5,830,539 0.145789 45.025407
mercurial zstd 22 47,330,076 5,830,539 0.143996 44.690460
pypy zlib 1 370,830,572 28,462,425 2.679217 460.721984
pypy zlib 6 340,112,317 27,648,747 2.768691 467.537158
pypy zlib 9 338,360,736 27,639,003 2.763495 476.589918
pypy zstd 1 362,377,479 27,916,214 1.532317 270.545409
pypy zstd 3 354,137,693 27,905,988 1.686718 294.951509
pypy zstd 5 342,640,043 27,655,774 1.705044 301.002219
pypy zstd 10 334,224,327 27,164,493 1.567287 285.186239
pypy zstd 15 329,000,363 26,645,965 1.637729 299.561332
pypy zstd 20 324,534,039 26,199,547 1.526813 302.149827
pypy zstd 22 324,530,595 26,198,932 1.525718 307.821218
netbeans zlib 1 1,281,847,810 165,495,457 122.477027 520.560316
netbeans zlib 6 1,205,284,353 159,161,207 139.876147 715.930400
netbeans zlib 9 1,197,135,671 155,034,586 141.620281 678.297064
netbeans zstd 1 1,259,581,737 160,840,613 72.996346 370.356311
netbeans zstd 3 1,232,978,122 157,691,551 81.622317 396.733087
netbeans zstd 5 1,208,034,075 160,246,880 83.080549 364.342626
netbeans zstd 10 1,188,624,176 156,083,417 79.323935 403.594602
netbeans zstd 15 1,176,973,589 153,859,477 89.731560 428.329652
netbeans zstd 20 1,162,958,258 151,147,535 82.842667 392.335349
netbeans zstd 22 1,162,707,029 151,150,220 82.565695 402.840655
mozilla zlib 1 2,775,497,186 298,527,987 147.867662 751.263721
mozilla zlib 6 2,596,856,420 286,597,671 170.572118 987.056093
mozilla zlib 9 2,587,542,494 287,018,264 163.622338 739.803002
mozilla zstd 1 2,723,159,348 286,617,532 91.700995 570.042751
mozilla zstd 3 2,665,055,001 286,152,013 95.240155 561.412805
mozilla zstd 5 2,607,819,817 288,060,030 101.978048 505.152906
mozilla zstd 10 2,558,761,085 283,967,648 104.113481 497.771202
mozilla zstd 15 2,526,216,060 275,581,300 105.853099 591.930683
mozilla zstd 20 2,485,114,806 266,478,859 95.268795 576.515389
mozilla zstd 22 2,484,869,080 266,456,505 94.429282 572.785537
author | Pierre-Yves David <pierre-yves.david@octobus.net> |
---|---|
date | Wed, 27 Mar 2019 18:35:59 +0100 |
parents | 3e7cfa17df5d |
children | 526750cdd02d |
line wrap: on
line source
# store.py - repository store handling for Mercurial # # Copyright 2008 Matt Mackall <mpm@selenic.com> # # 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 functools import hashlib import os import stat from .i18n import _ from . import ( error, node, policy, pycompat, util, vfs as vfsmod, ) parsers = policy.importmod(r'parsers') # how much bytes should be read from fncache in one read # It is done to prevent loading large fncache files into memory fncache_chunksize = 10 ** 6 def _matchtrackedpath(path, matcher): """parses a fncache entry and returns whether the entry is tracking a path matched by matcher or not. If matcher is None, returns True""" if matcher is None: return True path = decodedir(path) if path.startswith('data/'): return matcher(path[len('data/'):-len('.i')]) elif path.startswith('meta/'): return matcher.visitdir(path[len('meta/'):-len('/00manifest.i')] or '.') raise error.ProgrammingError("cannot decode path %s" % path) # This avoids a collision between a file named foo and a dir named # foo.i or foo.d def _encodedir(path): ''' >>> _encodedir(b'data/foo.i') 'data/foo.i' >>> _encodedir(b'data/foo.i/bla.i') 'data/foo.i.hg/bla.i' >>> _encodedir(b'data/foo.i.hg/bla.i') 'data/foo.i.hg.hg/bla.i' >>> _encodedir(b'data/foo.i\\ndata/foo.i/bla.i\\ndata/foo.i.hg/bla.i\\n') 'data/foo.i\\ndata/foo.i.hg/bla.i\\ndata/foo.i.hg.hg/bla.i\\n' ''' return (path .replace(".hg/", ".hg.hg/") .replace(".i/", ".i.hg/") .replace(".d/", ".d.hg/")) encodedir = getattr(parsers, 'encodedir', _encodedir) def decodedir(path): ''' >>> decodedir(b'data/foo.i') 'data/foo.i' >>> decodedir(b'data/foo.i.hg/bla.i') 'data/foo.i/bla.i' >>> decodedir(b'data/foo.i.hg.hg/bla.i') 'data/foo.i.hg/bla.i' ''' if ".hg/" not in path: return path return (path .replace(".d.hg/", ".d/") .replace(".i.hg/", ".i/") .replace(".hg.hg/", ".hg/")) def _reserved(): ''' characters that are problematic for filesystems * ascii escapes (0..31) * ascii hi (126..255) * windows specials these characters will be escaped by encodefunctions ''' winreserved = [ord(x) for x in u'\\:*?"<>|'] for x in range(32): yield x for x in range(126, 256): yield x for x in winreserved: yield x def _buildencodefun(): ''' >>> enc, dec = _buildencodefun() >>> enc(b'nothing/special.txt') 'nothing/special.txt' >>> dec(b'nothing/special.txt') 'nothing/special.txt' >>> enc(b'HELLO') '_h_e_l_l_o' >>> dec(b'_h_e_l_l_o') 'HELLO' >>> enc(b'hello:world?') 'hello~3aworld~3f' >>> dec(b'hello~3aworld~3f') 'hello:world?' >>> enc(b'the\\x07quick\\xADshot') 'the~07quick~adshot' >>> dec(b'the~07quick~adshot') 'the\\x07quick\\xadshot' ''' e = '_' xchr = pycompat.bytechr asciistr = list(map(xchr, range(127))) capitals = list(range(ord("A"), ord("Z") + 1)) cmap = dict((x, x) for x in asciistr) for x in _reserved(): cmap[xchr(x)] = "~%02x" % x for x in capitals + [ord(e)]: cmap[xchr(x)] = e + xchr(x).lower() dmap = {} for k, v in cmap.iteritems(): dmap[v] = k def decode(s): i = 0 while i < len(s): for l in pycompat.xrange(1, 4): try: yield dmap[s[i:i + l]] i += l break except KeyError: pass else: raise KeyError return (lambda s: ''.join([cmap[s[c:c + 1]] for c in pycompat.xrange(len(s))]), lambda s: ''.join(list(decode(s)))) _encodefname, _decodefname = _buildencodefun() def encodefilename(s): ''' >>> encodefilename(b'foo.i/bar.d/bla.hg/hi:world?/HELLO') 'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o' ''' return _encodefname(encodedir(s)) def decodefilename(s): ''' >>> decodefilename(b'foo.i.hg/bar.d.hg/bla.hg.hg/hi~3aworld~3f/_h_e_l_l_o') 'foo.i/bar.d/bla.hg/hi:world?/HELLO' ''' return decodedir(_decodefname(s)) def _buildlowerencodefun(): ''' >>> f = _buildlowerencodefun() >>> f(b'nothing/special.txt') 'nothing/special.txt' >>> f(b'HELLO') 'hello' >>> f(b'hello:world?') 'hello~3aworld~3f' >>> f(b'the\\x07quick\\xADshot') 'the~07quick~adshot' ''' xchr = pycompat.bytechr cmap = dict([(xchr(x), xchr(x)) for x in pycompat.xrange(127)]) for x in _reserved(): cmap[xchr(x)] = "~%02x" % x for x in range(ord("A"), ord("Z") + 1): cmap[xchr(x)] = xchr(x).lower() def lowerencode(s): return "".join([cmap[c] for c in pycompat.iterbytestr(s)]) return lowerencode lowerencode = getattr(parsers, 'lowerencode', None) or _buildlowerencodefun() # Windows reserved names: con, prn, aux, nul, com1..com9, lpt1..lpt9 _winres3 = ('aux', 'con', 'prn', 'nul') # length 3 _winres4 = ('com', 'lpt') # length 4 (with trailing 1..9) def _auxencode(path, dotencode): ''' Encodes filenames containing names reserved by Windows or which end in period or space. Does not touch other single reserved characters c. Specifically, c in '\\:*?"<>|' or ord(c) <= 31 are *not* encoded here. Additionally encodes space or period at the beginning, if dotencode is True. Parameter path is assumed to be all lowercase. A segment only needs encoding if a reserved name appears as a basename (e.g. "aux", "aux.foo"). A directory or file named "foo.aux" doesn't need encoding. >>> s = b'.foo/aux.txt/txt.aux/con/prn/nul/foo.' >>> _auxencode(s.split(b'/'), True) ['~2efoo', 'au~78.txt', 'txt.aux', 'co~6e', 'pr~6e', 'nu~6c', 'foo~2e'] >>> s = b'.com1com2/lpt9.lpt4.lpt1/conprn/com0/lpt0/foo.' >>> _auxencode(s.split(b'/'), False) ['.com1com2', 'lp~749.lpt4.lpt1', 'conprn', 'com0', 'lpt0', 'foo~2e'] >>> _auxencode([b'foo. '], True) ['foo.~20'] >>> _auxencode([b' .foo'], True) ['~20.foo'] ''' for i, n in enumerate(path): if not n: continue if dotencode and n[0] in '. ': n = "~%02x" % ord(n[0:1]) + n[1:] path[i] = n else: l = n.find('.') if l == -1: l = len(n) if ((l == 3 and n[:3] in _winres3) or (l == 4 and n[3:4] <= '9' and n[3:4] >= '1' and n[:3] in _winres4)): # encode third letter ('aux' -> 'au~78') ec = "~%02x" % ord(n[2:3]) n = n[0:2] + ec + n[3:] path[i] = n if n[-1] in '. ': # encode last period or space ('foo...' -> 'foo..~2e') path[i] = n[:-1] + "~%02x" % ord(n[-1:]) return path _maxstorepathlen = 120 _dirprefixlen = 8 _maxshortdirslen = 8 * (_dirprefixlen + 1) - 4 def _hashencode(path, dotencode): digest = node.hex(hashlib.sha1(path).digest()) le = lowerencode(path[5:]).split('/') # skips prefix 'data/' or 'meta/' parts = _auxencode(le, dotencode) basename = parts[-1] _root, ext = os.path.splitext(basename) sdirs = [] sdirslen = 0 for p in parts[:-1]: d = p[:_dirprefixlen] if d[-1] in '. ': # Windows can't access dirs ending in period or space d = d[:-1] + '_' if sdirslen == 0: t = len(d) else: t = sdirslen + 1 + len(d) if t > _maxshortdirslen: break sdirs.append(d) sdirslen = t dirs = '/'.join(sdirs) if len(dirs) > 0: dirs += '/' res = 'dh/' + dirs + digest + ext spaceleft = _maxstorepathlen - len(res) if spaceleft > 0: filler = basename[:spaceleft] res = 'dh/' + dirs + filler + digest + ext return res def _hybridencode(path, dotencode): '''encodes path with a length limit Encodes all paths that begin with 'data/', according to the following. Default encoding (reversible): Encodes all uppercase letters 'X' as '_x'. All reserved or illegal characters are encoded as '~xx', where xx is the two digit hex code of the character (see encodefilename). Relevant path components consisting of Windows reserved filenames are masked by encoding the third character ('aux' -> 'au~78', see _auxencode). Hashed encoding (not reversible): If the default-encoded path is longer than _maxstorepathlen, a non-reversible hybrid hashing of the path is done instead. This encoding uses up to _dirprefixlen characters of all directory levels of the lowerencoded path, but not more levels than can fit into _maxshortdirslen. Then follows the filler followed by the sha digest of the full path. The filler is the beginning of the basename of the lowerencoded path (the basename is everything after the last path separator). The filler is as long as possible, filling in characters from the basename until the encoded path has _maxstorepathlen characters (or all chars of the basename have been taken). The extension (e.g. '.i' or '.d') is preserved. The string 'data/' at the beginning is replaced with 'dh/', if the hashed encoding was used. ''' path = encodedir(path) ef = _encodefname(path).split('/') res = '/'.join(_auxencode(ef, dotencode)) if len(res) > _maxstorepathlen: res = _hashencode(path, dotencode) return res def _pathencode(path): de = encodedir(path) if len(path) > _maxstorepathlen: return _hashencode(de, True) ef = _encodefname(de).split('/') res = '/'.join(_auxencode(ef, True)) if len(res) > _maxstorepathlen: return _hashencode(de, True) return res _pathencode = getattr(parsers, 'pathencode', _pathencode) def _plainhybridencode(f): return _hybridencode(f, False) def _calcmode(vfs): try: # files in .hg/ will be created using this mode mode = vfs.stat().st_mode # avoid some useless chmods if (0o777 & ~util.umask) == (0o777 & mode): mode = None except OSError: mode = None return mode _data = ('narrowspec data meta 00manifest.d 00manifest.i' ' 00changelog.d 00changelog.i phaseroots obsstore') def isrevlog(f, kind, st): return kind == stat.S_IFREG and f[-2:] in ('.i', '.d') class basicstore(object): '''base class for local repository stores''' def __init__(self, path, vfstype): vfs = vfstype(path) self.path = vfs.base self.createmode = _calcmode(vfs) vfs.createmode = self.createmode self.rawvfs = vfs self.vfs = vfsmod.filtervfs(vfs, encodedir) self.opener = self.vfs def join(self, f): return self.path + '/' + encodedir(f) def _walk(self, relpath, recurse, filefilter=isrevlog): '''yields (unencoded, encoded, size)''' path = self.path if relpath: path += '/' + relpath striplen = len(self.path) + 1 l = [] if self.rawvfs.isdir(path): visit = [path] readdir = self.rawvfs.readdir while visit: p = visit.pop() for f, kind, st in readdir(p, stat=True): fp = p + '/' + f if filefilter(f, kind, st): n = util.pconvert(fp[striplen:]) l.append((decodedir(n), n, st.st_size)) elif kind == stat.S_IFDIR and recurse: visit.append(fp) l.sort() return l def datafiles(self, matcher=None): return self._walk('data', True) + self._walk('meta', True) def topfiles(self): # yield manifest before changelog return reversed(self._walk('', False)) def walk(self, matcher=None): '''yields (unencoded, encoded, size) if a matcher is passed, storage files of only those tracked paths are passed with matches the matcher ''' # yield data files first for x in self.datafiles(matcher): yield x for x in self.topfiles(): yield x def copylist(self): return ['requires'] + _data.split() def write(self, tr): pass def invalidatecaches(self): pass def markremoved(self, fn): pass def __contains__(self, path): '''Checks if the store contains path''' path = "/".join(("data", path)) # file? if self.vfs.exists(path + ".i"): return True # dir? if not path.endswith("/"): path = path + "/" return self.vfs.exists(path) class encodedstore(basicstore): def __init__(self, path, vfstype): vfs = vfstype(path + '/store') self.path = vfs.base self.createmode = _calcmode(vfs) vfs.createmode = self.createmode self.rawvfs = vfs self.vfs = vfsmod.filtervfs(vfs, encodefilename) self.opener = self.vfs def datafiles(self, matcher=None): for a, b, size in super(encodedstore, self).datafiles(): try: a = decodefilename(a) except KeyError: a = None if a is not None and not _matchtrackedpath(a, matcher): continue yield a, b, size def join(self, f): return self.path + '/' + encodefilename(f) def copylist(self): return (['requires', '00changelog.i'] + ['store/' + f for f in _data.split()]) class fncache(object): # the filename used to be partially encoded # hence the encodedir/decodedir dance def __init__(self, vfs): self.vfs = vfs self.entries = None self._dirty = False # set of new additions to fncache self.addls = set() def _load(self): '''fill the entries from the fncache file''' self._dirty = False try: fp = self.vfs('fncache', mode='rb') except IOError: # skip nonexistent file self.entries = set() return self.entries = set() chunk = b'' for c in iter(functools.partial(fp.read, fncache_chunksize), b''): chunk += c try: p = chunk.rindex(b'\n') self.entries.update(decodedir(chunk[:p + 1]).splitlines()) chunk = chunk[p + 1:] except ValueError: # substring '\n' not found, maybe the entry is bigger than the # chunksize, so let's keep iterating pass if chunk: raise error.Abort(_("fncache does not ends with a newline"), hint=_("use 'hg debugrebuildfncache' to rebuild" " the fncache")) self._checkentries(fp) fp.close() def _checkentries(self, fp): """ make sure there is no empty string in entries """ if '' in self.entries: fp.seek(0) for n, line in enumerate(util.iterfile(fp)): if not line.rstrip('\n'): t = _('invalid entry in fncache, line %d') % (n + 1) raise error.Abort(t) def write(self, tr): if self._dirty: assert self.entries is not None self.entries = self.entries | self.addls self.addls = set() tr.addbackup('fncache') fp = self.vfs('fncache', mode='wb', atomictemp=True) if self.entries: fp.write(encodedir('\n'.join(self.entries) + '\n')) fp.close() self._dirty = False if self.addls: # if we have just new entries, let's append them to the fncache tr.addbackup('fncache') fp = self.vfs('fncache', mode='ab', atomictemp=True) if self.addls: fp.write(encodedir('\n'.join(self.addls) + '\n')) fp.close() self.entries = None self.addls = set() def add(self, fn): if self.entries is None: self._load() if fn not in self.entries: self.addls.add(fn) def remove(self, fn): if self.entries is None: self._load() if fn in self.addls: self.addls.remove(fn) return try: self.entries.remove(fn) self._dirty = True except KeyError: pass def __contains__(self, fn): if fn in self.addls: return True if self.entries is None: self._load() return fn in self.entries def __iter__(self): if self.entries is None: self._load() return iter(self.entries | self.addls) class _fncachevfs(vfsmod.proxyvfs): def __init__(self, vfs, fnc, encode): vfsmod.proxyvfs.__init__(self, vfs) self.fncache = fnc self.encode = encode def __call__(self, path, mode='r', *args, **kw): encoded = self.encode(path) if mode not in ('r', 'rb') and (path.startswith('data/') or path.startswith('meta/')): # do not trigger a fncache load when adding a file that already is # known to exist. notload = self.fncache.entries is None and self.vfs.exists(encoded) if notload and 'a' in mode and not self.vfs.stat(encoded).st_size: # when appending to an existing file, if the file has size zero, # it should be considered as missing. Such zero-size files are # the result of truncation when a transaction is aborted. notload = False if not notload: self.fncache.add(path) return self.vfs(encoded, mode, *args, **kw) def join(self, path): if path: return self.vfs.join(self.encode(path)) else: return self.vfs.join(path) class fncachestore(basicstore): def __init__(self, path, vfstype, dotencode): if dotencode: encode = _pathencode else: encode = _plainhybridencode self.encode = encode vfs = vfstype(path + '/store') self.path = vfs.base self.pathsep = self.path + '/' self.createmode = _calcmode(vfs) vfs.createmode = self.createmode self.rawvfs = vfs fnc = fncache(vfs) self.fncache = fnc self.vfs = _fncachevfs(vfs, fnc, encode) self.opener = self.vfs def join(self, f): return self.pathsep + self.encode(f) def getsize(self, path): return self.rawvfs.stat(path).st_size def datafiles(self, matcher=None): for f in sorted(self.fncache): if not _matchtrackedpath(f, matcher): continue ef = self.encode(f) try: yield f, ef, self.getsize(ef) except OSError as err: if err.errno != errno.ENOENT: raise def copylist(self): d = ('narrowspec data meta dh fncache phaseroots obsstore' ' 00manifest.d 00manifest.i 00changelog.d 00changelog.i') return (['requires', '00changelog.i'] + ['store/' + f for f in d.split()]) def write(self, tr): self.fncache.write(tr) def invalidatecaches(self): self.fncache.entries = None self.fncache.addls = set() def markremoved(self, fn): self.fncache.remove(fn) def _exists(self, f): ef = self.encode(f) try: self.getsize(ef) return True except OSError as err: if err.errno != errno.ENOENT: raise # nonexistent entry return False def __contains__(self, path): '''Checks if the store contains path''' path = "/".join(("data", path)) # check for files (exact match) e = path + '.i' if e in self.fncache and self._exists(e): return True # now check for directories (prefix match) if not path.endswith('/'): path += '/' for e in self.fncache: if e.startswith(path) and self._exists(e): return True return False