Mercurial > hg
view mercurial/mdiff.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 | e7aa113b14f7 |
children | 2372284d9457 |
line wrap: on
line source
# mdiff.py - diff and patch routines for mercurial # # Copyright 2005, 2006 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 re import struct import zlib from .i18n import _ from . import ( encoding, error, policy, pycompat, util, ) from .utils import dateutil _missing_newline_marker = "\\ No newline at end of file\n" bdiff = policy.importmod(r'bdiff') mpatch = policy.importmod(r'mpatch') blocks = bdiff.blocks fixws = bdiff.fixws patches = mpatch.patches patchedsize = mpatch.patchedsize textdiff = bdiff.bdiff splitnewlines = bdiff.splitnewlines class diffopts(object): '''context is the number of context lines text treats all files as text showfunc enables diff -p output git enables the git extended patch format nodates removes dates from diff headers nobinary ignores binary files noprefix disables the 'a/' and 'b/' prefixes (ignored in plain mode) ignorews ignores all whitespace changes in the diff ignorewsamount ignores changes in the amount of whitespace ignoreblanklines ignores changes whose lines are all blank upgrade generates git diffs to avoid data loss ''' defaults = { 'context': 3, 'text': False, 'showfunc': False, 'git': False, 'nodates': False, 'nobinary': False, 'noprefix': False, 'index': 0, 'ignorews': False, 'ignorewsamount': False, 'ignorewseol': False, 'ignoreblanklines': False, 'upgrade': False, 'showsimilarity': False, 'worddiff': False, 'xdiff': False, } def __init__(self, **opts): opts = pycompat.byteskwargs(opts) for k in self.defaults.keys(): v = opts.get(k) if v is None: v = self.defaults[k] setattr(self, k, v) try: self.context = int(self.context) except ValueError: raise error.Abort(_('diff context lines count must be ' 'an integer, not %r') % pycompat.bytestr(self.context)) def copy(self, **kwargs): opts = dict((k, getattr(self, k)) for k in self.defaults) opts = pycompat.strkwargs(opts) opts.update(kwargs) return diffopts(**opts) defaultopts = diffopts() def wsclean(opts, text, blank=True): if opts.ignorews: text = bdiff.fixws(text, 1) elif opts.ignorewsamount: text = bdiff.fixws(text, 0) if blank and opts.ignoreblanklines: text = re.sub('\n+', '\n', text).strip('\n') if opts.ignorewseol: text = re.sub(br'[ \t\r\f]+\n', br'\n', text) return text def splitblock(base1, lines1, base2, lines2, opts): # The input lines matches except for interwoven blank lines. We # transform it into a sequence of matching blocks and blank blocks. lines1 = [(wsclean(opts, l) and 1 or 0) for l in lines1] lines2 = [(wsclean(opts, l) and 1 or 0) for l in lines2] s1, e1 = 0, len(lines1) s2, e2 = 0, len(lines2) while s1 < e1 or s2 < e2: i1, i2, btype = s1, s2, '=' if (i1 >= e1 or lines1[i1] == 0 or i2 >= e2 or lines2[i2] == 0): # Consume the block of blank lines btype = '~' while i1 < e1 and lines1[i1] == 0: i1 += 1 while i2 < e2 and lines2[i2] == 0: i2 += 1 else: # Consume the matching lines while i1 < e1 and lines1[i1] == 1 and lines2[i2] == 1: i1 += 1 i2 += 1 yield [base1 + s1, base1 + i1, base2 + s2, base2 + i2], btype s1 = i1 s2 = i2 def hunkinrange(hunk, linerange): """Return True if `hunk` defined as (start, length) is in `linerange` defined as (lowerbound, upperbound). >>> hunkinrange((5, 10), (2, 7)) True >>> hunkinrange((5, 10), (6, 12)) True >>> hunkinrange((5, 10), (13, 17)) True >>> hunkinrange((5, 10), (3, 17)) True >>> hunkinrange((5, 10), (1, 3)) False >>> hunkinrange((5, 10), (18, 20)) False >>> hunkinrange((5, 10), (1, 5)) False >>> hunkinrange((5, 10), (15, 27)) False """ start, length = hunk lowerbound, upperbound = linerange return lowerbound < start + length and start < upperbound def blocksinrange(blocks, rangeb): """filter `blocks` like (a1, a2, b1, b2) from items outside line range `rangeb` from ``(b1, b2)`` point of view. Return `filteredblocks, rangea` where: * `filteredblocks` is list of ``block = (a1, a2, b1, b2), stype`` items of `blocks` that are inside `rangeb` from ``(b1, b2)`` point of view; a block ``(b1, b2)`` being inside `rangeb` if ``rangeb[0] < b2 and b1 < rangeb[1]``; * `rangea` is the line range w.r.t. to ``(a1, a2)`` parts of `blocks`. """ lbb, ubb = rangeb lba, uba = None, None filteredblocks = [] for block in blocks: (a1, a2, b1, b2), stype = block if lbb >= b1 and ubb <= b2 and stype == '=': # rangeb is within a single "=" hunk, restrict back linerange1 # by offsetting rangeb lba = lbb - b1 + a1 uba = ubb - b1 + a1 else: if b1 <= lbb < b2: if stype == '=': lba = a2 - (b2 - lbb) else: lba = a1 if b1 < ubb <= b2: if stype == '=': uba = a1 + (ubb - b1) else: uba = a2 if hunkinrange((b1, (b2 - b1)), rangeb): filteredblocks.append(block) if lba is None or uba is None or uba < lba: raise error.Abort(_('line range exceeds file size')) return filteredblocks, (lba, uba) def chooseblocksfunc(opts=None): if (opts is None or not opts.xdiff or not util.safehasattr(bdiff, 'xdiffblocks')): return bdiff.blocks else: return bdiff.xdiffblocks def allblocks(text1, text2, opts=None, lines1=None, lines2=None): """Return (block, type) tuples, where block is an mdiff.blocks line entry. type is '=' for blocks matching exactly one another (bdiff blocks), '!' for non-matching blocks and '~' for blocks matching only after having filtered blank lines. line1 and line2 are text1 and text2 split with splitnewlines() if they are already available. """ if opts is None: opts = defaultopts if opts.ignorews or opts.ignorewsamount or opts.ignorewseol: text1 = wsclean(opts, text1, False) text2 = wsclean(opts, text2, False) diff = chooseblocksfunc(opts)(text1, text2) for i, s1 in enumerate(diff): # The first match is special. # we've either found a match starting at line 0 or a match later # in the file. If it starts later, old and new below will both be # empty and we'll continue to the next match. if i > 0: s = diff[i - 1] else: s = [0, 0, 0, 0] s = [s[1], s1[0], s[3], s1[2]] # bdiff sometimes gives huge matches past eof, this check eats them, # and deals with the special first match case described above if s[0] != s[1] or s[2] != s[3]: type = '!' if opts.ignoreblanklines: if lines1 is None: lines1 = splitnewlines(text1) if lines2 is None: lines2 = splitnewlines(text2) old = wsclean(opts, "".join(lines1[s[0]:s[1]])) new = wsclean(opts, "".join(lines2[s[2]:s[3]])) if old == new: type = '~' yield s, type yield s1, '=' def unidiff(a, ad, b, bd, fn1, fn2, binary, opts=defaultopts): """Return a unified diff as a (headers, hunks) tuple. If the diff is not null, `headers` is a list with unified diff header lines "--- <original>" and "+++ <new>" and `hunks` is a generator yielding (hunkrange, hunklines) coming from _unidiff(). Otherwise, `headers` and `hunks` are empty. Set binary=True if either a or b should be taken as a binary file. """ def datetag(date, fn=None): if not opts.git and not opts.nodates: return '\t%s' % date if fn and ' ' in fn: return '\t' return '' sentinel = [], () if not a and not b: return sentinel if opts.noprefix: aprefix = bprefix = '' else: aprefix = 'a/' bprefix = 'b/' epoch = dateutil.datestr((0, 0)) fn1 = util.pconvert(fn1) fn2 = util.pconvert(fn2) if binary: if a and b and len(a) == len(b) and a == b: return sentinel headerlines = [] hunks = (None, ['Binary file %s has changed\n' % fn1]), elif not a: without_newline = not b.endswith('\n') b = splitnewlines(b) if a is None: l1 = '--- /dev/null%s' % datetag(epoch) else: l1 = "--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1)) l2 = "+++ %s%s" % (bprefix + fn2, datetag(bd, fn2)) headerlines = [l1, l2] size = len(b) hunkrange = (0, 0, 1, size) hunklines = ["@@ -0,0 +1,%d @@\n" % size] + ["+" + e for e in b] if without_newline: hunklines[-1] += '\n' hunklines.append(_missing_newline_marker) hunks = (hunkrange, hunklines), elif not b: without_newline = not a.endswith('\n') a = splitnewlines(a) l1 = "--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1)) if b is None: l2 = '+++ /dev/null%s' % datetag(epoch) else: l2 = "+++ %s%s%s" % (bprefix, fn2, datetag(bd, fn2)) headerlines = [l1, l2] size = len(a) hunkrange = (1, size, 0, 0) hunklines = ["@@ -1,%d +0,0 @@\n" % size] + ["-" + e for e in a] if without_newline: hunklines[-1] += '\n' hunklines.append(_missing_newline_marker) hunks = (hunkrange, hunklines), else: hunks = _unidiff(a, b, opts=opts) if not next(hunks): return sentinel headerlines = [ "--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1)), "+++ %s%s%s" % (bprefix, fn2, datetag(bd, fn2)), ] return headerlines, hunks def _unidiff(t1, t2, opts=defaultopts): """Yield hunks of a headerless unified diff from t1 and t2 texts. Each hunk consists of a (hunkrange, hunklines) tuple where `hunkrange` is a tuple (s1, l1, s2, l2) representing the range information of the hunk to form the '@@ -s1,l1 +s2,l2 @@' header and `hunklines` is a list of lines of the hunk combining said header followed by line additions and deletions. The hunks are prefixed with a bool. """ l1 = splitnewlines(t1) l2 = splitnewlines(t2) def contextend(l, len): ret = l + opts.context if ret > len: ret = len return ret def contextstart(l): ret = l - opts.context if ret < 0: return 0 return ret lastfunc = [0, ''] def yieldhunk(hunk): (astart, a2, bstart, b2, delta) = hunk aend = contextend(a2, len(l1)) alen = aend - astart blen = b2 - bstart + aend - a2 func = "" if opts.showfunc: lastpos, func = lastfunc # walk backwards from the start of the context up to the start of # the previous hunk context until we find a line starting with an # alphanumeric char. for i in pycompat.xrange(astart - 1, lastpos - 1, -1): if l1[i][0:1].isalnum(): func = b' ' + l1[i].rstrip() # split long function name if ASCII. otherwise we have no # idea where the multi-byte boundary is, so just leave it. if encoding.isasciistr(func): func = func[:41] lastfunc[1] = func break # by recording this hunk's starting point as the next place to # start looking for function lines, we avoid reading any line in # the file more than once. lastfunc[0] = astart # zero-length hunk ranges report their start line as one less if alen: astart += 1 if blen: bstart += 1 hunkrange = astart, alen, bstart, blen hunklines = ( ["@@ -%d,%d +%d,%d @@%s\n" % (hunkrange + (func,))] + delta + [' ' + l1[x] for x in pycompat.xrange(a2, aend)] ) # If either file ends without a newline and the last line of # that file is part of a hunk, a marker is printed. If the # last line of both files is identical and neither ends in # a newline, print only one marker. That's the only case in # which the hunk can end in a shared line without a newline. skip = False if not t1.endswith('\n') and astart + alen == len(l1) + 1: for i in pycompat.xrange(len(hunklines) - 1, -1, -1): if hunklines[i].startswith(('-', ' ')): if hunklines[i].startswith(' '): skip = True hunklines[i] += '\n' hunklines.insert(i + 1, _missing_newline_marker) break if not skip and not t2.endswith('\n') and bstart + blen == len(l2) + 1: for i in pycompat.xrange(len(hunklines) - 1, -1, -1): if hunklines[i].startswith('+'): hunklines[i] += '\n' hunklines.insert(i + 1, _missing_newline_marker) break yield hunkrange, hunklines # bdiff.blocks gives us the matching sequences in the files. The loop # below finds the spaces between those matching sequences and translates # them into diff output. # hunk = None ignoredlines = 0 has_hunks = False for s, stype in allblocks(t1, t2, opts, l1, l2): a1, a2, b1, b2 = s if stype != '!': if stype == '~': # The diff context lines are based on t1 content. When # blank lines are ignored, the new lines offsets must # be adjusted as if equivalent blocks ('~') had the # same sizes on both sides. ignoredlines += (b2 - b1) - (a2 - a1) continue delta = [] old = l1[a1:a2] new = l2[b1:b2] b1 -= ignoredlines b2 -= ignoredlines astart = contextstart(a1) bstart = contextstart(b1) prev = None if hunk: # join with the previous hunk if it falls inside the context if astart < hunk[1] + opts.context + 1: prev = hunk astart = hunk[1] bstart = hunk[3] else: if not has_hunks: has_hunks = True yield True for x in yieldhunk(hunk): yield x if prev: # we've joined the previous hunk, record the new ending points. hunk[1] = a2 hunk[3] = b2 delta = hunk[4] else: # create a new hunk hunk = [astart, a2, bstart, b2, delta] delta[len(delta):] = [' ' + x for x in l1[astart:a1]] delta[len(delta):] = ['-' + x for x in old] delta[len(delta):] = ['+' + x for x in new] if hunk: if not has_hunks: has_hunks = True yield True for x in yieldhunk(hunk): yield x elif not has_hunks: yield False def b85diff(to, tn): '''print base85-encoded binary diff''' def fmtline(line): l = len(line) if l <= 26: l = pycompat.bytechr(ord('A') + l - 1) else: l = pycompat.bytechr(l - 26 + ord('a') - 1) return '%c%s\n' % (l, util.b85encode(line, True)) def chunk(text, csize=52): l = len(text) i = 0 while i < l: yield text[i:i + csize] i += csize if to is None: to = '' if tn is None: tn = '' if to == tn: return '' # TODO: deltas ret = [] ret.append('GIT binary patch\n') ret.append('literal %d\n' % len(tn)) for l in chunk(zlib.compress(tn)): ret.append(fmtline(l)) ret.append('\n') return ''.join(ret) def patchtext(bin): pos = 0 t = [] while pos < len(bin): p1, p2, l = struct.unpack(">lll", bin[pos:pos + 12]) pos += 12 t.append(bin[pos:pos + l]) pos += l return "".join(t) def patch(a, bin): if len(a) == 0: # skip over trivial delta header return util.buffer(bin, 12) return mpatch.patches(a, [bin]) # similar to difflib.SequenceMatcher.get_matching_blocks def get_matching_blocks(a, b): return [(d[0], d[2], d[1] - d[0]) for d in bdiff.blocks(a, b)] def trivialdiffheader(length): return struct.pack(">lll", 0, 0, length) if length else '' def replacediffheader(oldlen, newlen): return struct.pack(">lll", 0, oldlen, newlen)