tests/test-pathencode.py

sparse-revlog: implement algorithm to write sparse delta chains (issue5480)

The classic behavior of revlog._isgooddeltainfo is to consider the span size
of the whole delta chain, and limit it to 4 * textlen.
Once sparse-revlog writing is allowed (and enforced with a requirement),
revlog._isgooddeltainfo considers the span of the largest chunk as the
distance used in the verification, instead of using the span of the whole
delta chain.

In order to compute the span of the largest chunk, we need to slice into
chunks a chain with the new revision at the top of the revlog, and take the
maximal span of these chunks. The sparse read density is a parameter to the
slicing, as it will stop when the global read density reaches this threshold.
For instance, a density of 50% means that 2 of 4 read bytes are actually used
for the reconstruction of the revision (the others are part of other chains).

This allows a new revision to be potentially stored with a diff against
another revision anywhere in the history, instead of forcing it in the last 4
* textlen. The result is a much better compression on repositories that have
many concurrent branches. Here are a comparison between using deltas from
current upstream (aggressive-merge-deltas on by default) and deltas from a
sparse-revlog

Comparison of `.hg/store/` size:

    mercurial (6.74% merges):
        before:     46,831,873 bytes
        after:      46,795,992 bytes (no relevant change)
    pypy (8.30% merges):
        before:    333,524,651 bytes
        after:     308,417,511 bytes -8%
    netbeans (34.21% merges):
        before:  1,141,847,554 bytes
        after:   1,131,093,161 bytes -1%
    mozilla-central (4.84% merges):
        before:  2,344,248,850 bytes
        after:   2,328,459,258 bytes -1%
    large-private-repo-A (merge 19.73%)
        before: 41,510,550,163 bytes
        after:   8,121,763,428 bytes -80%
    large-private-repo-B (23.77%)
        before: 58,702,221,709 bytes
        after:   8,351,588,828 bytes -76%

Comparison of `00manifest.d` size:

    mercurial (6.74% merges):
        before:      6,143,044 bytes
        after:       6,107,163 bytes
    pypy (8.30% merges):
        before:     52,941,780 bytes
        after:      27,834,082 bytes -48%
    netbeans (34.21% merges):
        before:    130,088,982 bytes
        after:     119,337,636 bytes -10%
    mozilla-central (4.84% merges):
        before:    215,096,339 bytes
        after:     199,496,863 bytes -8%
    large-private-repo-A (merge 19.73%)
        before: 33,725,285,081 bytes
        after:     390,302,545 bytes -99%
    large-private-repo-B (23.77%)
        before: 49,457,701,645 bytes
        after:   1,366,752,187 bytes -97%


The better delta chains provide a performance boost in relevant repositories:

    pypy, bundling 1000 revisions:
        before: 1.670s
        after:  1.149s -31%

Unbundling got a bit slower. probably because the sparse algorithm is still
pure
python.

    pypy, unbundling 1000 revisions:
        before: 4.062s
        after:  4.507s +10%

Performance of bundle/unbundle in repository with few concurrent branches (eg:
mercurial) are unaffected.

No significant differences have been noticed then timing `hg push` and `hg
pull` locally. More state timings are being gathered.

Same as for aggressive-merge-delta, better delta comes with longer delta
chains. Longer chains have a performance impact. For example. The length of
the chain needed to get the manifest of pypy's tip moves from 82 item to 1929
items. This moves the restore time from 3.88ms to 11.3ms.

Delta chain length is an independent issue that affects repository without
this changes. It will be dealt with independently.

No significant differences have been observed on repositories where
`sparse-revlog` have not much effect (mercurial, unity, netbeans). On pypy,
small differences have been observed on some operation affected by delta chain
building and retrieval.


    pypy, perfmanifest
        before: 0.006162s
        after:  0.017899s +190%

    pypy, commit:
        before: 0.382
        after:  0.376 -1%

    pypy, status:
        before: 0.157
        after:  0.168 +7%

More comprehensive and stable timing comparisons are in progress.

# This is a randomized test that generates different pathnames every
# time it is invoked, and tests the encoding of those pathnames.
#
# It uses a simple probabilistic model to generate valid pathnames
# that have proven likely to expose bugs and divergent behavior in
# different encoding implementations.

from __future__ import absolute_import, print_function

import binascii
import collections
import itertools
import math
import os
import random
import sys
import time
from mercurial import (
    pycompat,
    store,
)

try:
    xrange
except NameError:
    xrange = range

validchars = set(map(pycompat.bytechr, range(0, 256)))
alphanum = range(ord('A'), ord('Z'))

for c in (b'\0', b'/'):
    validchars.remove(c)

winreserved = (b'aux con prn nul'.split() +
               [b'com%d' % i for i in xrange(1, 10)] +
               [b'lpt%d' % i for i in xrange(1, 10)])

def casecombinations(names):
    '''Build all case-diddled combinations of names.'''

    combos = set()

    for r in names:
        for i in xrange(len(r) + 1):
            for c in itertools.combinations(xrange(len(r)), i):
                d = r
                for j in c:
                    d = b''.join((d[:j], d[j:j + 1].upper(), d[j + 1:]))
                combos.add(d)
    return sorted(combos)

def buildprobtable(fp, cmd='hg manifest tip'):
    '''Construct and print a table of probabilities for path name
    components.  The numbers are percentages.'''

    counts = collections.defaultdict(lambda: 0)
    for line in os.popen(cmd).read().splitlines():
        if line[-2:] in ('.i', '.d'):
            line = line[:-2]
        if line.startswith('data/'):
            line = line[5:]
        for c in line:
            counts[c] += 1
    for c in '\r/\n':
        counts.pop(c, None)
    t = sum(counts.itervalues()) / 100.0
    fp.write('probtable = (')
    for i, (k, v) in enumerate(sorted(counts.items(), key=lambda x: x[1],
                                      reverse=True)):
        if (i % 5) == 0:
            fp.write('\n    ')
        vt = v / t
        if vt < 0.0005:
            break
        fp.write('(%r, %.03f), ' % (k, vt))
    fp.write('\n    )\n')

# A table of character frequencies (as percentages), gleaned by
# looking at filelog names from a real-world, very large repo.

probtable = (
    (b't', 9.828), (b'e', 9.042), (b's', 8.011), (b'a', 6.801), (b'i', 6.618),
    (b'g', 5.053), (b'r', 5.030), (b'o', 4.887), (b'p', 4.363), (b'n', 4.258),
    (b'l', 3.830), (b'h', 3.693), (b'_', 3.659), (b'.', 3.377), (b'm', 3.194),
    (b'u', 2.364), (b'd', 2.296), (b'c', 2.163), (b'b', 1.739), (b'f', 1.625),
    (b'6', 0.666), (b'j', 0.610), (b'y', 0.554), (b'x', 0.487), (b'w', 0.477),
    (b'k', 0.476), (b'v', 0.473), (b'3', 0.336), (b'1', 0.335), (b'2', 0.326),
    (b'4', 0.310), (b'5', 0.305), (b'9', 0.302), (b'8', 0.300), (b'7', 0.299),
    (b'q', 0.298), (b'0', 0.250), (b'z', 0.223), (b'-', 0.118), (b'C', 0.095),
    (b'T', 0.087), (b'F', 0.085), (b'B', 0.077), (b'S', 0.076), (b'P', 0.076),
    (b'L', 0.059), (b'A', 0.058), (b'N', 0.051), (b'D', 0.049), (b'M', 0.046),
    (b'E', 0.039), (b'I', 0.035), (b'R', 0.035), (b'G', 0.028), (b'U', 0.026),
    (b'W', 0.025), (b'O', 0.017), (b'V', 0.015), (b'H', 0.013), (b'Q', 0.011),
    (b'J', 0.007), (b'K', 0.005), (b'+', 0.004), (b'X', 0.003), (b'Y', 0.001),
    )

for c, _ in probtable:
    validchars.remove(c)
validchars = list(validchars)

def pickfrom(rng, table):
    c = 0
    r = rng.random() * sum(i[1] for i in table)
    for i, p in table:
        c += p
        if c >= r:
            return i

reservedcombos = casecombinations(winreserved)

# The first component of a name following a slash.

firsttable = (
    (lambda rng: pickfrom(rng, probtable), 90),
    (lambda rng: rng.choice(validchars), 5),
    (lambda rng: rng.choice(reservedcombos), 5),
    )

# Components of a name following the first.

resttable = firsttable[:-1]

# Special suffixes.

internalsuffixcombos = casecombinations(b'.hg .i .d'.split())

# The last component of a path, before a slash or at the end of a name.

lasttable = resttable + (
    (lambda rng: b'', 95),
    (lambda rng: rng.choice(internalsuffixcombos), 5),
    )

def makepart(rng, k):
    '''Construct a part of a pathname, without slashes.'''

    p = pickfrom(rng, firsttable)(rng)
    l = len(p)
    ps = [p]
    maxl = rng.randint(1, k)
    while l < maxl:
        p = pickfrom(rng, resttable)(rng)
        l += len(p)
        ps.append(p)
    ps.append(pickfrom(rng, lasttable)(rng))
    return b''.join(ps)

def makepath(rng, j, k):
    '''Construct a complete pathname.'''

    return (b'data/' + b'/'.join(makepart(rng, k) for _ in xrange(j)) +
            rng.choice([b'.d', b'.i']))

def genpath(rng, count):
    '''Generate random pathnames with gradually increasing lengths.'''

    mink, maxk = 1, 4096
    def steps():
        for i in xrange(count):
            yield mink + int(round(math.sqrt((maxk - mink) * float(i) / count)))
    for k in steps():
        x = rng.randint(1, k)
        y = rng.randint(1, k)
        yield makepath(rng, x, y)

def runtests(rng, seed, count):
    nerrs = 0
    for p in genpath(rng, count):
        h = store._pathencode(p)    # uses C implementation, if available
        r = store._hybridencode(p, True) # reference implementation in Python
        if h != r:
            if nerrs == 0:
                print('seed:', hex(seed)[:-1], file=sys.stderr)
            print("\np: '%s'" % p.encode("string_escape"), file=sys.stderr)
            print("h: '%s'" % h.encode("string_escape"), file=sys.stderr)
            print("r: '%s'" % r.encode("string_escape"), file=sys.stderr)
            nerrs += 1
    return nerrs

def main():
    import getopt

    # Empirically observed to take about a second to run
    count = 100
    seed = None
    opts, args = getopt.getopt(sys.argv[1:], 'c:s:',
                               ['build', 'count=', 'seed='])
    for o, a in opts:
        if o in ('-c', '--count'):
            count = int(a)
        elif o in ('-s', '--seed'):
            seed = int(a, base=0) # accepts base 10 or 16 strings
        elif o == '--build':
            buildprobtable(sys.stdout,
                           'find .hg/store/data -type f && '
                           'cat .hg/store/fncache 2>/dev/null')
            sys.exit(0)

    if seed is None:
        try:
            seed = int(binascii.hexlify(os.urandom(16)), 16)
        except AttributeError:
            seed = int(time.time() * 1000)

    rng = random.Random(seed)
    if runtests(rng, seed, count):
        sys.exit(1)

if __name__ == '__main__':
    main()