view mercurial/store.py @ 46150:a132aa5979ec

copies: no longer cache the ChangedFiles during copy tracing Now that the copies information for both parents are processed all at once, we no longer needs to cache this information, so we simplify the code. The simpler code is also a (tiny) bit faster overall. Repo Case Source-Rev Dest-Rev # of revisions old time new time Difference Factor time per rev --------------------------------------------------------------------------------------------------------------------------------------------------------------- mercurial x_revs_x_added_0_copies ad6b123de1c7 39cfcef4f463 : 1 revs, 0.000041 s, 0.000041 s, +0.000000 s, × 1.0000, 41 µs/rev mercurial x_revs_x_added_x_copies 2b1c78674230 0c1d10351869 : 6 revs, 0.000102 s, 0.000096 s, -0.000006 s, × 0.9412, 16 µs/rev mercurial x000_revs_x000_added_x_copies 81f8ff2a9bf2 dd3267698d84 : 1032 revs, 0.004254 s, 0.004039 s, -0.000215 s, × 0.9495, 3 µs/rev pypy x_revs_x_added_0_copies aed021ee8ae8 099ed31b181b : 9 revs, 0.000282 s, 0.000189 s, -0.000093 s, × 0.6702, 21 µs/rev pypy x_revs_x000_added_0_copies 4aa4e1f8e19a 359343b9ac0e : 1 revs, 0.000048 s, 0.000047 s, -0.000001 s, × 0.9792, 47 µs/rev pypy x_revs_x_added_x_copies ac52eb7bbbb0 72e022663155 : 7 revs, 0.000211 s, 0.000103 s, -0.000108 s, × 0.4882, 14 µs/rev pypy x_revs_x00_added_x_copies c3b14617fbd7 ace7255d9a26 : 1 revs, 0.000375 s, 0.000286 s, -0.000089 s, × 0.7627, 286 µs/rev pypy x_revs_x000_added_x000_copies df6f7a526b60 a83dc6a2d56f : 6 revs, 0.010574 s, 0.010436 s, -0.000138 s, × 0.9869, 1739 µs/rev pypy x000_revs_xx00_added_0_copies 89a76aede314 2f22446ff07e : 4785 revs, 0.049974 s, 0.047465 s, -0.002509 s, × 0.9498, 9 µs/rev pypy x000_revs_x000_added_x_copies 8a3b5bfd266e 2c68e87c3efe : 6780 revs, 0.084300 s, 0.082351 s, -0.001949 s, × 0.9769, 12 µs/rev pypy x000_revs_x000_added_x000_copies 89a76aede314 7b3dda341c84 : 5441 revs, 0.060128 s, 0.058757 s, -0.001371 s, × 0.9772, 10 µs/rev pypy x0000_revs_x_added_0_copies d1defd0dc478 c9cb1334cc78 : 43645 revs, 0.686542 s, 0.674129 s, -0.012413 s, × 0.9819, 15 µs/rev pypy x0000_revs_xx000_added_0_copies bf2c629d0071 4ffed77c095c : 2 revs, 0.009277 s, 0.009434 s, +0.000157 s, × 1.0169, 4717 µs/rev pypy x0000_revs_xx000_added_x000_copies 08ea3258278e d9fa043f30c0 : 11316 revs, 0.114733 s, 0.111935 s, -0.002798 s, × 0.9756, 9 µs/rev netbeans x_revs_x_added_0_copies fb0955ffcbcd a01e9239f9e7 : 2 revs, 0.000081 s, 0.000078 s, -0.000003 s, × 0.9630, 39 µs/rev netbeans x_revs_x000_added_0_copies 6f360122949f 20eb231cc7d0 : 2 revs, 0.000107 s, 0.000106 s, -0.000001 s, × 0.9907, 53 µs/rev netbeans x_revs_x_added_x_copies 1ada3faf6fb6 5a39d12eecf4 : 3 revs, 0.000173 s, 0.000162 s, -0.000011 s, × 0.9364, 54 µs/rev netbeans x_revs_x00_added_x_copies 35be93ba1e2c 9eec5e90c05f : 9 revs, 0.000698 s, 0.000695 s, -0.000003 s, × 0.9957, 77 µs/rev netbeans x000_revs_xx00_added_0_copies eac3045b4fdd 51d4ae7f1290 : 1421 revs, 0.009248 s, 0.008901 s, -0.000347 s, × 0.9625, 6 µs/rev netbeans x000_revs_x000_added_x_copies e2063d266acd 6081d72689dc : 1533 revs, 0.015446 s, 0.014333 s, -0.001113 s, × 0.9279, 9 µs/rev netbeans x000_revs_x000_added_x000_copies ff453e9fee32 411350406ec2 : 5750 revs, 0.074373 s, 0.071998 s, -0.002375 s, × 0.9681, 12 µs/rev netbeans x0000_revs_xx000_added_x000_copies 588c2d1ced70 1aad62e59ddd : 66949 revs, 0.639870 s, 0.615346 s, -0.024524 s, × 0.9617, 9 µs/rev mozilla-central x_revs_x_added_0_copies 3697f962bb7b 7015fcdd43a2 : 2 revs, 0.000088 s, 0.000085 s, -0.000003 s, × 0.9659, 42 µs/rev mozilla-central x_revs_x000_added_0_copies dd390860c6c9 40d0c5bed75d : 8 revs, 0.000199 s, 0.000199 s, +0.000000 s, × 1.0000, 24 µs/rev mozilla-central x_revs_x_added_x_copies 8d198483ae3b 14207ffc2b2f : 9 revs, 0.000171 s, 0.000169 s, -0.000002 s, × 0.9883, 18 µs/rev mozilla-central x_revs_x00_added_x_copies 98cbc58cc6bc 446a150332c3 : 7 revs, 0.000592 s, 0.000590 s, -0.000002 s, × 0.9966, 84 µs/rev mozilla-central x_revs_x000_added_x000_copies 3c684b4b8f68 0a5e72d1b479 : 3 revs, 0.003151 s, 0.003122 s, -0.000029 s, × 0.9908, 1040 µs/rev mozilla-central x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 6 revs, 0.061612 s, 0.061192 s, -0.000420 s, × 0.9932, 10198 µs/rev mozilla-central x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 1593 revs, 0.005381 s, 0.005137 s, -0.000244 s, × 0.9547, 3 µs/rev mozilla-central x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 41 revs, 0.003742 s, 0.003585 s, -0.000157 s, × 0.9580, 87 µs/rev mozilla-central x000_revs_x000_added_x000_copies 7c97034feb78 4407bd0c6330 : 7839 revs, 0.061983 s, 0.060592 s, -0.001391 s, × 0.9776, 7 µs/rev mozilla-central x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 615 revs, 0.019861 s, 0.019596 s, -0.000265 s, × 0.9867, 31 µs/rev mozilla-central x0000_revs_xx000_added_x000_copies f78c615a656c 96a38b690156 : 30263 revs, 0.188101 s, 0.183558 s, -0.004543 s, × 0.9758, 6 µs/rev mozilla-central x00000_revs_x0000_added_x0000_copies 6832ae71433c 4c222a1d9a00 : 153721 revs, 1.806696 s, 1.758083 s, -0.048613 s, × 0.9731, 11 µs/rev mozilla-central x00000_revs_x00000_added_x000_copies 76caed42cf7c 1daa622bbe42 : 204976 revs, 2.682987 s, 2.592955 s, -0.090032 s, × 0.9664, 12 µs/rev mozilla-try x_revs_x_added_0_copies aaf6dde0deb8 9790f499805a : 2 revs, 0.000852 s, 0.000844 s, -0.000008 s, × 0.9906, 422 µs/rev mozilla-try x_revs_x000_added_0_copies d8d0222927b4 5bb8ce8c7450 : 2 revs, 0.000859 s, 0.000861 s, +0.000002 s, × 1.0023, 430 µs/rev mozilla-try x_revs_x_added_x_copies 092fcca11bdb 936255a0384a : 4 revs, 0.000150 s, 0.000150 s, +0.000000 s, × 1.0000, 37 µs/rev mozilla-try x_revs_x00_added_x_copies b53d2fadbdb5 017afae788ec : 2 revs, 0.001158 s, 0.001166 s, +0.000008 s, × 1.0069, 583 µs/rev mozilla-try x_revs_x000_added_x000_copies 20408ad61ce5 6f0ee96e21ad : 1 revs, 0.027240 s, 0.027359 s, +0.000119 s, × 1.0044, 27359 µs/rev mozilla-try x_revs_x0000_added_x0000_copies effb563bb7e5 c07a39dc4e80 : 6 revs, 0.062824 s, 0.061848 s, -0.000976 s, × 0.9845, 10308 µs/rev mozilla-try x000_revs_xx00_added_0_copies 6100d773079a 04a55431795e : 1593 revs, 0.005463 s, 0.005110 s, -0.000353 s, × 0.9354, 3 µs/rev mozilla-try x000_revs_x000_added_x_copies 9f17a6fc04f9 2d37b966abed : 41 revs, 0.004238 s, 0.004168 s, -0.000070 s, × 0.9835, 101 µs/rev mozilla-try x000_revs_x000_added_x000_copies 1346fd0130e4 4c65cbdabc1f : 6657 revs, 0.064113 s, 0.063414 s, -0.000699 s, × 0.9891, 9 µs/rev mozilla-try x0000_revs_x_added_0_copies 63519bfd42ee a36a2a865d92 : 40314 revs, 0.294063 s, 0.288301 s, -0.005762 s, × 0.9804, 7 µs/rev mozilla-try x0000_revs_x_added_x_copies 9fe69ff0762d bcabf2a78927 : 38690 revs, 0.281493 s, 0.275798 s, -0.005695 s, × 0.9798, 7 µs/rev mozilla-try x0000_revs_xx000_added_x_copies 156f6e2674f2 4d0f2c178e66 : 8598 revs, 0.076323 s, 0.074640 s, -0.001683 s, × 0.9779, 8 µs/rev mozilla-try x0000_revs_xx000_added_0_copies 9eec5917337d 67118cc6dcad : 615 revs, 0.020390 s, 0.020327 s, -0.000063 s, × 0.9969, 33 µs/rev mozilla-try x0000_revs_xx000_added_x000_copies 89294cd501d9 7ccb2fc7ccb5 : 97052 revs, 3.023879 s, 2.970385 s, -0.053494 s, × 0.9823, 30 µs/rev mozilla-try x0000_revs_x0000_added_x0000_copies e928c65095ed e951f4ad123a : 52031 revs, 0.735549 s, 0.719432 s, -0.016117 s, × 0.9781, 13 µs/rev mozilla-try x00000_revs_x_added_0_copies 6a320851d377 1ebb79acd503 : 363753 revs, 18.568900 s, 18.165143 s, -0.403757 s, × 0.9783, 49 µs/rev mozilla-try x00000_revs_x00000_added_0_copies dc8a3ca7010e d16fde900c9c : 34414 revs, 0.502584 s, 0.486769 s, -0.015815 s, × 0.9685, 14 µs/rev mozilla-try x00000_revs_x_added_x_copies 5173c4b6f97c 95d83ee7242d : 362229 revs, 18.356645 s, 17.913924 s, -0.442721 s, × 0.9759, 49 µs/rev mozilla-try x00000_revs_x000_added_x_copies 9126823d0e9c ca82787bb23c : 359344 revs, 18.250393 s, 17.660113 s, -0.590280 s, × 0.9677, 49 µs/rev mozilla-try x00000_revs_x0000_added_x0000_copies 8d3fafa80d4b eb884023b810 : 192665 revs, 2.792459 s, 2.709446 s, -0.083013 s, × 0.9703, 14 µs/rev mozilla-try x00000_revs_x00000_added_x0000_copies 1b661134e2ca 1ae03d022d6d : 228985 revs, 107.697264 s, 107.796891 s, +0.099627 s, × 1.0009, 470 µs/rev mozilla-try x00000_revs_x00000_added_x000_copies 9b2a99adc05e 8e29777b48e6 : 382065 revs, 63.961040 s, 63.575217 s, -0.385823 s, × 0.9940, 166 µs/rev Differential Revision: https://phab.mercurial-scm.org/D9423
author Pierre-Yves David <pierre-yves.david@octobus.net>
date Fri, 20 Nov 2020 13:46:14 +0100
parents 59fa3890d40a
children 1cebad969d88
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 os
import stat

from .i18n import _
from .pycompat import getattr
from .node import hex
from . import (
    changelog,
    error,
    manifest,
    policy,
    pycompat,
    util,
    vfs as vfsmod,
)
from .utils import hashutil

parsers = policy.importmod('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(b'data/'):
        return matcher(path[len(b'data/') : -len(b'.i')])
    elif path.startswith(b'meta/'):
        return matcher.visitdir(path[len(b'meta/') : -len(b'/00manifest.i')])

    raise error.ProgrammingError(b"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(b".hg/", b".hg.hg/")
        .replace(b".i/", b".i.hg/")
        .replace(b".d/", b".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 b".hg/" not in path:
        return path
    return (
        path.replace(b".d.hg/", b".d/")
        .replace(b".i.hg/", b".i/")
        .replace(b".hg.hg/", b".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 = b'_'
    xchr = pycompat.bytechr
    asciistr = list(map(xchr, range(127)))
    capitals = list(range(ord(b"A"), ord(b"Z") + 1))

    cmap = {x: x for x in asciistr}
    for x in _reserved():
        cmap[xchr(x)] = b"~%02x" % x
    for x in capitals + [ord(e)]:
        cmap[xchr(x)] = e + xchr(x).lower()

    dmap = {}
    for k, v in pycompat.iteritems(cmap):
        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: b''.join(
            [cmap[s[c : c + 1]] for c in pycompat.xrange(len(s))]
        ),
        lambda s: b''.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 = {xchr(x): xchr(x) for x in pycompat.xrange(127)}
    for x in _reserved():
        cmap[xchr(x)] = b"~%02x" % x
    for x in range(ord(b"A"), ord(b"Z") + 1):
        cmap[xchr(x)] = xchr(x).lower()

    def lowerencode(s):
        return b"".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 = (b'aux', b'con', b'prn', b'nul')  # length 3
_winres4 = (b'com', b'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 b'. ':
            n = b"~%02x" % ord(n[0:1]) + n[1:]
            path[i] = n
        else:
            l = n.find(b'.')
            if l == -1:
                l = len(n)
            if (l == 3 and n[:3] in _winres3) or (
                l == 4
                and n[3:4] <= b'9'
                and n[3:4] >= b'1'
                and n[:3] in _winres4
            ):
                # encode third letter ('aux' -> 'au~78')
                ec = b"~%02x" % ord(n[2:3])
                n = n[0:2] + ec + n[3:]
                path[i] = n
        if n[-1] in b'. ':
            # encode last period or space ('foo...' -> 'foo..~2e')
            path[i] = n[:-1] + b"~%02x" % ord(n[-1:])
    return path


_maxstorepathlen = 120
_dirprefixlen = 8
_maxshortdirslen = 8 * (_dirprefixlen + 1) - 4


def _hashencode(path, dotencode):
    digest = hex(hashutil.sha1(path).digest())
    le = lowerencode(path[5:]).split(b'/')  # 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 b'. ':
            # Windows can't access dirs ending in period or space
            d = d[:-1] + b'_'
        if sdirslen == 0:
            t = len(d)
        else:
            t = sdirslen + 1 + len(d)
            if t > _maxshortdirslen:
                break
        sdirs.append(d)
        sdirslen = t
    dirs = b'/'.join(sdirs)
    if len(dirs) > 0:
        dirs += b'/'
    res = b'dh/' + dirs + digest + ext
    spaceleft = _maxstorepathlen - len(res)
    if spaceleft > 0:
        filler = basename[:spaceleft]
        res = b'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(b'/')
    res = b'/'.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(b'/')
    res = b'/'.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 = [
    b'bookmarks',
    b'narrowspec',
    b'data',
    b'meta',
    b'00manifest.d',
    b'00manifest.i',
    b'00changelog.d',
    b'00changelog.i',
    b'phaseroots',
    b'obsstore',
    b'requires',
]


def isrevlog(f, kind, st):
    return kind == stat.S_IFREG and f[-2:] in (b'.i', b'.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 + b'/' + encodedir(f)

    def _walk(self, relpath, recurse, filefilter=isrevlog):
        '''yields (unencoded, encoded, size)'''
        path = self.path
        if relpath:
            path += b'/' + 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 + b'/' + 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 changelog(self, trypending):
        return changelog.changelog(self.vfs, trypending=trypending)

    def manifestlog(self, repo, storenarrowmatch):
        rootstore = manifest.manifestrevlog(self.vfs)
        return manifest.manifestlog(self.vfs, repo, rootstore, storenarrowmatch)

    def datafiles(self, matcher=None):
        return self._walk(b'data', True) + self._walk(b'meta', True)

    def topfiles(self):
        # yield manifest before changelog
        return reversed(self._walk(b'', 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 _data

    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 = b"/".join((b"data", path))
        # file?
        if self.vfs.exists(path + b".i"):
            return True
        # dir?
        if not path.endswith(b"/"):
            path = path + b"/"
        return self.vfs.exists(path)


class encodedstore(basicstore):
    def __init__(self, path, vfstype):
        vfs = vfstype(path + b'/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 + b'/' + encodefilename(f)

    def copylist(self):
        return [b'requires', b'00changelog.i'] + [b'store/' + f for f in _data]


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 ensureloaded(self, warn=None):
        """read the fncache file if not already read.

        If the file on disk is corrupted, raise. If warn is provided,
        warn and keep going instead."""
        if self.entries is None:
            self._load(warn)

    def _load(self, warn=None):
        '''fill the entries from the fncache file'''
        self._dirty = False
        try:
            fp = self.vfs(b'fncache', mode=b'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:
            msg = _(b"fncache does not ends with a newline")
            if warn:
                warn(msg + b'\n')
            else:
                raise error.Abort(
                    msg,
                    hint=_(
                        b"use 'hg debugrebuildfncache' to "
                        b"rebuild the fncache"
                    ),
                )
        self._checkentries(fp, warn)
        fp.close()

    def _checkentries(self, fp, warn):
        """ make sure there is no empty string in entries """
        if b'' in self.entries:
            fp.seek(0)
            for n, line in enumerate(util.iterfile(fp)):
                if not line.rstrip(b'\n'):
                    t = _(b'invalid entry in fncache, line %d') % (n + 1)
                    if warn:
                        warn(t + b'\n')
                    else:
                        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(b'fncache')
            fp = self.vfs(b'fncache', mode=b'wb', atomictemp=True)
            if self.entries:
                fp.write(encodedir(b'\n'.join(self.entries) + b'\n'))
            fp.close()
            self._dirty = False
        if self.addls:
            # if we have just new entries, let's append them to the fncache
            tr.addbackup(b'fncache')
            fp = self.vfs(b'fncache', mode=b'ab', atomictemp=True)
            if self.addls:
                fp.write(encodedir(b'\n'.join(self.addls) + b'\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=b'r', *args, **kw):
        encoded = self.encode(path)
        if mode not in (b'r', b'rb') and (
            path.startswith(b'data/') or path.startswith(b'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 b'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 + b'/store')
        self.path = vfs.base
        self.pathsep = self.path + b'/'
        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 = (
            b'bookmarks',
            b'narrowspec',
            b'data',
            b'meta',
            b'dh',
            b'fncache',
            b'phaseroots',
            b'obsstore',
            b'00manifest.d',
            b'00manifest.i',
            b'00changelog.d',
            b'00changelog.i',
            b'requires',
        )
        return [b'requires', b'00changelog.i'] + [b'store/' + f for f in d]

    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 = b"/".join((b"data", path))
        # check for files (exact match)
        e = path + b'.i'
        if e in self.fncache and self._exists(e):
            return True
        # now check for directories (prefix match)
        if not path.endswith(b'/'):
            path += b'/'
        for e in self.fncache:
            if e.startswith(path) and self._exists(e):
                return True
        return False