annotate mercurial/mdiff.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

# mdiff.py - diff and patch routines for mercurial
#
# Copyright 2005, 2006 Matt Mackall <[e-mail]>
#
# 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 .pycompat import (
    getattr,
    setattr,
)
from . import (
    diffhelper,
    encoding,
    error,
    policy,
    pycompat,
    util,
)
from .utils import dateutil

bdiff = policy.importmod('bdiff')
mpatch = policy.importmod('mpatch')

blocks = bdiff.blocks
fixws = bdiff.fixws
patches = mpatch.patches
patchedsize = mpatch.patchedsize
textdiff = bdiff.bdiff
splitnewlines = bdiff.splitnewlines


# TODO: this looks like it could be an attrs, which might help pytype
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
    """

    _HAS_DYNAMIC_ATTRIBUTES = True

    defaults = {
        b'context': 3,
        b'text': False,
        b'showfunc': False,
        b'git': False,
        b'nodates': False,
        b'nobinary': False,
        b'noprefix': False,
        b'index': 0,
        b'ignorews': False,
        b'ignorewsamount': False,
        b'ignorewseol': False,
        b'ignoreblanklines': False,
        b'upgrade': False,
        b'showsimilarity': False,
        b'worddiff': False,
        b'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(
                _(b'diff context lines count must be an integer, not %r')
                % pycompat.bytestr(self.context)
            )

    def copy(self, **kwargs):
        opts = {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(b'\n+', b'\n', text).strip(b'\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, b'='
        if i1 >= e1 or lines1[i1] == 0 or i2 >= e2 or lines2[i2] == 0:
            # Consume the block of blank lines
            btype = b'~'
            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 == b'=':
            # 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 == b'=':
                    lba = a2 - (b2 - lbb)
                else:
                    lba = a1
            if b1 < ubb <= b2:
                if stype == b'=':
                    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.InputError(_(b'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, b'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 = b'!'
            if opts.ignoreblanklines:
                if lines1 is None:
                    lines1 = splitnewlines(text1)
                if lines2 is None:
                    lines2 = splitnewlines(text2)
                old = wsclean(opts, b"".join(lines1[s[0] : s[1]]))
                new = wsclean(opts, b"".join(lines2[s[2] : s[3]]))
                if old == new:
                    type = b'~'
            yield s, type
        yield s1, b'='


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 b'\t%s' % date
        if fn and b' ' in fn:
            return b'\t'
        return b''

    sentinel = [], ()
    if not a and not b:
        return sentinel

    if opts.noprefix:
        aprefix = bprefix = b''
    else:
        aprefix = b'a/'
        bprefix = b'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, [b'Binary file %s has changed\n' % fn1]),)
    elif not a:
        without_newline = not b.endswith(b'\n')
        b = splitnewlines(b)
        if a is None:
            l1 = b'--- /dev/null%s' % datetag(epoch)
        else:
            l1 = b"--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1))
        l2 = b"+++ %s%s" % (bprefix + fn2, datetag(bd, fn2))
        headerlines = [l1, l2]
        size = len(b)
        hunkrange = (0, 0, 1, size)
        hunklines = [b"@@ -0,0 +1,%d @@\n" % size] + [b"+" + e for e in b]
        if without_newline:
            hunklines[-1] += b'\n'
            hunklines.append(diffhelper.MISSING_NEWLINE_MARKER)
        hunks = ((hunkrange, hunklines),)
    elif not b:
        without_newline = not a.endswith(b'\n')
        a = splitnewlines(a)
        l1 = b"--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1))
        if b is None:
            l2 = b'+++ /dev/null%s' % datetag(epoch)
        else:
            l2 = b"+++ %s%s%s" % (bprefix, fn2, datetag(bd, fn2))
        headerlines = [l1, l2]
        size = len(a)
        hunkrange = (1, size, 0, 0)
        hunklines = [b"@@ -1,%d +0,0 @@\n" % size] + [b"-" + e for e in a]
        if without_newline:
            hunklines[-1] += b'\n'
            hunklines.append(diffhelper.MISSING_NEWLINE_MARKER)
        hunks = ((hunkrange, hunklines),)
    else:
        hunks = _unidiff(a, b, opts=opts)
        if not next(hunks):
            return sentinel

        headerlines = [
            b"--- %s%s%s" % (aprefix, fn1, datetag(ad, fn1)),
            b"+++ %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, b'']

    def yieldhunk(hunk):
        (astart, a2, bstart, b2, delta) = hunk
        aend = contextend(a2, len(l1))
        alen = aend - astart
        blen = b2 - bstart + aend - a2

        func = b""
        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 = (
            [b"@@ -%d,%d +%d,%d @@%s\n" % (hunkrange + (func,))]
            + delta
            + [b' ' + 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(b'\n') and astart + alen == len(l1) + 1:
            for i in pycompat.xrange(len(hunklines) - 1, -1, -1):
                if hunklines[i].startswith((b'-', b' ')):
                    if hunklines[i].startswith(b' '):
                        skip = True
                    hunklines[i] += b'\n'
                    hunklines.insert(i + 1, diffhelper.MISSING_NEWLINE_MARKER)
                    break
        if not skip and not t2.endswith(b'\n') and bstart + blen == len(l2) + 1:
            for i in pycompat.xrange(len(hunklines) - 1, -1, -1):
                if hunklines[i].startswith(b'+'):
                    hunklines[i] += b'\n'
                    hunklines.insert(i + 1, diffhelper.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 != b'!':
            if stype == b'~':
                # 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) :] = [b' ' + x for x in l1[astart:a1]]
        delta[len(delta) :] = [b'-' + x for x in old]
        delta[len(delta) :] = [b'+' + 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(b'A') + l - 1)
        else:
            l = pycompat.bytechr(l - 26 + ord(b'a') - 1)
        return b'%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 = b''
    if tn is None:
        tn = b''

    if to == tn:
        return b''

    # TODO: deltas
    ret = []
    ret.append(b'GIT binary patch\n')
    ret.append(b'literal %d\n' % len(tn))
    for l in chunk(zlib.compress(tn)):
        ret.append(fmtline(l))
    ret.append(b'\n')

    return b''.join(ret)


def patchtext(bin):
    pos = 0
    t = []
    while pos < len(bin):
        p1, p2, l = struct.unpack(b">lll", bin[pos : pos + 12])
        pos += 12
        t.append(bin[pos : pos + l])
        pos += l
    return b"".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(b">lll", 0, 0, length) if length else b''


def replacediffheader(oldlen, newlen):
    return struct.pack(b">lll", 0, oldlen, newlen)