annotate mercurial/pure/bdiff.py @ 30212:260af19891f2

changegroup: increase write buffer size to 128k

By default, Python defers to the operating system for choosing the
default buffer size on opened files. On my Linux machine, the default
is 4k, which is really small for 2016.

This patch bumps the write buffer size when writing
changegroups/bundles to 128k. This matches the 128k read buffer
we already use on revlogs.

It's worth noting that this only impacts when writing to an explicit
file (such as during `hg bundle`). Buffers when writing to bundle
files via the repo vfs or to a temporary file are not impacted.

When producing a none-v2 bundle file of the mozilla-unified repository,
this change caused the number of write() system calls to drop from
952,449 to 29,788. After this change, the most frequent system
calls are fstat(), read(), lseek(), and open(). There were
2,523,672 system calls after this patch (so a net decrease of
~950k is statistically significant).

This change shows no performance change on my system. But I have a
high-end system with a fast SSD. It is quite possible this change
will have a significant impact on network file systems, where
extra network round trips due to excessive I/O system calls could
introduce significant latency.

# bdiff.py - Python implementation of bdiff.c
#
# Copyright 2009 Matt Mackall <[e-mail]> and others
#
# 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 array
import difflib
import re
import struct

from . import policy
policynocffi = policy.policynocffi
modulepolicy = policy.policy

def splitnewlines(text):
    '''like str.splitlines, but only split on newlines.'''
    lines = [l + '\n' for l in text.split('\n')]
    if lines:
        if lines[-1] == '\n':
            lines.pop()
        else:
            lines[-1] = lines[-1][:-1]
    return lines

def _normalizeblocks(a, b, blocks):
    prev = None
    r = []
    for curr in blocks:
        if prev is None:
            prev = curr
            continue
        shift = 0

        a1, b1, l1 = prev
        a1end = a1 + l1
        b1end = b1 + l1

        a2, b2, l2 = curr
        a2end = a2 + l2
        b2end = b2 + l2
        if a1end == a2:
            while (a1end + shift < a2end and
                   a[a1end + shift] == b[b1end + shift]):
                shift += 1
        elif b1end == b2:
            while (b1end + shift < b2end and
                   a[a1end + shift] == b[b1end + shift]):
                shift += 1
        r.append((a1, b1, l1 + shift))
        prev = a2 + shift, b2 + shift, l2 - shift
    r.append(prev)
    return r

def _tostring(c):
    if type(c) is array.array:
        # this copy overhead isn't ideal
        return c.tostring()
    return str(c)

def bdiff(a, b):
    a = _tostring(a).splitlines(True)
    b = _tostring(b).splitlines(True)

    if not a:
        s = "".join(b)
        return s and (struct.pack(">lll", 0, 0, len(s)) + s)

    bin = []
    p = [0]
    for i in a: p.append(p[-1] + len(i))

    d = difflib.SequenceMatcher(None, a, b).get_matching_blocks()
    d = _normalizeblocks(a, b, d)
    la = 0
    lb = 0
    for am, bm, size in d:
        s = "".join(b[lb:bm])
        if am > la or s:
            bin.append(struct.pack(">lll", p[la], p[am], len(s)) + s)
        la = am + size
        lb = bm + size

    return "".join(bin)

def blocks(a, b):
    an = splitnewlines(a)
    bn = splitnewlines(b)
    d = difflib.SequenceMatcher(None, an, bn).get_matching_blocks()
    d = _normalizeblocks(an, bn, d)
    return [(i, i + n, j, j + n) for (i, j, n) in d]

def fixws(text, allws):
    if allws:
        text = re.sub('[ \t\r]+', '', text)
    else:
        text = re.sub('[ \t\r]+', ' ', text)
        text = text.replace(' \n', '\n')
    return text

if modulepolicy not in policynocffi:
    try:
        from _bdiff_cffi import ffi, lib
    except ImportError:
        if modulepolicy == 'cffi': # strict cffi import
            raise
    else:
        def blocks(sa, sb):
            a = ffi.new("struct bdiff_line**")
            b = ffi.new("struct bdiff_line**")
            ac = ffi.new("char[]", str(sa))
            bc = ffi.new("char[]", str(sb))
            l = ffi.new("struct bdiff_hunk*")
            try:
                an = lib.bdiff_splitlines(ac, len(sa), a)
                bn = lib.bdiff_splitlines(bc, len(sb), b)
                if not a[0] or not b[0]:
                    raise MemoryError
                count = lib.bdiff_diff(a[0], an, b[0], bn, l)
                if count < 0:
                    raise MemoryError
                rl = [None] * count
                h = l.next
                i = 0
                while h:
                    rl[i] = (h.a1, h.a2, h.b1, h.b2)
                    h = h.next
                    i += 1
            finally:
                lib.free(a[0])
                lib.free(b[0])
                lib.bdiff_freehunks(l.next)
            return rl

        def bdiff(sa, sb):
            a = ffi.new("struct bdiff_line**")
            b = ffi.new("struct bdiff_line**")
            ac = ffi.new("char[]", str(sa))
            bc = ffi.new("char[]", str(sb))
            l = ffi.new("struct bdiff_hunk*")
            try:
                an = lib.bdiff_splitlines(ac, len(sa), a)
                bn = lib.bdiff_splitlines(bc, len(sb), b)
                if not a[0] or not b[0]:
                    raise MemoryError
                count = lib.bdiff_diff(a[0], an, b[0], bn, l)
                if count < 0:
                    raise MemoryError
                rl = []
                h = l.next
                la = lb = 0
                while h:
                    if h.a1 != la or h.b1 != lb:
                        lgt = (b[0] + h.b1).l - (b[0] + lb).l
                        rl.append(struct.pack(">lll", (a[0] + la).l - a[0].l,
                            (a[0] + h.a1).l - a[0].l, lgt))
                        rl.append(str(ffi.buffer((b[0] + lb).l, lgt)))
                    la = h.a2
                    lb = h.b2
                    h = h.next

            finally:
                lib.free(a[0])
                lib.free(b[0])
                lib.bdiff_freehunks(l.next)
            return "".join(rl)