view mercurial/linelog.py @ 39515:93486cc46125

treemanifest: introduce lazy loading of subdirs An earlier patch series made it so that what to load was up to the calling code, which works fine until manifests are copied - when they're copied, they're loaded completely and thus we lose the entire benefit. By lazy loading everything, we can avoid having to pass in the matcher to ~every manifest function, and handle copies correctly as well. This changeset doesn't go as far as it could with loading only the necessary subsets, that will happen in later changes in this series; at the moment, except in a few situations, we just load everything the moment we want to interact with treemanifest._dirs. This is thus most likely to be a small slowdown if treemanifests is in use regardless of whether narrow is in use, but hopefully easier to verify correctness and review. This is part of a series of speedups, it is not expected to produce any real speed improvements itself, but the numbers show that it doesn't produce a large speed penalty in any common case, and for the cases it does provide a penalty in, it is not a large absolute amount (even if it is a large percentage amount). Timing numbers according to command: hyperfine --prepare <preparation_script> 'hg status' HGRCPATH points to a file with the following contents: [extensions] narrow = strip = rebase = mozilla-unified (called m-u below) was at revision #468856. regular hash: eb39298e432d treemanifests hash: 0553b7f29eaf large-dir-repo (called l-d-r below) was generated with the following script: #!/bin/bash hg init large-dir-repo mkdir -p large-dir-repo/third_party/rust/log touch large-dir-repo/third_party/rust/log/foo.txt for i in $(seq 1 30000); do d=$(mktemp -d large-dir-repo/third_party/XXXXXXXXX) touch $d/file.txt done hg -R large-dir-repo ci -Am 'rev0' --user test --date '0 0' echo hi > large-dir-repo/third_party/rust/log/bar.txt hg -R large-dir-repo ci -Am 'rev1' --user test --date '0 0' echo hi > large-dir-repo/third_party/rust/log/baz.txt hg -R large-dir-repo ci -Am 'rev2' --user test --date '0 0' for the repos that use narrow, the narrowspec was this: [include] rootfilesin:accessible/jsat rootfilesin:accessible/tests/mochitest/jsat rootfilesin:mobile/android/chrome/content rootfilesin:mobile/android/modules/geckoview rootfilesin:third_party/rust/log [exclude] This narrowspec was chosen due to the size of the third_party/rust directory (this directory was *not* modified in revision #468856 in mozilla-unified), plus all the directories that *were* modified in revision #468856 of mozilla-unified. Importantly, when using narrow, these repos had everything checked out (in the case of large-dir-repo, that means all 30,001 directories), *before* adding the narrowspec. This is to simulate the behavior when using a virtual filesystem that shows everything for the user even if they haven't added it to the narrowspec yet. This is not a supported configuration, and `hg update` and `hg rebase` will not really do the "correct" thing if there are mutations outside of the narrowspec (which is not the case in these tests, due to a carefully crafted narrowspec), but non-mutating commands should behave correctly. I'm not claiming anything less than a 5% speed win as improvements due to this change; these are probably eiter measurement artifacts or constant time improvements. The numbers that aren't changing are shown primarily to prove that this doesn't make anything worse in any case I plan on testing during this series. 'before' is hg from commit 6268fed3 'N' indicates narrow in use 'T' indicates treemanifest in use Please note that these commands and the narrowspec are a little different than the ones in a similar table that I made in a3cabe9415e1. Important: it is my understanding that these numbers below are *not super reliable*, the large slowdowns may be artifacts of some odd interaction between GC and python module/code complexity. Another changeset of mine (D4351) had shown large timing differences when ~empty, uncalled functions were added to match.py, though only when using --color=never or redirecting to /dev/null. We seem to be on some cusp of complexity or code size that is causing, at my best guess (according to linux `perf` benchmarks) GC to alter behavior and cause a 200-400ms difference in timings. I haven't had a chance to replicate these results on another machine. diff --git: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 1.580 s +- 0.034 s | 1.576 s +- 0.022 s | 99.7% m-u | | x | 1.568 s +- 0.025 s | 1.584 s +- 0.044 s | 101.0% m-u | x | | 1.569 s +- 0.031 s | 1.554 s +- 0.025 s | 99.0% m-u | x | x | 107.3 ms +- 1.6 ms | 106.3 ms +- 1.5 ms | 99.1% l-d-r | | | 232.5 ms +- 5.9 ms | 233.5 ms +- 5.3 ms | 100.4% l-d-r | | x | 236.6 ms +- 6.3 ms | 233.6 ms +- 7.0 ms | 98.7% l-d-r | x | | 118.4 ms +- 2.1 ms | 118.4 ms +- 1.4 ms | 100.0% l-d-r | x | x | 116.8 ms +- 1.5 ms | 118.9 ms +- 1.6 ms | 101.8% diff -c . --git: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 354.4 ms +- 16.6 ms | 351.0 ms +- 6.9 ms | 99.0% m-u | | x | 207.2 ms +- 3.0 ms | 206.2 ms +- 2.7 ms | 99.5% m-u | x | | 422.0 ms +- 26.0 ms | 351.2 ms +- 6.4 ms | 83.2% <-- m-u | x | x | 166.7 ms +- 2.1 ms | 169.5 ms +- 4.1 ms | 101.7% l-d-r | | | 98.4 ms +- 4.5 ms | 98.5 ms +- 2.1 ms | 100.1% l-d-r | | x | 5.519 s +- 0.060 s | 5.149 s +- 0.042 s | 93.3% <-- l-d-r | x | | 99.1 ms +- 3.2 ms | 102.6 ms +- 9.7 ms | 103.5% <--? l-d-r | x | x | 994.9 ms +- 10.7 ms | 1.026 s +- 0.012 s | 103.1% <--? rebase -r . --keep -d .^^: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 6.639 s +- 0.168 s | 6.559 s +- 0.097 s | 98.8% m-u | | x | 6.601 s +- 0.143 s | 6.640 s +- 0.207 s | 100.6% m-u | x | | 6.582 s +- 0.098 s | 6.543 s +- 0.098 s | 99.4% m-u | x | x | 678.4 ms +- 57.7 ms | 703.7 ms +- 52.4 ms | 103.7% <--? l-d-r | | | 780.0 ms +- 23.9 ms | 776.0 ms +- 12.6 ms | 99.5% l-d-r | | x | 7.520 s +- 0.255 s | 7.395 s +- 0.044 s | 98.3% l-d-r | x | | 331.9 ms +- 16.5 ms | 327.0 ms +- 3.4 ms | 98.5% l-d-r | x | x | 6.228 s +- 0.113 s | 5.924 s +- 0.044 s | 95.1% status --change . --copies: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 330.8 ms +- 7.2 ms | 329.0 ms +- 7.1 ms | 99.5% m-u | | x | 182.9 ms +- 2.7 ms | 183.5 ms +- 2.7 ms | 100.3% m-u | x | | 330.0 ms +- 7.6 ms | 327.1 ms +- 5.4 ms | 99.1% m-u | x | x | 146.2 ms +- 2.4 ms | 147.1 ms +- 1.3 ms | 100.6% l-d-r | | | 95.3 ms +- 1.4 ms | 95.9 ms +- 1.5 ms | 100.6% l-d-r | | x | 5.157 s +- 0.035 s | 5.166 s +- 0.058 s | 100.2% l-d-r | x | | 99.7 ms +- 3.0 ms | 100.2 ms +- 4.4 ms | 100.5% l-d-r | x | x | 993.6 ms +- 13.1 ms | 1.025 s +- 0.015 s | 103.2% <--? status --copies: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 2.348 s +- 0.031 s | 2.329 s +- 0.019 s | 99.2% m-u | | x | 2.337 s +- 0.026 s | 2.346 s +- 0.034 s | 100.4% m-u | x | | 2.354 s +- 0.015 s | 2.342 s +- 0.021 s | 99.5% m-u | x | x | 120.6 ms +- 4.3 ms | 119.2 ms +- 2.1 ms | 98.8% l-d-r | | | 731.5 ms +- 11.1 ms | 719.6 ms +- 9.8 ms | 98.4% l-d-r | | x | 729.0 ms +- 15.5 ms | 725.7 ms +- 10.6 ms | 99.5% l-d-r | x | | 211.0 ms +- 3.9 ms | 212.8 ms +- 3.7 ms | 100.9% l-d-r | x | x | 211.5 ms +- 4.2 ms | 211.0 ms +- 3.3 ms | 99.8% update $rev^; ~/src/hg/hg{hg}/hg update $rev: repo | N | T | before (mean +- stdev) | after (mean +- stdev) | % of before ------+---+---+------------------------+-----------------------+------------ m-u | | | 3.910 s +- 0.055 s | 3.920 s +- 0.075 s | 100.3% m-u | | x | 3.613 s +- 0.056 s | 3.630 s +- 0.056 s | 100.5% m-u | x | | 3.873 s +- 0.055 s | 3.864 s +- 0.049 s | 99.8% m-u | x | x | 400.4 ms +- 7.4 ms | 403.6 ms +- 5.0 ms | 100.8% l-d-r | | | 531.6 ms +- 10.0 ms | 528.8 ms +- 9.6 ms | 99.5% l-d-r | | x | 10.377 s +- 0.049 s | 9.955 s +- 0.046 s | 95.9% l-d-r | x | | 308.3 ms +- 4.4 ms | 306.8 ms +- 3.7 ms | 99.5% l-d-r | x | x | 1.805 s +- 0.015 s | 1.834 s +- 0.020 s | 101.6% Differential Revision: https://phab.mercurial-scm.org/D4366
author spectral <spectral@google.com>
date Thu, 16 Aug 2018 12:31:52 -0700
parents ee97f7a677f3
children 2372284d9457
line wrap: on
line source

# linelog - efficient cache for annotate data
#
# Copyright 2018 Google LLC.
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.
"""linelog is an efficient cache for annotate data inspired by SCCS Weaves.

SCCS Weaves are an implementation of
https://en.wikipedia.org/wiki/Interleaved_deltas. See
mercurial/help/internals/linelog.txt for an exploration of SCCS weaves
and how linelog works in detail.

Here's a hacker's summary: a linelog is a program which is executed in
the context of a revision. Executing the program emits information
about lines, including the revision that introduced them and the line
number in the file at the introducing revision. When an insertion or
deletion is performed on the file, a jump instruction is used to patch
in a new body of annotate information.
"""
from __future__ import absolute_import, print_function

import abc
import struct

from .thirdparty import (
    attr,
)
from . import (
    pycompat,
)

_llentry = struct.Struct('>II')

class LineLogError(Exception):
    """Error raised when something bad happens internally in linelog."""

@attr.s
class lineinfo(object):
    # Introducing revision of this line.
    rev = attr.ib()
    # Line number for this line in its introducing revision.
    linenum = attr.ib()
    # Private. Offset in the linelog program of this line. Used internally.
    _offset = attr.ib()

@attr.s
class annotateresult(object):
    rev = attr.ib()
    lines = attr.ib()
    _eof = attr.ib()

    def __iter__(self):
        return iter(self.lines)

class _llinstruction(object):

    __metaclass__ = abc.ABCMeta

    @abc.abstractmethod
    def __init__(self, op1, op2):
        pass

    @abc.abstractmethod
    def __str__(self):
        pass

    def __repr__(self):
        return str(self)

    @abc.abstractmethod
    def __eq__(self, other):
        pass

    @abc.abstractmethod
    def encode(self):
        """Encode this instruction to the binary linelog format."""

    @abc.abstractmethod
    def execute(self, rev, pc, emit):
        """Execute this instruction.

        Args:
          rev: The revision we're annotating.
          pc: The current offset in the linelog program.
          emit: A function that accepts a single lineinfo object.

        Returns:
          The new value of pc. Returns None if exeuction should stop
          (that is, we've found the end of the file.)
        """

class _jge(_llinstruction):
    """If the current rev is greater than or equal to op1, jump to op2."""

    def __init__(self, op1, op2):
        self._cmprev = op1
        self._target = op2

    def __str__(self):
        return r'JGE %d %d' % (self._cmprev, self._target)

    def __eq__(self, other):
        return (type(self) == type(other)
                and self._cmprev == other._cmprev
                and self._target == other._target)

    def encode(self):
        return _llentry.pack(self._cmprev << 2, self._target)

    def execute(self, rev, pc, emit):
        if rev >= self._cmprev:
            return self._target
        return pc + 1

class _jump(_llinstruction):
    """Unconditional jumps are expressed as a JGE with op1 set to 0."""

    def __init__(self, op1, op2):
        if op1 != 0:
            raise LineLogError("malformed JUMP, op1 must be 0, got %d" % op1)
        self._target = op2

    def __str__(self):
        return r'JUMP %d' % (self._target)

    def __eq__(self, other):
        return (type(self) == type(other)
                and self._target == other._target)

    def encode(self):
        return _llentry.pack(0, self._target)

    def execute(self, rev, pc, emit):
        return self._target

class _eof(_llinstruction):
    """EOF is expressed as a JGE that always jumps to 0."""

    def __init__(self, op1, op2):
        if op1 != 0:
            raise LineLogError("malformed EOF, op1 must be 0, got %d" % op1)
        if op2 != 0:
            raise LineLogError("malformed EOF, op2 must be 0, got %d" % op2)

    def __str__(self):
        return r'EOF'

    def __eq__(self, other):
        return type(self) == type(other)

    def encode(self):
        return _llentry.pack(0, 0)

    def execute(self, rev, pc, emit):
        return None

class _jl(_llinstruction):
    """If the current rev is less than op1, jump to op2."""

    def __init__(self, op1, op2):
        self._cmprev = op1
        self._target = op2

    def __str__(self):
        return r'JL %d %d' % (self._cmprev, self._target)

    def __eq__(self, other):
        return (type(self) == type(other)
                and self._cmprev == other._cmprev
                and self._target == other._target)

    def encode(self):
        return _llentry.pack(1 | (self._cmprev << 2), self._target)

    def execute(self, rev, pc, emit):
        if rev < self._cmprev:
            return self._target
        return pc + 1

class _line(_llinstruction):
    """Emit a line."""

    def __init__(self, op1, op2):
        # This line was introduced by this revision number.
        self._rev = op1
        # This line had the specified line number in the introducing revision.
        self._origlineno = op2

    def __str__(self):
        return r'LINE %d %d' % (self._rev, self._origlineno)

    def __eq__(self, other):
        return (type(self) == type(other)
                and self._rev == other._rev
                and self._origlineno == other._origlineno)

    def encode(self):
        return _llentry.pack(2 | (self._rev << 2), self._origlineno)

    def execute(self, rev, pc, emit):
        emit(lineinfo(self._rev, self._origlineno, pc))
        return pc + 1

def _decodeone(data, offset):
    """Decode a single linelog instruction from an offset in a buffer."""
    try:
        op1, op2 = _llentry.unpack_from(data, offset)
    except struct.error as e:
        raise LineLogError('reading an instruction failed: %r' % e)
    opcode = op1 & 0b11
    op1 = op1 >> 2
    if opcode == 0:
        if op1 == 0:
            if op2 == 0:
                return _eof(op1, op2)
            return _jump(op1, op2)
        return _jge(op1, op2)
    elif opcode == 1:
        return _jl(op1, op2)
    elif opcode == 2:
        return _line(op1, op2)
    raise NotImplementedError('Unimplemented opcode %r' % opcode)

class linelog(object):
    """Efficient cache for per-line history information."""

    def __init__(self, program=None, maxrev=0):
        if program is None:
            # We pad the program with an extra leading EOF so that our
            # offsets will match the C code exactly. This means we can
            # interoperate with the C code.
            program = [_eof(0, 0), _eof(0, 0)]
        self._program = program
        self._lastannotate = None
        self._maxrev = maxrev

    def __eq__(self, other):
        return (type(self) == type(other)
                and self._program == other._program
                and self._maxrev == other._maxrev)

    def __repr__(self):
        return '<linelog at %s: maxrev=%d size=%d>' % (
            hex(id(self)), self._maxrev, len(self._program))

    def debugstr(self):
        fmt = r'%%%dd %%s' % len(str(len(self._program)))
        return pycompat.sysstr('\n').join(
            fmt % (idx, i) for idx, i in enumerate(self._program[1:], 1))

    @classmethod
    def fromdata(cls, buf):
        if len(buf) % _llentry.size != 0:
            raise LineLogError(
                "invalid linelog buffer size %d (must be a multiple of %d)" % (
                    len(buf), _llentry.size))
        expected = len(buf) / _llentry.size
        fakejge = _decodeone(buf, 0)
        if isinstance(fakejge, _jump):
            maxrev = 0
        else:
            maxrev = fakejge._cmprev
        numentries = fakejge._target
        if expected != numentries:
            raise LineLogError("corrupt linelog data: claimed"
                               " %d entries but given data for %d entries" % (
                                   expected, numentries))
        instructions = [_eof(0, 0)]
        for offset in pycompat.xrange(1, numentries):
            instructions.append(_decodeone(buf, offset * _llentry.size))
        return cls(instructions, maxrev=maxrev)

    def encode(self):
        hdr = _jge(self._maxrev, len(self._program)).encode()
        return hdr + ''.join(i.encode() for i in self._program[1:])

    def clear(self):
        self._program = []
        self._maxrev = 0
        self._lastannotate = None

    def replacelines_vec(self, rev, a1, a2, blines):
        return self.replacelines(rev, a1, a2, 0, len(blines),
                                 _internal_blines=blines)

    def replacelines(self, rev, a1, a2, b1, b2, _internal_blines=None):
        """Replace lines [a1, a2) with lines [b1, b2)."""
        if self._lastannotate:
            # TODO(augie): make replacelines() accept a revision at
            # which we're editing as well as a revision to mark
            # responsible for the edits. In hg-experimental it's
            # stateful like this, so we're doing the same thing to
            # retain compatibility with absorb until that's imported.
            ar = self._lastannotate
        else:
            ar = self.annotate(rev)
            #        ar = self.annotate(self._maxrev)
        if a1 > len(ar.lines):
            raise LineLogError(
                '%d contains %d lines, tried to access line %d' % (
                    rev, len(ar.lines), a1))
        elif a1 == len(ar.lines):
            # Simulated EOF instruction since we're at EOF, which
            # doesn't have a "real" line.
            a1inst = _eof(0, 0)
            a1info = lineinfo(0, 0, ar._eof)
        else:
            a1info = ar.lines[a1]
            a1inst = self._program[a1info._offset]
        programlen = self._program.__len__
        oldproglen = programlen()
        appendinst = self._program.append

        # insert
        blineinfos = []
        bappend = blineinfos.append
        if b1 < b2:
            # Determine the jump target for the JGE at the start of
            # the new block.
            tgt = oldproglen + (b2 - b1 + 1)
            # Jump to skip the insert if we're at an older revision.
            appendinst(_jl(rev, tgt))
            for linenum in pycompat.xrange(b1, b2):
                if _internal_blines is None:
                    bappend(lineinfo(rev, linenum, programlen()))
                    appendinst(_line(rev, linenum))
                else:
                    newrev, newlinenum = _internal_blines[linenum]
                    bappend(lineinfo(newrev, newlinenum, programlen()))
                    appendinst(_line(newrev, newlinenum))
        # delete
        if a1 < a2:
            if a2 > len(ar.lines):
                raise LineLogError(
                    '%d contains %d lines, tried to access line %d' % (
                        rev, len(ar.lines), a2))
            elif a2 == len(ar.lines):
                endaddr = ar._eof
            else:
                endaddr = ar.lines[a2]._offset
            if a2 > 0 and rev < self._maxrev:
                # If we're here, we're deleting a chunk of an old
                # commit, so we need to be careful and not touch
                # invisible lines between a2-1 and a2 (IOW, lines that
                # are added later).
                endaddr = ar.lines[a2 - 1]._offset + 1
            appendinst(_jge(rev, endaddr))
        # copy instruction from a1
        a1instpc = programlen()
        appendinst(a1inst)
        # if a1inst isn't a jump or EOF, then we need to add an unconditional
        # jump back into the program here.
        if not isinstance(a1inst, (_jump, _eof)):
            appendinst(_jump(0, a1info._offset + 1))
        # Patch instruction at a1, which makes our patch live.
        self._program[a1info._offset] = _jump(0, oldproglen)

        # Update self._lastannotate in place. This serves as a cache to avoid
        # expensive "self.annotate" in this function, when "replacelines" is
        # used continuously.
        if len(self._lastannotate.lines) > a1:
            self._lastannotate.lines[a1]._offset = a1instpc
        else:
            assert isinstance(a1inst, _eof)
            self._lastannotate._eof = a1instpc
        self._lastannotate.lines[a1:a2] = blineinfos
        self._lastannotate.rev = max(self._lastannotate.rev, rev)

        if rev > self._maxrev:
            self._maxrev = rev

    def annotate(self, rev):
        pc = 1
        lines = []
        executed = 0
        # Sanity check: if instructions executed exceeds len(program), we
        # hit an infinite loop in the linelog program somehow and we
        # should stop.
        while pc is not None and executed < len(self._program):
            inst = self._program[pc]
            lastpc = pc
            pc = inst.execute(rev, pc, lines.append)
            executed += 1
        if pc is not None:
            raise LineLogError(
                r'Probably hit an infinite loop in linelog. Program:\n' +
                self.debugstr())
        ar = annotateresult(rev, lines, lastpc)
        self._lastannotate = ar
        return ar

    @property
    def maxrev(self):
        return self._maxrev

    # Stateful methods which depend on the value of the last
    # annotation run. This API is for compatiblity with the original
    # linelog, and we should probably consider refactoring it.
    @property
    def annotateresult(self):
        """Return the last annotation result. C linelog code exposed this."""
        return [(l.rev, l.linenum) for l in self._lastannotate.lines]

    def getoffset(self, line):
        return self._lastannotate.lines[line]._offset

    def getalllines(self, start=0, end=0):
        """Get all lines that ever occurred in [start, end).

        Passing start == end == 0 means "all lines ever".

        This works in terms of *internal* program offsets, not line numbers.
        """
        pc = start or 1
        lines = []
        # only take as many steps as there are instructions in the
        # program - if we don't find an EOF or our stop-line before
        # then, something is badly broken.
        for step in pycompat.xrange(len(self._program)):
            inst = self._program[pc]
            nextpc = pc + 1
            if isinstance(inst, _jump):
                nextpc = inst._target
            elif isinstance(inst, _eof):
                return lines
            elif isinstance(inst, (_jl, _jge)):
                pass
            elif isinstance(inst, _line):
                lines.append((inst._rev, inst._origlineno))
            else:
                raise LineLogError("Illegal instruction %r" % inst)
            if nextpc == end:
                return lines
            pc = nextpc
        raise LineLogError("Failed to perform getalllines")