view hgext/convert/darcs.py @ 26380:56a640b0f656

revlog: don't flush data file after every added revision

The current behavior of revlogs is to flush the data file when writing
data to it. Tracing system calls revealed that changegroup processing
incurred numerous write(2) calls for values much smaller than the
default buffer size (Python defaults to 4096, but it can be adjusted
based on detected block size at run time by CPython).

The reason we flush revlogs is so readers have all data available.
For example, the current code in revlog.py will re-open the revlog
file (instead of seeking an existing file handle) to read the text
of a revision. This happens when starting a new delta chain when
adding several revisions from changegroups, for example. Yes, this
is likely sub-optimal (we should probably be sharing file descriptors
between readers and writers to avoid the flushing and associated
overhead of re-opening files).

While flushing revlogs is necessary, it appears all callers are
diligent about flushing files before a read is performed (see
buildtext() in _addrevision()), making the flush in
_writeentry() redundant and unncessary. So, we remove it. In practice,
this means we incur a write(2) a) when the buffer is full (typically
4096 bytes) b) when a new delta chain is created rather than after
every added revision. This applies to every revlog, but by volume
it mostly impacts filelogs.

Removing the redundant flush from _writeentry() significantly
reduces the number of write(2) calls during changegroup processing on
my Linux machine. When applying a changegroup of the hg repo based on
my local repo, the total number of write(2) calls during application
of the mercurial/localrepo.py revlogs dropped from 1,320 to 217 with
this patch applied. Total I/O related system calls dropped from 1,577
to 474.

When unbundling a mozilla-central gzipped bundle (264,403 changesets
with 1,492,215 changes to 222,507 files), total write(2) calls
dropped from 1,252,881 to 827,106 and total system calls dropped from
3,601,259 to 3,178,636 - a reduction of 425,775!

While the system call reduction is significant, it appears
to have no impact on wall time on my Linux and Windows machines. Still,
fewer syscalls is fewer syscalls. Surely this can't hurt. If nothing
else, it makes examining remaining system call usage simpler and opens
the door to experimenting with the performance impact of different
buffer sizes.

# darcs.py - darcs support for the convert extension
#
#  Copyright 2007-2009 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 common import NoRepo, checktool, commandline, commit, converter_source
from mercurial.i18n import _
from mercurial import util
import os, shutil, tempfile, re, errno

# The naming drift of ElementTree is fun!

try:
    from xml.etree.cElementTree import ElementTree, XMLParser
except ImportError:
    try:
        from xml.etree.ElementTree import ElementTree, XMLParser
    except ImportError:
        try:
            from elementtree.cElementTree import ElementTree, XMLParser
        except ImportError:
            try:
                from elementtree.ElementTree import ElementTree, XMLParser
            except ImportError:
                pass

class darcs_source(converter_source, commandline):
    def __init__(self, ui, path, revs=None):
        converter_source.__init__(self, ui, path, revs=revs)
        commandline.__init__(self, ui, 'darcs')

        # check for _darcs, ElementTree so that we can easily skip
        # test-convert-darcs if ElementTree is not around
        if not os.path.exists(os.path.join(path, '_darcs')):
            raise NoRepo(_("%s does not look like a darcs repository") % path)

        checktool('darcs')
        version = self.run0('--version').splitlines()[0].strip()
        if version < '2.1':
            raise util.Abort(_('darcs version 2.1 or newer needed (found %r)') %
                             version)

        if "ElementTree" not in globals():
            raise util.Abort(_("Python ElementTree module is not available"))

        self.path = os.path.realpath(path)

        self.lastrev = None
        self.changes = {}
        self.parents = {}
        self.tags = {}

        # Check darcs repository format
        format = self.format()
        if format:
            if format in ('darcs-1.0', 'hashed'):
                raise NoRepo(_("%s repository format is unsupported, "
                               "please upgrade") % format)
        else:
            self.ui.warn(_('failed to detect repository format!'))

    def before(self):
        self.tmppath = tempfile.mkdtemp(
            prefix='convert-' + os.path.basename(self.path) + '-')
        output, status = self.run('init', repodir=self.tmppath)
        self.checkexit(status)

        tree = self.xml('changes', xml_output=True, summary=True,
                        repodir=self.path)
        tagname = None
        child = None
        for elt in tree.findall('patch'):
            node = elt.get('hash')
            name = elt.findtext('name', '')
            if name.startswith('TAG '):
                tagname = name[4:].strip()
            elif tagname is not None:
                self.tags[tagname] = node
                tagname = None
            self.changes[node] = elt
            self.parents[child] = [node]
            child = node
        self.parents[child] = []

    def after(self):
        self.ui.debug('cleaning up %s\n' % self.tmppath)
        shutil.rmtree(self.tmppath, ignore_errors=True)

    def recode(self, s, encoding=None):
        if isinstance(s, unicode):
            # XMLParser returns unicode objects for anything it can't
            # encode into ASCII. We convert them back to str to get
            # recode's normal conversion behavior.
            s = s.encode('latin-1')
        return super(darcs_source, self).recode(s, encoding)

    def xml(self, cmd, **kwargs):
        # NOTE: darcs is currently encoding agnostic and will print
        # patch metadata byte-for-byte, even in the XML changelog.
        etree = ElementTree()
        # While we are decoding the XML as latin-1 to be as liberal as
        # possible, etree will still raise an exception if any
        # non-printable characters are in the XML changelog.
        parser = XMLParser(encoding='latin-1')
        p = self._run(cmd, **kwargs)
        etree.parse(p.stdout, parser=parser)
        p.wait()
        self.checkexit(p.returncode)
        return etree.getroot()

    def format(self):
        output, status = self.run('show', 'repo', no_files=True,
                                  repodir=self.path)
        self.checkexit(status)
        m = re.search(r'^\s*Format:\s*(.*)$', output, re.MULTILINE)
        if not m:
            return None
        return ','.join(sorted(f.strip() for f in m.group(1).split(',')))

    def manifest(self):
        man = []
        output, status = self.run('show', 'files', no_directories=True,
                                  repodir=self.tmppath)
        self.checkexit(status)
        for line in output.split('\n'):
            path = line[2:]
            if path:
                man.append(path)
        return man

    def getheads(self):
        return self.parents[None]

    def getcommit(self, rev):
        elt = self.changes[rev]
        date = util.strdate(elt.get('local_date'), '%a %b %d %H:%M:%S %Z %Y')
        desc = elt.findtext('name') + '\n' + elt.findtext('comment', '')
        # etree can return unicode objects for name, comment, and author,
        # so recode() is used to ensure str objects are emitted.
        return commit(author=self.recode(elt.get('author')),
                      date=util.datestr(date, '%Y-%m-%d %H:%M:%S %1%2'),
                      desc=self.recode(desc).strip(),
                      parents=self.parents[rev])

    def pull(self, rev):
        output, status = self.run('pull', self.path, all=True,
                                  match='hash %s' % rev,
                                  no_test=True, no_posthook=True,
                                  external_merge='/bin/false',
                                  repodir=self.tmppath)
        if status:
            if output.find('We have conflicts in') == -1:
                self.checkexit(status, output)
            output, status = self.run('revert', all=True, repodir=self.tmppath)
            self.checkexit(status, output)

    def getchanges(self, rev, full):
        if full:
            raise util.Abort(_("convert from darcs do not support --full"))
        copies = {}
        changes = []
        man = None
        for elt in self.changes[rev].find('summary').getchildren():
            if elt.tag in ('add_directory', 'remove_directory'):
                continue
            if elt.tag == 'move':
                if man is None:
                    man = self.manifest()
                source, dest = elt.get('from'), elt.get('to')
                if source in man:
                    # File move
                    changes.append((source, rev))
                    changes.append((dest, rev))
                    copies[dest] = source
                else:
                    # Directory move, deduce file moves from manifest
                    source = source + '/'
                    for f in man:
                        if not f.startswith(source):
                            continue
                        fdest = dest + '/' + f[len(source):]
                        changes.append((f, rev))
                        changes.append((fdest, rev))
                        copies[fdest] = f
            else:
                changes.append((elt.text.strip(), rev))
        self.pull(rev)
        self.lastrev = rev
        return sorted(changes), copies, set()

    def getfile(self, name, rev):
        if rev != self.lastrev:
            raise util.Abort(_('internal calling inconsistency'))
        path = os.path.join(self.tmppath, name)
        try:
            data = util.readfile(path)
            mode = os.lstat(path).st_mode
        except IOError as inst:
            if inst.errno == errno.ENOENT:
                return None, None
            raise
        mode = (mode & 0o111) and 'x' or ''
        return data, mode

    def gettags(self):
        return self.tags