Mercurial > hg
view hgext/journal.py @ 48068:bf8837e3d7ce
dirstate: Remove the flat Rust DirstateMap implementation
Before this changeset we had two Rust implementations of `DirstateMap`.
This removes the "flat" DirstateMap so that the "tree" DirstateMap is always
used when Rust enabled. This simplifies the code a lot, and will enable
(in the next changeset) further removal of a trait abstraction.
This is a performance regression when:
* Rust is enabled, and
* The repository uses the legacy dirstate-v1 file format, and
* For `hg status`, unknown files are not listed (such as with `-mard`)
The regression is about 100 milliseconds for `hg status -mard` on a
semi-large repository (mozilla-central), from ~320ms to ~420ms.
We deem this to be small enough to be worth it.
The new dirstate-v2 is still experimental at this point, but we aim to
stabilize it (though not yet enable it by default for new repositories)
in Mercurial 6.0. Eventually, upgrating repositories to dirsate-v2 will
eliminate this regression (and enable other performance improvements).
# Background
The flat DirstateMap was introduced with the first Rust implementation of the
status algorithm. It works similarly to the previous Python + C one, with a
single `HashMap` that associates file paths to a `DirstateEntry` (where Python
has a dict).
We later added the tree DirstateMap where the root of the tree contains nodes
for files and directories that are directly at the root of the repository,
and nodes for directories can contain child nodes representing the files and
directly that *they* contain directly. The shape of this tree mirrors that of
the working directory in the filesystem. This enables the status algorithm to
traverse this tree in tandem with traversing the filesystem tree, which in
turns enables a more efficient algorithm.
Furthermore, the new dirstate-v2 file format is also based on a tree of the
same shape. The tree DirstateMap can access a dirstate-v2 file without parsing
it: binary data in a single large (possibly memory-mapped) bytes buffer is
traversed on demand. This allows `DirstateMap` creation to take `O(1)` time.
(Mutation works by creating new in-memory nodes with copy-on-write semantics,
and serialization is append-mostly.)
The tradeoff is that for "legacy" repositories that use the dirstate-v1 file
format, parsing that file into a tree DirstateMap takes more time. Profiling
shows that this time is dominated by `HashMap`. For a dirstate containing `F`
files with an average `D` directory depth, the flat DirstateMap does parsing
in `O(F)` number of HashMap operations but the tree DirstateMap in `O(F × D)`
operations, since each node has its own HashMap containing its child nodes.
This slower costs ~140ms on an old snapshot of mozilla-central, and ~80ms
on an old snapshot of the Netbeans repository.
The status algorithm is faster, but with `-mard` (when not listing unknown
files) it is typically not faster *enough* to compensate the slower parsing.
Both Rust implementations are always faster than the Python + C implementation
# Benchmark results
All benchmarks are run on changeset 98c0408324e6, with repositories that use
the dirstate-v1 file format, on a server with 4 CPU cores and 4 CPU threads
(no HyperThreading).
`hg status` benchmarks show wall clock times of the entire command as the
average and standard deviation of serveral runs, collected by
https://github.com/sharkdp/hyperfine and reformated.
Parsing benchmarks are wall clock time of the Rust function that converts a
bytes buffer of the dirstate file into the `DirstateMap` data structure as
used by the status algorithm. A single run each, collected by running
`hg status` this environment variable:
RUST_LOG=hg::dirstate::dirstate_map=trace,hg::dirstate_tree::dirstate_map=trace
Benchmark 1: Rust flat DirstateMap → Rust tree DirstateMap
hg status
mozilla-clean 562.3 ms ± 2.0 ms → 462.5 ms ± 0.6 ms 1.22 ± 0.00 times faster
mozilla-dirty 859.6 ms ± 2.2 ms → 719.5 ms ± 3.2 ms 1.19 ± 0.01 times faster
mozilla-ignored 558.2 ms ± 3.0 ms → 457.9 ms ± 2.9 ms 1.22 ± 0.01 times faster
mozilla-unknowns 859.4 ms ± 5.7 ms → 716.0 ms ± 4.7 ms 1.20 ± 0.01 times faster
netbeans-clean 336.5 ms ± 0.9 ms → 339.5 ms ± 0.4 ms 0.99 ± 0.00 times faster
netbeans-dirty 491.4 ms ± 1.6 ms → 475.1 ms ± 1.2 ms 1.03 ± 0.00 times faster
netbeans-ignored 343.7 ms ± 1.0 ms → 347.8 ms ± 0.4 ms 0.99 ± 0.00 times faster
netbeans-unknowns 484.3 ms ± 1.0 ms → 466.0 ms ± 1.2 ms 1.04 ± 0.00 times faster
hg status -mard
mozilla-clean 317.3 ms ± 0.6 ms → 422.5 ms ± 1.2 ms 0.75 ± 0.00 times faster
mozilla-dirty 315.4 ms ± 0.6 ms → 417.7 ms ± 1.1 ms 0.76 ± 0.00 times faster
mozilla-ignored 314.6 ms ± 0.6 ms → 417.4 ms ± 1.0 ms 0.75 ± 0.00 times faster
mozilla-unknowns 312.9 ms ± 0.9 ms → 417.3 ms ± 1.6 ms 0.75 ± 0.00 times faster
netbeans-clean 212.0 ms ± 0.6 ms → 283.6 ms ± 0.8 ms 0.75 ± 0.00 times faster
netbeans-dirty 211.4 ms ± 1.0 ms → 283.4 ms ± 1.6 ms 0.75 ± 0.01 times faster
netbeans-ignored 211.4 ms ± 0.9 ms → 283.9 ms ± 0.8 ms 0.74 ± 0.01 times faster
netbeans-unknowns 211.1 ms ± 0.6 ms → 283.4 ms ± 1.0 ms 0.74 ± 0.00 times faster
Parsing
mozilla-clean 38.4ms → 177.6ms
mozilla-dirty 38.8ms → 177.0ms
mozilla-ignored 38.8ms → 178.0ms
mozilla-unknowns 38.7ms → 176.9ms
netbeans-clean 16.5ms → 97.3ms
netbeans-dirty 16.5ms → 98.4ms
netbeans-ignored 16.9ms → 97.4ms
netbeans-unknowns 16.9ms → 96.3ms
Benchmark 2: Python + C dirstatemap → Rust tree DirstateMap
hg status
mozilla-clean 1261.0 ms ± 3.6 ms → 461.1 ms ± 0.5 ms 2.73 ± 0.00 times faster
mozilla-dirty 2293.4 ms ± 9.1 ms → 719.6 ms ± 3.6 ms 3.19 ± 0.01 times faster
mozilla-ignored 1240.4 ms ± 2.3 ms → 457.7 ms ± 1.9 ms 2.71 ± 0.00 times faster
mozilla-unknowns 2283.3 ms ± 9.0 ms → 719.7 ms ± 3.8 ms 3.17 ± 0.01 times faster
netbeans-clean 879.7 ms ± 3.5 ms → 339.9 ms ± 0.5 ms 2.59 ± 0.00 times faster
netbeans-dirty 1257.3 ms ± 4.7 ms → 474.6 ms ± 1.6 ms 2.65 ± 0.01 times faster
netbeans-ignored 943.9 ms ± 1.9 ms → 347.3 ms ± 1.1 ms 2.72 ± 0.00 times faster
netbeans-unknowns 1188.1 ms ± 5.0 ms → 465.2 ms ± 2.3 ms 2.55 ± 0.01 times faster
hg status -mard
mozilla-clean 903.2 ms ± 3.6 ms → 423.4 ms ± 2.2 ms 2.13 ± 0.01 times faster
mozilla-dirty 884.6 ms ± 4.5 ms → 417.3 ms ± 1.4 ms 2.12 ± 0.01 times faster
mozilla-ignored 881.9 ms ± 1.3 ms → 417.3 ms ± 0.8 ms 2.11 ± 0.00 times faster
mozilla-unknowns 878.5 ms ± 1.9 ms → 416.4 ms ± 0.9 ms 2.11 ± 0.00 times faster
netbeans-clean 434.9 ms ± 1.8 ms → 284.0 ms ± 0.8 ms 1.53 ± 0.01 times faster
netbeans-dirty 434.1 ms ± 0.8 ms → 283.1 ms ± 0.8 ms 1.53 ± 0.00 times faster
netbeans-ignored 431.7 ms ± 1.1 ms → 283.6 ms ± 1.8 ms 1.52 ± 0.01 times faster
netbeans-unknowns 433.0 ms ± 1.3 ms → 283.5 ms ± 0.7 ms 1.53 ± 0.00 times faster
Differential Revision: https://phab.mercurial-scm.org/D11516
author | Simon Sapin <simon.sapin@octobus.net> |
---|---|
date | Mon, 27 Sep 2021 12:09:15 +0200 |
parents | d55b71393907 |
children | 6000f5b25c9b |
line wrap: on
line source
# journal.py # # Copyright 2014-2016 Facebook, Inc. # # This software may be used and distributed according to the terms of the # GNU General Public License version 2 or any later version. """track previous positions of bookmarks (EXPERIMENTAL) This extension adds a new command: `hg journal`, which shows you where bookmarks were previously located. """ from __future__ import absolute_import import collections import errno import os import weakref from mercurial.i18n import _ from mercurial.node import ( bin, hex, ) from mercurial import ( bookmarks, cmdutil, dispatch, encoding, error, extensions, hg, localrepo, lock, logcmdutil, pycompat, registrar, util, ) from mercurial.utils import ( dateutil, procutil, stringutil, ) cmdtable = {} command = registrar.command(cmdtable) # Note for extension authors: ONLY specify testedwith = 'ships-with-hg-core' for # extensions which SHIP WITH MERCURIAL. Non-mainline extensions should # be specifying the version(s) of Mercurial they are tested with, or # leave the attribute unspecified. testedwith = b'ships-with-hg-core' # storage format version; increment when the format changes storageversion = 0 # namespaces bookmarktype = b'bookmark' wdirparenttype = b'wdirparent' # In a shared repository, what shared feature name is used # to indicate this namespace is shared with the source? sharednamespaces = { bookmarktype: hg.sharedbookmarks, } # Journal recording, register hooks and storage object def extsetup(ui): extensions.wrapfunction(dispatch, b'runcommand', runcommand) extensions.wrapfunction(bookmarks.bmstore, b'_write', recordbookmarks) extensions.wrapfilecache( localrepo.localrepository, b'dirstate', wrapdirstate ) extensions.wrapfunction(hg, b'postshare', wrappostshare) extensions.wrapfunction(hg, b'copystore', unsharejournal) def reposetup(ui, repo): if repo.local(): repo.journal = journalstorage(repo) repo._wlockfreeprefix.add(b'namejournal') dirstate, cached = localrepo.isfilecached(repo, b'dirstate') if cached: # already instantiated dirstate isn't yet marked as # "journal"-ing, even though repo.dirstate() was already # wrapped by own wrapdirstate() _setupdirstate(repo, dirstate) def runcommand(orig, lui, repo, cmd, fullargs, *args): """Track the command line options for recording in the journal""" journalstorage.recordcommand(*fullargs) return orig(lui, repo, cmd, fullargs, *args) def _setupdirstate(repo, dirstate): dirstate.journalstorage = repo.journal dirstate.addparentchangecallback(b'journal', recorddirstateparents) # hooks to record dirstate changes def wrapdirstate(orig, repo): """Make journal storage available to the dirstate object""" dirstate = orig(repo) if util.safehasattr(repo, 'journal'): _setupdirstate(repo, dirstate) return dirstate def recorddirstateparents(dirstate, old, new): """Records all dirstate parent changes in the journal.""" old = list(old) new = list(new) if util.safehasattr(dirstate, 'journalstorage'): # only record two hashes if there was a merge oldhashes = old[:1] if old[1] == dirstate._nodeconstants.nullid else old newhashes = new[:1] if new[1] == dirstate._nodeconstants.nullid else new dirstate.journalstorage.record( wdirparenttype, b'.', oldhashes, newhashes ) # hooks to record bookmark changes (both local and remote) def recordbookmarks(orig, store, fp): """Records all bookmark changes in the journal.""" repo = store._repo if util.safehasattr(repo, 'journal'): oldmarks = bookmarks.bmstore(repo) for mark, value in pycompat.iteritems(store): oldvalue = oldmarks.get(mark, repo.nullid) if value != oldvalue: repo.journal.record(bookmarktype, mark, oldvalue, value) return orig(store, fp) # shared repository support def _readsharedfeatures(repo): """A set of shared features for this repository""" try: return set(repo.vfs.read(b'shared').splitlines()) except IOError as inst: if inst.errno != errno.ENOENT: raise return set() def _mergeentriesiter(*iterables, **kwargs): """Given a set of sorted iterables, yield the next entry in merged order Note that by default entries go from most recent to oldest. """ order = kwargs.pop('order', max) iterables = [iter(it) for it in iterables] # this tracks still active iterables; iterables are deleted as they are # exhausted, which is why this is a dictionary and why each entry also # stores the key. Entries are mutable so we can store the next value each # time. iterable_map = {} for key, it in enumerate(iterables): try: iterable_map[key] = [next(it), key, it] except StopIteration: # empty entry, can be ignored pass while iterable_map: value, key, it = order(pycompat.itervalues(iterable_map)) yield value try: iterable_map[key][0] = next(it) except StopIteration: # this iterable is empty, remove it from consideration del iterable_map[key] def wrappostshare(orig, sourcerepo, destrepo, **kwargs): """Mark this shared working copy as sharing journal information""" with destrepo.wlock(): orig(sourcerepo, destrepo, **kwargs) with destrepo.vfs(b'shared', b'a') as fp: fp.write(b'journal\n') def unsharejournal(orig, ui, repo, repopath): """Copy shared journal entries into this repo when unsharing""" if ( repo.path == repopath and repo.shared() and util.safehasattr(repo, 'journal') ): sharedrepo = hg.sharedreposource(repo) sharedfeatures = _readsharedfeatures(repo) if sharedrepo and sharedfeatures > {b'journal'}: # there is a shared repository and there are shared journal entries # to copy. move shared date over from source to destination but # move the local file first if repo.vfs.exists(b'namejournal'): journalpath = repo.vfs.join(b'namejournal') util.rename(journalpath, journalpath + b'.bak') storage = repo.journal local = storage._open( repo.vfs, filename=b'namejournal.bak', _newestfirst=False ) shared = ( e for e in storage._open(sharedrepo.vfs, _newestfirst=False) if sharednamespaces.get(e.namespace) in sharedfeatures ) for entry in _mergeentriesiter(local, shared, order=min): storage._write(repo.vfs, entry) return orig(ui, repo, repopath) class journalentry( collections.namedtuple( 'journalentry', 'timestamp user command namespace name oldhashes newhashes', ) ): """Individual journal entry * timestamp: a mercurial (time, timezone) tuple * user: the username that ran the command * namespace: the entry namespace, an opaque string * name: the name of the changed item, opaque string with meaning in the namespace * command: the hg command that triggered this record * oldhashes: a tuple of one or more binary hashes for the old location * newhashes: a tuple of one or more binary hashes for the new location Handles serialisation from and to the storage format. Fields are separated by newlines, hashes are written out in hex separated by commas, timestamp and timezone are separated by a space. """ @classmethod def fromstorage(cls, line): ( time, user, command, namespace, name, oldhashes, newhashes, ) = line.split(b'\n') timestamp, tz = time.split() timestamp, tz = float(timestamp), int(tz) oldhashes = tuple(bin(hash) for hash in oldhashes.split(b',')) newhashes = tuple(bin(hash) for hash in newhashes.split(b',')) return cls( (timestamp, tz), user, command, namespace, name, oldhashes, newhashes, ) def __bytes__(self): """bytes representation for storage""" time = b' '.join(map(pycompat.bytestr, self.timestamp)) oldhashes = b','.join([hex(hash) for hash in self.oldhashes]) newhashes = b','.join([hex(hash) for hash in self.newhashes]) return b'\n'.join( ( time, self.user, self.command, self.namespace, self.name, oldhashes, newhashes, ) ) __str__ = encoding.strmethod(__bytes__) class journalstorage(object): """Storage for journal entries Entries are divided over two files; one with entries that pertain to the local working copy *only*, and one with entries that are shared across multiple working copies when shared using the share extension. Entries are stored with NUL bytes as separators. See the journalentry class for the per-entry structure. The file format starts with an integer version, delimited by a NUL. This storage uses a dedicated lock; this makes it easier to avoid issues with adding entries that added when the regular wlock is unlocked (e.g. the dirstate). """ _currentcommand = () _lockref = None def __init__(self, repo): self.user = procutil.getuser() self.ui = repo.ui self.vfs = repo.vfs # is this working copy using a shared storage? self.sharedfeatures = self.sharedvfs = None if repo.shared(): features = _readsharedfeatures(repo) sharedrepo = hg.sharedreposource(repo) if sharedrepo is not None and b'journal' in features: self.sharedvfs = sharedrepo.vfs self.sharedfeatures = features # track the current command for recording in journal entries @property def command(self): commandstr = b' '.join( map(procutil.shellquote, journalstorage._currentcommand) ) if b'\n' in commandstr: # truncate multi-line commands commandstr = commandstr.partition(b'\n')[0] + b' ...' return commandstr @classmethod def recordcommand(cls, *fullargs): """Set the current hg arguments, stored with recorded entries""" # Set the current command on the class because we may have started # with a non-local repo (cloning for example). cls._currentcommand = fullargs def _currentlock(self, lockref): """Returns the lock if it's held, or None if it's not. (This is copied from the localrepo class) """ if lockref is None: return None l = lockref() if l is None or not l.held: return None return l def jlock(self, vfs): """Create a lock for the journal file""" if self._currentlock(self._lockref) is not None: raise error.Abort(_(b'journal lock does not support nesting')) desc = _(b'journal of %s') % vfs.base try: l = lock.lock(vfs, b'namejournal.lock', 0, desc=desc) except error.LockHeld as inst: self.ui.warn( _(b"waiting for lock on %s held by %r\n") % (desc, inst.locker) ) # default to 600 seconds timeout l = lock.lock( vfs, b'namejournal.lock', self.ui.configint(b"ui", b"timeout"), desc=desc, ) self.ui.warn(_(b"got lock after %s seconds\n") % l.delay) self._lockref = weakref.ref(l) return l def record(self, namespace, name, oldhashes, newhashes): """Record a new journal entry * namespace: an opaque string; this can be used to filter on the type of recorded entries. * name: the name defining this entry; for bookmarks, this is the bookmark name. Can be filtered on when retrieving entries. * oldhashes and newhashes: each a single binary hash, or a list of binary hashes. These represent the old and new position of the named item. """ if not isinstance(oldhashes, list): oldhashes = [oldhashes] if not isinstance(newhashes, list): newhashes = [newhashes] entry = journalentry( dateutil.makedate(), self.user, self.command, namespace, name, oldhashes, newhashes, ) vfs = self.vfs if self.sharedvfs is not None: # write to the shared repository if this feature is being # shared between working copies. if sharednamespaces.get(namespace) in self.sharedfeatures: vfs = self.sharedvfs self._write(vfs, entry) def _write(self, vfs, entry): with self.jlock(vfs): # open file in amend mode to ensure it is created if missing with vfs(b'namejournal', mode=b'a+b') as f: f.seek(0, os.SEEK_SET) # Read just enough bytes to get a version number (up to 2 # digits plus separator) version = f.read(3).partition(b'\0')[0] if version and version != b"%d" % storageversion: # different version of the storage. Exit early (and not # write anything) if this is not a version we can handle or # the file is corrupt. In future, perhaps rotate the file # instead? self.ui.warn( _(b"unsupported journal file version '%s'\n") % version ) return if not version: # empty file, write version first f.write((b"%d" % storageversion) + b'\0') f.seek(0, os.SEEK_END) f.write(bytes(entry) + b'\0') def filtered(self, namespace=None, name=None): """Yield all journal entries with the given namespace or name Both the namespace and the name are optional; if neither is given all entries in the journal are produced. Matching supports regular expressions by using the `re:` prefix (use `literal:` to match names or namespaces that start with `re:`) """ if namespace is not None: namespace = stringutil.stringmatcher(namespace)[-1] if name is not None: name = stringutil.stringmatcher(name)[-1] for entry in self: if namespace is not None and not namespace(entry.namespace): continue if name is not None and not name(entry.name): continue yield entry def __iter__(self): """Iterate over the storage Yields journalentry instances for each contained journal record. """ local = self._open(self.vfs) if self.sharedvfs is None: return local # iterate over both local and shared entries, but only those # shared entries that are among the currently shared features shared = ( e for e in self._open(self.sharedvfs) if sharednamespaces.get(e.namespace) in self.sharedfeatures ) return _mergeentriesiter(local, shared) def _open(self, vfs, filename=b'namejournal', _newestfirst=True): if not vfs.exists(filename): return with vfs(filename) as f: raw = f.read() lines = raw.split(b'\0') version = lines and lines[0] if version != b"%d" % storageversion: version = version or _(b'not available') raise error.Abort(_(b"unknown journal file version '%s'") % version) # Skip the first line, it's a version number. Normally we iterate over # these in reverse order to list newest first; only when copying across # a shared storage do we forgo reversing. lines = lines[1:] if _newestfirst: lines = reversed(lines) for line in lines: if not line: continue yield journalentry.fromstorage(line) # journal reading # log options that don't make sense for journal _ignoreopts = (b'no-merges', b'graph') @command( b'journal', [ (b'', b'all', None, b'show history for all names'), (b'c', b'commits', None, b'show commit metadata'), ] + [opt for opt in cmdutil.logopts if opt[1] not in _ignoreopts], b'[OPTION]... [BOOKMARKNAME]', helpcategory=command.CATEGORY_CHANGE_ORGANIZATION, ) def journal(ui, repo, *args, **opts): """show the previous position of bookmarks and the working copy The journal is used to see the previous commits that bookmarks and the working copy pointed to. By default the previous locations for the working copy. Passing a bookmark name will show all the previous positions of that bookmark. Use the --all switch to show previous locations for all bookmarks and the working copy; each line will then include the bookmark name, or '.' for the working copy, as well. If `name` starts with `re:`, the remainder of the name is treated as a regular expression. To match a name that actually starts with `re:`, use the prefix `literal:`. By default hg journal only shows the commit hash and the command that was running at that time. -v/--verbose will show the prior hash, the user, and the time at which it happened. Use -c/--commits to output log information on each commit hash; at this point you can use the usual `--patch`, `--git`, `--stat` and `--template` switches to alter the log output for these. `hg journal -T json` can be used to produce machine readable output. """ opts = pycompat.byteskwargs(opts) name = b'.' if opts.get(b'all'): if args: raise error.Abort( _(b"You can't combine --all and filtering on a name") ) name = None if args: name = args[0] fm = ui.formatter(b'journal', opts) def formatnodes(nodes): return fm.formatlist(map(fm.hexfunc, nodes), name=b'node', sep=b',') if opts.get(b"template") != b"json": if name is None: displayname = _(b'the working copy and bookmarks') else: displayname = b"'%s'" % name ui.status(_(b"previous locations of %s:\n") % displayname) limit = logcmdutil.getlimit(opts) entry = None ui.pager(b'journal') for count, entry in enumerate(repo.journal.filtered(name=name)): if count == limit: break fm.startitem() fm.condwrite( ui.verbose, b'oldnodes', b'%s -> ', formatnodes(entry.oldhashes) ) fm.write(b'newnodes', b'%s', formatnodes(entry.newhashes)) fm.condwrite(ui.verbose, b'user', b' %-8s', entry.user) fm.condwrite( opts.get(b'all') or name.startswith(b're:'), b'name', b' %-8s', entry.name, ) fm.condwrite( ui.verbose, b'date', b' %s', fm.formatdate(entry.timestamp, b'%Y-%m-%d %H:%M %1%2'), ) fm.write(b'command', b' %s\n', entry.command) if opts.get(b"commits"): if fm.isplain(): displayer = logcmdutil.changesetdisplayer(ui, repo, opts) else: displayer = logcmdutil.changesetformatter( ui, repo, fm.nested(b'changesets'), diffopts=opts ) for hash in entry.newhashes: try: ctx = repo[hash] displayer.show(ctx) except error.RepoLookupError as e: fm.plain(b"%s\n\n" % pycompat.bytestr(e)) displayer.close() fm.end() if entry is None: ui.status(_(b"no recorded locations\n"))