Mercurial > public > mercurial-scm > hg
view mercurial/subrepoutil.py @ 44118:f81c17ec303c
hgdemandimport: apply lazy module loading to sys.meta_path finders
Python's `sys.meta_path` finders are the primary objects whose job it
is to find a module at import time. When `import` is called, Python
iterates objects in this list and calls `o.find_spec(...)` to find
a `ModuleSpec` (or None if the module couldn't be found by that
finder). If no meta path finder can find a module, import fails.
One of the default meta path finders is `PathFinder`. Its job is to
import modules from the filesystem and is probably the most important
importer. This finder looks at `sys.path` and `sys.path_hooks` to do
its job.
The `ModuleSpec` returned by `MetaPathImporter.find_spec()` has a
`loader` attribute, which defines the concrete module loader to use.
`sys.path_hooks` is a hook point for teaching `PathFinder` to
instantiate custom loader types.
Previously, we injected a custom `sys.path_hook` that told `PathFinder`
to wrap the default loaders with a loader that creates a module object
that is lazy.
This approach worked. But its main limitation was that it only applied
to the `PathFinder` meta path importer. There are other meta path
importers that are registered. And in the case of PyOxidizer loading
modules from memory, `PathFinder` doesn't come into play since
PyOxidizer's own meta path importer was handling all imports.
This commit changes our approach to lazy module loading by proxying
all meta path importers. Specifically, we overload the `find_spec()`
method to swap in a wrapped loader on the `ModuleSpec` before it
is returned. The end result of this is all meta path importers should
be lazy.
As much as I would have loved to utilize .__class__ manipulation to
achieve this, some meta path importers are implemented in C/Rust
in such a way that they cannot be monkeypatched. This is why we
use __getattribute__ to define a proxy.
Also, this change could theoretically open us up to regressions in
meta path importers whose loader is creating module objects which
can't be monkeypatched. But I'm not aware of any of these in the
wild. So I think we'll be safe.
According to hyperfine, this change yields a decent startup time win of
5-6ms:
```
Benchmark #1: ~/.pyenv/versions/3.6.10/bin/python ./hg version
Time (mean ? ?): 86.8 ms ? 0.5 ms [User: 78.0 ms, System: 8.7 ms]
Range (min ? max): 86.0 ms ? 89.1 ms 50 runs
Time (mean ? ?): 81.1 ms ? 2.7 ms [User: 74.5 ms, System: 6.5 ms]
Range (min ? max): 77.8 ms ? 90.5 ms 50 runs
Benchmark #2: ~/.pyenv/versions/3.7.6/bin/python ./hg version
Time (mean ? ?): 78.9 ms ? 0.6 ms [User: 70.2 ms, System: 8.7 ms]
Range (min ? max): 78.1 ms ? 81.2 ms 50 runs
Time (mean ? ?): 73.4 ms ? 0.6 ms [User: 65.3 ms, System: 8.0 ms]
Range (min ? max): 72.4 ms ? 75.7 ms 50 runs
Benchmark #3: ~/.pyenv/versions/3.8.1/bin/python ./hg version
Time (mean ? ?): 78.1 ms ? 0.6 ms [User: 70.2 ms, System: 7.9 ms]
Range (min ? max): 77.4 ms ? 80.9 ms 50 runs
Time (mean ? ?): 72.1 ms ? 0.4 ms [User: 64.4 ms, System: 7.6 ms]
Range (min ? max): 71.4 ms ? 74.1 ms 50 runs
```
Differential Revision: https://phab.mercurial-scm.org/D7954
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Mon, 20 Jan 2020 23:51:25 -0800 |
parents | d783f945a701 |
children | f7f142d74df3 |
line wrap: on
line source
# subrepoutil.py - sub-repository operations and substate handling # # Copyright 2009-2010 Matt Mackall <mpm@selenic.com> # # 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 errno import os import posixpath import re from .i18n import _ from .pycompat import getattr from . import ( config, error, filemerge, pathutil, phases, pycompat, util, ) from .utils import stringutil nullstate = (b'', b'', b'empty') def state(ctx, ui): """return a state dict, mapping subrepo paths configured in .hgsub to tuple: (source from .hgsub, revision from .hgsubstate, kind (key in types dict)) """ p = config.config() repo = ctx.repo() def read(f, sections=None, remap=None): if f in ctx: try: data = ctx[f].data() except IOError as err: if err.errno != errno.ENOENT: raise # handle missing subrepo spec files as removed ui.warn( _(b"warning: subrepo spec file \'%s\' not found\n") % repo.pathto(f) ) return p.parse(f, data, sections, remap, read) else: raise error.Abort( _(b"subrepo spec file \'%s\' not found") % repo.pathto(f) ) if b'.hgsub' in ctx: read(b'.hgsub') for path, src in ui.configitems(b'subpaths'): p.set(b'subpaths', path, src, ui.configsource(b'subpaths', path)) rev = {} if b'.hgsubstate' in ctx: try: for i, l in enumerate(ctx[b'.hgsubstate'].data().splitlines()): l = l.lstrip() if not l: continue try: revision, path = l.split(b" ", 1) except ValueError: raise error.Abort( _( b"invalid subrepository revision " b"specifier in \'%s\' line %d" ) % (repo.pathto(b'.hgsubstate'), (i + 1)) ) rev[path] = revision except IOError as err: if err.errno != errno.ENOENT: raise def remap(src): for pattern, repl in p.items(b'subpaths'): # Turn r'C:\foo\bar' into r'C:\\foo\\bar' since re.sub # does a string decode. repl = stringutil.escapestr(repl) # However, we still want to allow back references to go # through unharmed, so we turn r'\\1' into r'\1'. Again, # extra escapes are needed because re.sub string decodes. repl = re.sub(br'\\\\([0-9]+)', br'\\\1', repl) try: src = re.sub(pattern, repl, src, 1) except re.error as e: raise error.Abort( _(b"bad subrepository pattern in %s: %s") % ( p.source(b'subpaths', pattern), stringutil.forcebytestr(e), ) ) return src state = {} for path, src in p[b''].items(): kind = b'hg' if src.startswith(b'['): if b']' not in src: raise error.Abort(_(b'missing ] in subrepository source')) kind, src = src.split(b']', 1) kind = kind[1:] src = src.lstrip() # strip any extra whitespace after ']' if not util.url(src).isabs(): parent = _abssource(repo, abort=False) if parent: parent = util.url(parent) parent.path = posixpath.join(parent.path or b'', src) parent.path = posixpath.normpath(parent.path) joined = bytes(parent) # Remap the full joined path and use it if it changes, # else remap the original source. remapped = remap(joined) if remapped == joined: src = remap(src) else: src = remapped src = remap(src) state[util.pconvert(path)] = (src.strip(), rev.get(path, b''), kind) return state def writestate(repo, state): """rewrite .hgsubstate in (outer) repo with these subrepo states""" lines = [ b'%s %s\n' % (state[s][1], s) for s in sorted(state) if state[s][1] != nullstate[1] ] repo.wwrite(b'.hgsubstate', b''.join(lines), b'') def submerge(repo, wctx, mctx, actx, overwrite, labels=None): """delegated from merge.applyupdates: merging of .hgsubstate file in working context, merging context and ancestor context""" if mctx == actx: # backwards? actx = wctx.p1() s1 = wctx.substate s2 = mctx.substate sa = actx.substate sm = {} repo.ui.debug(b"subrepo merge %s %s %s\n" % (wctx, mctx, actx)) def debug(s, msg, r=b""): if r: r = b"%s:%s:%s" % r repo.ui.debug(b" subrepo %s: %s %s\n" % (s, msg, r)) promptssrc = filemerge.partextras(labels) for s, l in sorted(pycompat.iteritems(s1)): a = sa.get(s, nullstate) ld = l # local state with possible dirty flag for compares if wctx.sub(s).dirty(): ld = (l[0], l[1] + b"+") if wctx == actx: # overwrite a = ld prompts = promptssrc.copy() prompts[b's'] = s if s in s2: r = s2[s] if ld == r or r == a: # no change or local is newer sm[s] = l continue elif ld == a: # other side changed debug(s, b"other changed, get", r) wctx.sub(s).get(r, overwrite) sm[s] = r elif ld[0] != r[0]: # sources differ prompts[b'lo'] = l[0] prompts[b'ro'] = r[0] if repo.ui.promptchoice( _( b' subrepository sources for %(s)s differ\n' b'you can use (l)ocal%(l)s source (%(lo)s)' b' or (r)emote%(o)s source (%(ro)s).\n' b'what do you want to do?' b'$$ &Local $$ &Remote' ) % prompts, 0, ): debug(s, b"prompt changed, get", r) wctx.sub(s).get(r, overwrite) sm[s] = r elif ld[1] == a[1]: # local side is unchanged debug(s, b"other side changed, get", r) wctx.sub(s).get(r, overwrite) sm[s] = r else: debug(s, b"both sides changed") srepo = wctx.sub(s) prompts[b'sl'] = srepo.shortid(l[1]) prompts[b'sr'] = srepo.shortid(r[1]) option = repo.ui.promptchoice( _( b' subrepository %(s)s diverged (local revision: %(sl)s, ' b'remote revision: %(sr)s)\n' b'you can (m)erge, keep (l)ocal%(l)s or keep ' b'(r)emote%(o)s.\n' b'what do you want to do?' b'$$ &Merge $$ &Local $$ &Remote' ) % prompts, 0, ) if option == 0: wctx.sub(s).merge(r) sm[s] = l debug(s, b"merge with", r) elif option == 1: sm[s] = l debug(s, b"keep local subrepo revision", l) else: wctx.sub(s).get(r, overwrite) sm[s] = r debug(s, b"get remote subrepo revision", r) elif ld == a: # remote removed, local unchanged debug(s, b"remote removed, remove") wctx.sub(s).remove() elif a == nullstate: # not present in remote or ancestor debug(s, b"local added, keep") sm[s] = l continue else: if repo.ui.promptchoice( _( b' local%(l)s changed subrepository %(s)s' b' which remote%(o)s removed\n' b'use (c)hanged version or (d)elete?' b'$$ &Changed $$ &Delete' ) % prompts, 0, ): debug(s, b"prompt remove") wctx.sub(s).remove() for s, r in sorted(s2.items()): if s in s1: continue elif s not in sa: debug(s, b"remote added, get", r) mctx.sub(s).get(r) sm[s] = r elif r != sa[s]: prompts = promptssrc.copy() prompts[b's'] = s if ( repo.ui.promptchoice( _( b' remote%(o)s changed subrepository %(s)s' b' which local%(l)s removed\n' b'use (c)hanged version or (d)elete?' b'$$ &Changed $$ &Delete' ) % prompts, 0, ) == 0 ): debug(s, b"prompt recreate", r) mctx.sub(s).get(r) sm[s] = r # record merged .hgsubstate writestate(repo, sm) return sm def precommit(ui, wctx, status, match, force=False): """Calculate .hgsubstate changes that should be applied before committing Returns (subs, commitsubs, newstate) where - subs: changed subrepos (including dirty ones) - commitsubs: dirty subrepos which the caller needs to commit recursively - newstate: new state dict which the caller must write to .hgsubstate This also updates the given status argument. """ subs = [] commitsubs = set() newstate = wctx.substate.copy() # only manage subrepos and .hgsubstate if .hgsub is present if b'.hgsub' in wctx: # we'll decide whether to track this ourselves, thanks for c in status.modified, status.added, status.removed: if b'.hgsubstate' in c: c.remove(b'.hgsubstate') # compare current state to last committed state # build new substate based on last committed state oldstate = wctx.p1().substate for s in sorted(newstate.keys()): if not match(s): # ignore working copy, use old state if present if s in oldstate: newstate[s] = oldstate[s] continue if not force: raise error.Abort( _(b"commit with new subrepo %s excluded") % s ) dirtyreason = wctx.sub(s).dirtyreason(True) if dirtyreason: if not ui.configbool(b'ui', b'commitsubrepos'): raise error.Abort( dirtyreason, hint=_(b"use --subrepos for recursive commit"), ) subs.append(s) commitsubs.add(s) else: bs = wctx.sub(s).basestate() newstate[s] = (newstate[s][0], bs, newstate[s][2]) if oldstate.get(s, (None, None, None))[1] != bs: subs.append(s) # check for removed subrepos for p in wctx.parents(): r = [s for s in p.substate if s not in newstate] subs += [s for s in r if match(s)] if subs: if not match(b'.hgsub') and b'.hgsub' in ( wctx.modified() + wctx.added() ): raise error.Abort(_(b"can't commit subrepos without .hgsub")) status.modified.insert(0, b'.hgsubstate') elif b'.hgsub' in status.removed: # clean up .hgsubstate when .hgsub is removed if b'.hgsubstate' in wctx and b'.hgsubstate' not in ( status.modified + status.added + status.removed ): status.removed.insert(0, b'.hgsubstate') return subs, commitsubs, newstate def reporelpath(repo): """return path to this (sub)repo as seen from outermost repo""" parent = repo while util.safehasattr(parent, b'_subparent'): parent = parent._subparent return repo.root[len(pathutil.normasprefix(parent.root)) :] def subrelpath(sub): """return path to this subrepo as seen from outermost repo""" return sub._relpath def _abssource(repo, push=False, abort=True): """return pull/push path of repo - either based on parent repo .hgsub info or on the top repo config. Abort or return None if no source found.""" if util.safehasattr(repo, b'_subparent'): source = util.url(repo._subsource) if source.isabs(): return bytes(source) source.path = posixpath.normpath(source.path) parent = _abssource(repo._subparent, push, abort=False) if parent: parent = util.url(util.pconvert(parent)) parent.path = posixpath.join(parent.path or b'', source.path) parent.path = posixpath.normpath(parent.path) return bytes(parent) else: # recursion reached top repo path = None if util.safehasattr(repo, b'_subtoppath'): path = repo._subtoppath elif push and repo.ui.config(b'paths', b'default-push'): path = repo.ui.config(b'paths', b'default-push') elif repo.ui.config(b'paths', b'default'): path = repo.ui.config(b'paths', b'default') elif repo.shared(): # chop off the .hg component to get the default path form. This has # already run through vfsmod.vfs(..., realpath=True), so it doesn't # have problems with 'C:' return os.path.dirname(repo.sharedpath) if path: # issue5770: 'C:\' and 'C:' are not equivalent paths. The former is # as expected: an absolute path to the root of the C: drive. The # latter is a relative path, and works like so: # # C:\>cd C:\some\path # C:\>D: # D:\>python -c "import os; print os.path.abspath('C:')" # C:\some\path # # D:\>python -c "import os; print os.path.abspath('C:relative')" # C:\some\path\relative if util.hasdriveletter(path): if len(path) == 2 or path[2:3] not in br'\/': path = os.path.abspath(path) return path if abort: raise error.Abort(_(b"default path for subrepository not found")) def newcommitphase(ui, ctx): commitphase = phases.newcommitphase(ui) substate = getattr(ctx, "substate", None) if not substate: return commitphase check = ui.config(b'phases', b'checksubrepos') if check not in (b'ignore', b'follow', b'abort'): raise error.Abort( _(b'invalid phases.checksubrepos configuration: %s') % check ) if check == b'ignore': return commitphase maxphase = phases.public maxsub = None for s in sorted(substate): sub = ctx.sub(s) subphase = sub.phase(substate[s][1]) if maxphase < subphase: maxphase = subphase maxsub = s if commitphase < maxphase: if check == b'abort': raise error.Abort( _( b"can't commit in %s phase" b" conflicting %s from subrepository %s" ) % ( phases.phasenames[commitphase], phases.phasenames[maxphase], maxsub, ) ) ui.warn( _( b"warning: changes are committed in" b" %s phase from subrepository %s\n" ) % (phases.phasenames[maxphase], maxsub) ) return maxphase return commitphase