##// END OF EJS Templates
demandimport: replace more references to _demandmod instances...
demandimport: replace more references to _demandmod instances _demandmod instances may be referenced by multiple importing modules. Before this patch, the _demandmod instance only maintained a reference to its first consumer when using the "from X import Y" syntax. This is because we only created a single _demandmod instance (attached to the parent X module). If multiple modules A and B performed "from X import Y", we'd produce a single _demandmod instance "demandmod" with the following references: X.Y = <demandmod> A.Y = <demandmod> B.Y = <demandmod> The locals from the first consumer (A) would be stored in <demandmod1>. When <demandmod1> was loaded, we'd look at the locals for the first consumer and replace the symbol, if necessary. This resulted in state: X.Y = <module> A.Y = <module> B.Y = <demandmod> B's reference to Y wasn't updated and was still using the proxy object because we just didn't record that B had a reference to <demandmod> that needed updating! With this patch, we add support for tracking which modules in addition to the initial importer have a reference to the _demandmod instance and we replace those references at module load time. In the case of posix.py, this fixes an issue where the "encoding" module was being proxied, resulting in hundreds of thousands of __getattribute__ lookups on the _demandmod instance during dirstate operations on mozilla-central, speeding up execution by many milliseconds. There are likely several other operation that benefit from this change as well. The new mechanism isn't perfect: references in locals (not globals) may likely linger. So, if there is an import inside a function and a symbol from that module is used in a hot loop, we could have unwanted overhead from proxying through _demandmod. Non-global imports are discouraged anyway. So hopefully this isn't a big deal in practice. We could potentially deploy a code checker that bans use of attribute lookups of function-level-imported modules inside loops. This deficiency in theory could be avoided by storing the set of globals and locals dicts to update in the _demandmod instance. However, I tried this and it didn't work. One reason is that some globals are _demandmod instances. We could work around this, but it's a bit more work. There also might be other module import foo at play. The solution as implemented is better than what we had and IMO is good enough for the time being. It's worth noting that this sub-optimal behavior was made worse by the introduction of absolute_import and its recommended "from . import X" syntax for importing modules from the "mercurial" package. If we ever wrote performance tests, measuring the amount of module imports and __getattribute__ proxy calls through _demandmod instances would be something I'd have it check.

File last commit:

r26421:4b0fc75f default
r26457:7e813050 default
Show More
hgweb_mod.py
441 lines | 16.5 KiB | text/x-python | PythonLexer
# hgweb/hgweb_mod.py - Web interface for a repository.
#
# Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net>
# Copyright 2005-2007 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.
import contextlib
import os
from mercurial import ui, hg, hook, error, encoding, templater, util, repoview
from mercurial.templatefilters import websub
from common import ErrorResponse, permhooks, caching
from common import HTTP_OK, HTTP_NOT_MODIFIED, HTTP_BAD_REQUEST
from common import HTTP_NOT_FOUND, HTTP_SERVER_ERROR
from request import wsgirequest
import webcommands, protocol, webutil
perms = {
'changegroup': 'pull',
'changegroupsubset': 'pull',
'getbundle': 'pull',
'stream_out': 'pull',
'listkeys': 'pull',
'unbundle': 'push',
'pushkey': 'push',
}
def makebreadcrumb(url, prefix=''):
'''Return a 'URL breadcrumb' list
A 'URL breadcrumb' is a list of URL-name pairs,
corresponding to each of the path items on a URL.
This can be used to create path navigation entries.
'''
if url.endswith('/'):
url = url[:-1]
if prefix:
url = '/' + prefix + url
relpath = url
if relpath.startswith('/'):
relpath = relpath[1:]
breadcrumb = []
urlel = url
pathitems = [''] + relpath.split('/')
for pathel in reversed(pathitems):
if not pathel or not urlel:
break
breadcrumb.append({'url': urlel, 'name': pathel})
urlel = os.path.dirname(urlel)
return reversed(breadcrumb)
class requestcontext(object):
"""Holds state/context for an individual request.
Servers can be multi-threaded. Holding state on the WSGI application
is prone to race conditions. Instances of this class exist to hold
mutable and race-free state for requests.
"""
def __init__(self, app, repo):
self.repo = repo
self.reponame = app.reponame
self.archives = ('zip', 'gz', 'bz2')
self.maxchanges = self.configint('web', 'maxchanges', 10)
self.stripecount = self.configint('web', 'stripes', 1)
self.maxshortchanges = self.configint('web', 'maxshortchanges', 60)
self.maxfiles = self.configint('web', 'maxfiles', 10)
self.allowpull = self.configbool('web', 'allowpull', True)
# we use untrusted=False to prevent a repo owner from using
# web.templates in .hg/hgrc to get access to any file readable
# by the user running the CGI script
self.templatepath = self.config('web', 'templates', untrusted=False)
# This object is more expensive to build than simple config values.
# It is shared across requests. The app will replace the object
# if it is updated. Since this is a reference and nothing should
# modify the underlying object, it should be constant for the lifetime
# of the request.
self.websubtable = app.websubtable
# Trust the settings from the .hg/hgrc files by default.
def config(self, section, name, default=None, untrusted=True):
return self.repo.ui.config(section, name, default,
untrusted=untrusted)
def configbool(self, section, name, default=False, untrusted=True):
return self.repo.ui.configbool(section, name, default,
untrusted=untrusted)
def configint(self, section, name, default=None, untrusted=True):
return self.repo.ui.configint(section, name, default,
untrusted=untrusted)
def configlist(self, section, name, default=None, untrusted=True):
return self.repo.ui.configlist(section, name, default,
untrusted=untrusted)
archivespecs = {
'bz2': ('application/x-bzip2', 'tbz2', '.tar.bz2', None),
'gz': ('application/x-gzip', 'tgz', '.tar.gz', None),
'zip': ('application/zip', 'zip', '.zip', None),
}
def archivelist(self, nodeid):
allowed = self.configlist('web', 'allow_archive')
for typ, spec in self.archivespecs.iteritems():
if typ in allowed or self.configbool('web', 'allow%s' % typ):
yield {'type': typ, 'extension': spec[2], 'node': nodeid}
def templater(self, req):
# determine scheme, port and server name
# this is needed to create absolute urls
proto = req.env.get('wsgi.url_scheme')
if proto == 'https':
proto = 'https'
default_port = '443'
else:
proto = 'http'
default_port = '80'
port = req.env['SERVER_PORT']
port = port != default_port and (':' + port) or ''
urlbase = '%s://%s%s' % (proto, req.env['SERVER_NAME'], port)
logourl = self.config('web', 'logourl', 'https://mercurial-scm.org/')
logoimg = self.config('web', 'logoimg', 'hglogo.png')
staticurl = self.config('web', 'staticurl') or req.url + 'static/'
if not staticurl.endswith('/'):
staticurl += '/'
# some functions for the templater
def motd(**map):
yield self.config('web', 'motd', '')
# figure out which style to use
vars = {}
styles = (
req.form.get('style', [None])[0],
self.config('web', 'style'),
'paper',
)
style, mapfile = templater.stylemap(styles, self.templatepath)
if style == styles[0]:
vars['style'] = style
start = req.url[-1] == '?' and '&' or '?'
sessionvars = webutil.sessionvars(vars, start)
if not self.reponame:
self.reponame = (self.config('web', 'name')
or req.env.get('REPO_NAME')
or req.url.strip('/') or self.repo.root)
def websubfilter(text):
return websub(text, self.websubtable)
# create the templater
tmpl = templater.templater(mapfile,
filters={'websub': websubfilter},
defaults={'url': req.url,
'logourl': logourl,
'logoimg': logoimg,
'staticurl': staticurl,
'urlbase': urlbase,
'repo': self.reponame,
'encoding': encoding.encoding,
'motd': motd,
'sessionvars': sessionvars,
'pathdef': makebreadcrumb(req.url),
'style': style,
})
return tmpl
class hgweb(object):
"""HTTP server for individual repositories.
Instances of this class serve HTTP responses for a particular
repository.
Instances are typically used as WSGI applications.
Some servers are multi-threaded. On these servers, there may
be multiple active threads inside __call__.
"""
def __init__(self, repo, name=None, baseui=None):
if isinstance(repo, str):
if baseui:
u = baseui.copy()
else:
u = ui.ui()
r = hg.repository(u, repo)
else:
# we trust caller to give us a private copy
r = repo
r.ui.setconfig('ui', 'report_untrusted', 'off', 'hgweb')
r.baseui.setconfig('ui', 'report_untrusted', 'off', 'hgweb')
r.ui.setconfig('ui', 'nontty', 'true', 'hgweb')
r.baseui.setconfig('ui', 'nontty', 'true', 'hgweb')
# resolve file patterns relative to repo root
r.ui.setconfig('ui', 'forcecwd', r.root, 'hgweb')
r.baseui.setconfig('ui', 'forcecwd', r.root, 'hgweb')
# displaying bundling progress bar while serving feel wrong and may
# break some wsgi implementation.
r.ui.setconfig('progress', 'disable', 'true', 'hgweb')
r.baseui.setconfig('progress', 'disable', 'true', 'hgweb')
self._repos = [hg.cachedlocalrepo(self._webifyrepo(r))]
self._lastrepo = self._repos[0]
hook.redirect(True)
self.reponame = name
def _webifyrepo(self, repo):
repo = getwebview(repo)
self.websubtable = webutil.getwebsubs(repo)
return repo
@contextlib.contextmanager
def _obtainrepo(self):
"""Obtain a repo unique to the caller.
Internally we maintain a stack of cachedlocalrepo instances
to be handed out. If one is available, we pop it and return it,
ensuring it is up to date in the process. If one is not available,
we clone the most recently used repo instance and return it.
It is currently possible for the stack to grow without bounds
if the server allows infinite threads. However, servers should
have a thread limit, thus establishing our limit.
"""
if self._repos:
cached = self._repos.pop()
r, created = cached.fetch()
else:
cached = self._lastrepo.copy()
r, created = cached.fetch()
if created:
r = self._webifyrepo(r)
self._lastrepo = cached
self.mtime = cached.mtime
try:
yield r
finally:
self._repos.append(cached)
def run(self):
"""Start a server from CGI environment.
Modern servers should be using WSGI and should avoid this
method, if possible.
"""
if not os.environ.get('GATEWAY_INTERFACE', '').startswith("CGI/1."):
raise RuntimeError("This function is only intended to be "
"called while running as a CGI script.")
import mercurial.hgweb.wsgicgi as wsgicgi
wsgicgi.launch(self)
def __call__(self, env, respond):
"""Run the WSGI application.
This may be called by multiple threads.
"""
req = wsgirequest(env, respond)
return self.run_wsgi(req)
def run_wsgi(self, req):
"""Internal method to run the WSGI application.
This is typically only called by Mercurial. External consumers
should be using instances of this class as the WSGI application.
"""
with self._obtainrepo() as repo:
for r in self._runwsgi(req, repo):
yield r
def _runwsgi(self, req, repo):
rctx = requestcontext(self, repo)
# This state is global across all threads.
encoding.encoding = rctx.config('web', 'encoding', encoding.encoding)
rctx.repo.ui.environ = req.env
# work with CGI variables to create coherent structure
# use SCRIPT_NAME, PATH_INFO and QUERY_STRING as well as our REPO_NAME
req.url = req.env['SCRIPT_NAME']
if not req.url.endswith('/'):
req.url += '/'
if 'REPO_NAME' in req.env:
req.url += req.env['REPO_NAME'] + '/'
if 'PATH_INFO' in req.env:
parts = req.env['PATH_INFO'].strip('/').split('/')
repo_parts = req.env.get('REPO_NAME', '').split('/')
if parts[:len(repo_parts)] == repo_parts:
parts = parts[len(repo_parts):]
query = '/'.join(parts)
else:
query = req.env['QUERY_STRING'].split('&', 1)[0]
query = query.split(';', 1)[0]
# process this if it's a protocol request
# protocol bits don't need to create any URLs
# and the clients always use the old URL structure
cmd = req.form.get('cmd', [''])[0]
if protocol.iscmd(cmd):
try:
if query:
raise ErrorResponse(HTTP_NOT_FOUND)
if cmd in perms:
self.check_perm(rctx, req, perms[cmd])
return protocol.call(rctx.repo, req, cmd)
except ErrorResponse as inst:
# A client that sends unbundle without 100-continue will
# break if we respond early.
if (cmd == 'unbundle' and
(req.env.get('HTTP_EXPECT',
'').lower() != '100-continue') or
req.env.get('X-HgHttp2', '')):
req.drain()
else:
req.headers.append(('Connection', 'Close'))
req.respond(inst, protocol.HGTYPE,
body='0\n%s\n' % inst)
return ''
# translate user-visible url structure to internal structure
args = query.split('/', 2)
if 'cmd' not in req.form and args and args[0]:
cmd = args.pop(0)
style = cmd.rfind('-')
if style != -1:
req.form['style'] = [cmd[:style]]
cmd = cmd[style + 1:]
# avoid accepting e.g. style parameter as command
if util.safehasattr(webcommands, cmd):
req.form['cmd'] = [cmd]
if cmd == 'static':
req.form['file'] = ['/'.join(args)]
else:
if args and args[0]:
node = args.pop(0).replace('%2F', '/')
req.form['node'] = [node]
if args:
req.form['file'] = args
ua = req.env.get('HTTP_USER_AGENT', '')
if cmd == 'rev' and 'mercurial' in ua:
req.form['style'] = ['raw']
if cmd == 'archive':
fn = req.form['node'][0]
for type_, spec in rctx.archivespecs.iteritems():
ext = spec[2]
if fn.endswith(ext):
req.form['node'] = [fn[:-len(ext)]]
req.form['type'] = [type_]
# process the web interface request
try:
tmpl = rctx.templater(req)
ctype = tmpl('mimetype', encoding=encoding.encoding)
ctype = templater.stringify(ctype)
# check read permissions non-static content
if cmd != 'static':
self.check_perm(rctx, req, None)
if cmd == '':
req.form['cmd'] = [tmpl.cache['default']]
cmd = req.form['cmd'][0]
if rctx.configbool('web', 'cache', True):
caching(self, req) # sets ETag header or raises NOT_MODIFIED
if cmd not in webcommands.__all__:
msg = 'no such method: %s' % cmd
raise ErrorResponse(HTTP_BAD_REQUEST, msg)
elif cmd == 'file' and 'raw' in req.form.get('style', []):
rctx.ctype = ctype
content = webcommands.rawfile(rctx, req, tmpl)
else:
content = getattr(webcommands, cmd)(rctx, req, tmpl)
req.respond(HTTP_OK, ctype)
return content
except (error.LookupError, error.RepoLookupError) as err:
req.respond(HTTP_NOT_FOUND, ctype)
msg = str(err)
if (util.safehasattr(err, 'name') and
not isinstance(err, error.ManifestLookupError)):
msg = 'revision not found: %s' % err.name
return tmpl('error', error=msg)
except (error.RepoError, error.RevlogError) as inst:
req.respond(HTTP_SERVER_ERROR, ctype)
return tmpl('error', error=str(inst))
except ErrorResponse as inst:
req.respond(inst, ctype)
if inst.code == HTTP_NOT_MODIFIED:
# Not allowed to return a body on a 304
return ['']
return tmpl('error', error=str(inst))
def check_perm(self, rctx, req, op):
for permhook in permhooks:
permhook(rctx, req, op)
def getwebview(repo):
"""The 'web.view' config controls changeset filter to hgweb. Possible
values are ``served``, ``visible`` and ``all``. Default is ``served``.
The ``served`` filter only shows changesets that can be pulled from the
hgweb instance. The``visible`` filter includes secret changesets but
still excludes "hidden" one.
See the repoview module for details.
The option has been around undocumented since Mercurial 2.5, but no
user ever asked about it. So we better keep it undocumented for now."""
viewconfig = repo.ui.config('web', 'view', 'served',
untrusted=True)
if viewconfig == 'all':
return repo.unfiltered()
elif viewconfig in repoview.filtertable:
return repo.filtered(viewconfig)
else:
return repo.filtered('served')