##// END OF EJS Templates
wireproto: add streams to frame-based protocol...
wireproto: add streams to frame-based protocol Previously, the frame-based protocol was just a series of frames, with each frame associated with a request ID. In order to scale the protocol, we'll want to enable the use of compression. While it is possible to enable compression at the socket/pipe level, this has its disadvantages. The big one is it undermines the point of frames being standalone, atomic units that can be read and written: if you add compression above the framing protocol, you are back to having a stream-based protocol as opposed to something frame-based. So in order to preserve frames, compression needs to occur at the frame payload level. Compressing each frame's payload individually will limit compression ratios because the window size of the compressor will be limited by the max frame size, which is 32-64kb as currently defined. It will also add CPU overhead, as it is more efficient for compressors to operate on fewer, larger blocks of data than more, smaller blocks. So compressing each frame independently is out. This means we need to compress each frame's payload as if it is part of a larger stream. The simplest approach is to have 1 stream per connection. This could certainly work. However, it has disadvantages (documented below). We could also have 1 stream per RPC/command invocation. (This is the model HTTP/2 goes with.) This also has disadvantages. The main disadvantage to one global stream is that it has the very real potential to create CPU bottlenecks doing compression. Networks are only getting faster and the performance of single CPU cores has been relatively flat. Newer compression formats like zstandard offer better CPU cycle efficiency than predecessors like zlib. But it still all too common to saturate your CPU with compression overhead long before you saturate the network pipe. The main disadvantage with streams per request is that you can't reap the benefits of the compression context for multiple requests. For example, if you send 1000 RPC requests (or HTTP/2 requests for that matter), the response to each would have its own compression context. The overall size of the raw responses would be larger because compression contexts wouldn't be able to reference data from another request or response. The approach for streams as implemented in this commit is to support N streams per connection and for streams to potentially span requests and responses. As explained by the added internals docs, this facilitates servers and clients delegating independent streams and compression to independent threads / CPU cores. This helps alleviate the CPU bottleneck of compression. This design also allows compression contexts to be reused across requests/responses. This can result in improved compression ratios and less overhead for compressors and decompressors having to build new contexts. Another feature that was defined was the ability for individual frames within a stream to declare whether that individual frame's payload uses the content encoding (read: compression) defined by the stream. The idea here is that some servers may serve data from a combination of caches and dynamic resolution. Data coming from caches may be pre-compressed. We want to facilitate servers being able to essentially stream bytes from caches to the wire with minimal overhead. Being able to mix and match with frames are compressed within a stream enables these types of advanced server functionality. This commit defines the new streams mechanism. Basic code for supporting streams in frames has been added. But that code is seriously lacking and doesn't fully conform to the defined protocol. For example, we don't close any streams. And support for content encoding within streams is not yet implemented. The change was rather invasive and I didn't think it would be reasonable to implement the entire feature in a single commit. For the record, I would have loved to reuse an existing multiplexing protocol to build the new wire protocol on top of. However, I couldn't find a protocol that offers the performance and scaling characteristics that I desired. Namely, it should support multiple compression contexts to facilitate scaling out to multiple CPU cores and compression contexts should be able to live longer than single RPC requests. HTTP/2 *almost* fits the bill. But the semantics of HTTP message exchange state that streams can only live for a single request-response. We /could/ tunnel on top of HTTP/2 streams and frames with HEADER and DATA frames. But there's no guarantee that HTTP/2 libraries and proxies would allow us to use HTTP/2 streams and frames without the HTTP message exchange semantics defined in RFC 7540 Section 8. Other RPC protocols like gRPC tunnel are built on top of HTTP/2 and thus preserve its semantics of stream per RPC invocation. Even QUIC does this. We could attempt to invent a higher-level stream that spans HTTP/2 streams. But this would be violating HTTP/2 because there is no guarantee that HTTP/2 streams are routed to the same server. The best we can do - which is what this protocol does - is shoehorn all request and response data into a single HTTP message and create streams within. At that point, we've defined a Content-Type in HTTP parlance. It just so happens our media type can also work as a standalone, stream-based protocol, without leaning on HTTP or similar protocol. Differential Revision: https://phab.mercurial-scm.org/D2907

File last commit:

r35760:7a1806e0 default
r37304:9bfcbe4f default
Show More
sparse.py
693 lines | 23.2 KiB | text/x-python | PythonLexer
# sparse.py - functionality for sparse checkouts
#
# Copyright 2014 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.
from __future__ import absolute_import
import collections
import hashlib
import os
from .i18n import _
from .node import (
hex,
nullid,
)
from . import (
error,
match as matchmod,
merge as mergemod,
pathutil,
pycompat,
scmutil,
util,
)
# Whether sparse features are enabled. This variable is intended to be
# temporary to facilitate porting sparse to core. It should eventually be
# a per-repo option, possibly a repo requirement.
enabled = False
def parseconfig(ui, raw):
"""Parse sparse config file content.
Returns a tuple of includes, excludes, and profiles.
"""
includes = set()
excludes = set()
profiles = set()
current = None
havesection = False
for line in raw.split('\n'):
line = line.strip()
if not line or line.startswith('#'):
# empty or comment line, skip
continue
elif line.startswith('%include '):
line = line[9:].strip()
if line:
profiles.add(line)
elif line == '[include]':
if havesection and current != includes:
# TODO pass filename into this API so we can report it.
raise error.Abort(_('sparse config cannot have includes ' +
'after excludes'))
havesection = True
current = includes
continue
elif line == '[exclude]':
havesection = True
current = excludes
elif line:
if current is None:
raise error.Abort(_('sparse config entry outside of '
'section: %s') % line,
hint=_('add an [include] or [exclude] line '
'to declare the entry type'))
if line.strip().startswith('/'):
ui.warn(_('warning: sparse profile cannot use' +
' paths starting with /, ignoring %s\n') % line)
continue
current.add(line)
return includes, excludes, profiles
# Exists as separate function to facilitate monkeypatching.
def readprofile(repo, profile, changeid):
"""Resolve the raw content of a sparse profile file."""
# TODO add some kind of cache here because this incurs a manifest
# resolve and can be slow.
return repo.filectx(profile, changeid=changeid).data()
def patternsforrev(repo, rev):
"""Obtain sparse checkout patterns for the given rev.
Returns a tuple of iterables representing includes, excludes, and
patterns.
"""
# Feature isn't enabled. No-op.
if not enabled:
return set(), set(), set()
raw = repo.vfs.tryread('sparse')
if not raw:
return set(), set(), set()
if rev is None:
raise error.Abort(_('cannot parse sparse patterns from working '
'directory'))
includes, excludes, profiles = parseconfig(repo.ui, raw)
ctx = repo[rev]
if profiles:
visited = set()
while profiles:
profile = profiles.pop()
if profile in visited:
continue
visited.add(profile)
try:
raw = readprofile(repo, profile, rev)
except error.ManifestLookupError:
msg = (
"warning: sparse profile '%s' not found "
"in rev %s - ignoring it\n" % (profile, ctx))
# experimental config: sparse.missingwarning
if repo.ui.configbool(
'sparse', 'missingwarning'):
repo.ui.warn(msg)
else:
repo.ui.debug(msg)
continue
pincludes, pexcludes, subprofs = parseconfig(repo.ui, raw)
includes.update(pincludes)
excludes.update(pexcludes)
profiles.update(subprofs)
profiles = visited
if includes:
includes.add('.hg*')
return includes, excludes, profiles
def activeconfig(repo):
"""Determine the active sparse config rules.
Rules are constructed by reading the current sparse config and bringing in
referenced profiles from parents of the working directory.
"""
revs = [repo.changelog.rev(node) for node in
repo.dirstate.parents() if node != nullid]
allincludes = set()
allexcludes = set()
allprofiles = set()
for rev in revs:
includes, excludes, profiles = patternsforrev(repo, rev)
allincludes |= includes
allexcludes |= excludes
allprofiles |= profiles
return allincludes, allexcludes, allprofiles
def configsignature(repo, includetemp=True):
"""Obtain the signature string for the current sparse configuration.
This is used to construct a cache key for matchers.
"""
cache = repo._sparsesignaturecache
signature = cache.get('signature')
if includetemp:
tempsignature = cache.get('tempsignature')
else:
tempsignature = '0'
if signature is None or (includetemp and tempsignature is None):
signature = hex(hashlib.sha1(repo.vfs.tryread('sparse')).digest())
cache['signature'] = signature
if includetemp:
raw = repo.vfs.tryread('tempsparse')
tempsignature = hex(hashlib.sha1(raw).digest())
cache['tempsignature'] = tempsignature
return '%s %s' % (signature, tempsignature)
def writeconfig(repo, includes, excludes, profiles):
"""Write the sparse config file given a sparse configuration."""
with repo.vfs('sparse', 'wb') as fh:
for p in sorted(profiles):
fh.write('%%include %s\n' % p)
if includes:
fh.write('[include]\n')
for i in sorted(includes):
fh.write(i)
fh.write('\n')
if excludes:
fh.write('[exclude]\n')
for e in sorted(excludes):
fh.write(e)
fh.write('\n')
repo._sparsesignaturecache.clear()
def readtemporaryincludes(repo):
raw = repo.vfs.tryread('tempsparse')
if not raw:
return set()
return set(raw.split('\n'))
def writetemporaryincludes(repo, includes):
repo.vfs.write('tempsparse', '\n'.join(sorted(includes)))
repo._sparsesignaturecache.clear()
def addtemporaryincludes(repo, additional):
includes = readtemporaryincludes(repo)
for i in additional:
includes.add(i)
writetemporaryincludes(repo, includes)
def prunetemporaryincludes(repo):
if not enabled or not repo.vfs.exists('tempsparse'):
return
s = repo.status()
if s.modified or s.added or s.removed or s.deleted:
# Still have pending changes. Don't bother trying to prune.
return
sparsematch = matcher(repo, includetemp=False)
dirstate = repo.dirstate
actions = []
dropped = []
tempincludes = readtemporaryincludes(repo)
for file in tempincludes:
if file in dirstate and not sparsematch(file):
message = _('dropping temporarily included sparse files')
actions.append((file, None, message))
dropped.append(file)
typeactions = collections.defaultdict(list)
typeactions['r'] = actions
mergemod.applyupdates(repo, typeactions, repo[None], repo['.'], False)
# Fix dirstate
for file in dropped:
dirstate.drop(file)
repo.vfs.unlink('tempsparse')
repo._sparsesignaturecache.clear()
msg = _('cleaned up %d temporarily added file(s) from the '
'sparse checkout\n')
repo.ui.status(msg % len(tempincludes))
def forceincludematcher(matcher, includes):
"""Returns a matcher that returns true for any of the forced includes
before testing against the actual matcher."""
kindpats = [('path', include, '') for include in includes]
includematcher = matchmod.includematcher('', '', kindpats)
return matchmod.unionmatcher([includematcher, matcher])
def matcher(repo, revs=None, includetemp=True):
"""Obtain a matcher for sparse working directories for the given revs.
If multiple revisions are specified, the matcher is the union of all
revs.
``includetemp`` indicates whether to use the temporary sparse profile.
"""
# If sparse isn't enabled, sparse matcher matches everything.
if not enabled:
return matchmod.always(repo.root, '')
if not revs or revs == [None]:
revs = [repo.changelog.rev(node)
for node in repo.dirstate.parents() if node != nullid]
signature = configsignature(repo, includetemp=includetemp)
key = '%s %s' % (signature, ' '.join(map(pycompat.bytestr, revs)))
result = repo._sparsematchercache.get(key)
if result:
return result
matchers = []
for rev in revs:
try:
includes, excludes, profiles = patternsforrev(repo, rev)
if includes or excludes:
matcher = matchmod.match(repo.root, '', [],
include=includes, exclude=excludes,
default='relpath')
matchers.append(matcher)
except IOError:
pass
if not matchers:
result = matchmod.always(repo.root, '')
elif len(matchers) == 1:
result = matchers[0]
else:
result = matchmod.unionmatcher(matchers)
if includetemp:
tempincludes = readtemporaryincludes(repo)
result = forceincludematcher(result, tempincludes)
repo._sparsematchercache[key] = result
return result
def filterupdatesactions(repo, wctx, mctx, branchmerge, actions):
"""Filter updates to only lay out files that match the sparse rules."""
if not enabled:
return actions
oldrevs = [pctx.rev() for pctx in wctx.parents()]
oldsparsematch = matcher(repo, oldrevs)
if oldsparsematch.always():
return actions
files = set()
prunedactions = {}
if branchmerge:
# If we're merging, use the wctx filter, since we're merging into
# the wctx.
sparsematch = matcher(repo, [wctx.parents()[0].rev()])
else:
# If we're updating, use the target context's filter, since we're
# moving to the target context.
sparsematch = matcher(repo, [mctx.rev()])
temporaryfiles = []
for file, action in actions.iteritems():
type, args, msg = action
files.add(file)
if sparsematch(file):
prunedactions[file] = action
elif type == 'm':
temporaryfiles.append(file)
prunedactions[file] = action
elif branchmerge:
if type != 'k':
temporaryfiles.append(file)
prunedactions[file] = action
elif type == 'f':
prunedactions[file] = action
elif file in wctx:
prunedactions[file] = ('r', args, msg)
if len(temporaryfiles) > 0:
repo.ui.status(_('temporarily included %d file(s) in the sparse '
'checkout for merging\n') % len(temporaryfiles))
addtemporaryincludes(repo, temporaryfiles)
# Add the new files to the working copy so they can be merged, etc
actions = []
message = 'temporarily adding to sparse checkout'
wctxmanifest = repo[None].manifest()
for file in temporaryfiles:
if file in wctxmanifest:
fctx = repo[None][file]
actions.append((file, (fctx.flags(), False), message))
typeactions = collections.defaultdict(list)
typeactions['g'] = actions
mergemod.applyupdates(repo, typeactions, repo[None], repo['.'],
False)
dirstate = repo.dirstate
for file, flags, msg in actions:
dirstate.normal(file)
profiles = activeconfig(repo)[2]
changedprofiles = profiles & files
# If an active profile changed during the update, refresh the checkout.
# Don't do this during a branch merge, since all incoming changes should
# have been handled by the temporary includes above.
if changedprofiles and not branchmerge:
mf = mctx.manifest()
for file in mf:
old = oldsparsematch(file)
new = sparsematch(file)
if not old and new:
flags = mf.flags(file)
prunedactions[file] = ('g', (flags, False), '')
elif old and not new:
prunedactions[file] = ('r', [], '')
return prunedactions
def refreshwdir(repo, origstatus, origsparsematch, force=False):
"""Refreshes working directory by taking sparse config into account.
The old status and sparse matcher is compared against the current sparse
matcher.
Will abort if a file with pending changes is being excluded or included
unless ``force`` is True.
"""
# Verify there are no pending changes
pending = set()
pending.update(origstatus.modified)
pending.update(origstatus.added)
pending.update(origstatus.removed)
sparsematch = matcher(repo)
abort = False
for f in pending:
if not sparsematch(f):
repo.ui.warn(_("pending changes to '%s'\n") % f)
abort = not force
if abort:
raise error.Abort(_('could not update sparseness due to pending '
'changes'))
# Calculate actions
dirstate = repo.dirstate
ctx = repo['.']
added = []
lookup = []
dropped = []
mf = ctx.manifest()
files = set(mf)
actions = {}
for file in files:
old = origsparsematch(file)
new = sparsematch(file)
# Add files that are newly included, or that don't exist in
# the dirstate yet.
if (new and not old) or (old and new and not file in dirstate):
fl = mf.flags(file)
if repo.wvfs.exists(file):
actions[file] = ('e', (fl,), '')
lookup.append(file)
else:
actions[file] = ('g', (fl, False), '')
added.append(file)
# Drop files that are newly excluded, or that still exist in
# the dirstate.
elif (old and not new) or (not old and not new and file in dirstate):
dropped.append(file)
if file not in pending:
actions[file] = ('r', [], '')
# Verify there are no pending changes in newly included files
abort = False
for file in lookup:
repo.ui.warn(_("pending changes to '%s'\n") % file)
abort = not force
if abort:
raise error.Abort(_('cannot change sparseness due to pending '
'changes (delete the files or use '
'--force to bring them back dirty)'))
# Check for files that were only in the dirstate.
for file, state in dirstate.iteritems():
if not file in files:
old = origsparsematch(file)
new = sparsematch(file)
if old and not new:
dropped.append(file)
# Apply changes to disk
typeactions = dict((m, [])
for m in 'a f g am cd dc r dm dg m e k p pr'.split())
for f, (m, args, msg) in actions.iteritems():
if m not in typeactions:
typeactions[m] = []
typeactions[m].append((f, args, msg))
mergemod.applyupdates(repo, typeactions, repo[None], repo['.'], False)
# Fix dirstate
for file in added:
dirstate.normal(file)
for file in dropped:
dirstate.drop(file)
for file in lookup:
# File exists on disk, and we're bringing it back in an unknown state.
dirstate.normallookup(file)
return added, dropped, lookup
def aftercommit(repo, node):
"""Perform actions after a working directory commit."""
# This function is called unconditionally, even if sparse isn't
# enabled.
ctx = repo[node]
profiles = patternsforrev(repo, ctx.rev())[2]
# profiles will only have data if sparse is enabled.
if profiles & set(ctx.files()):
origstatus = repo.status()
origsparsematch = matcher(repo)
refreshwdir(repo, origstatus, origsparsematch, force=True)
prunetemporaryincludes(repo)
def _updateconfigandrefreshwdir(repo, includes, excludes, profiles,
force=False, removing=False):
"""Update the sparse config and working directory state."""
raw = repo.vfs.tryread('sparse')
oldincludes, oldexcludes, oldprofiles = parseconfig(repo.ui, raw)
oldstatus = repo.status()
oldmatch = matcher(repo)
oldrequires = set(repo.requirements)
# TODO remove this try..except once the matcher integrates better
# with dirstate. We currently have to write the updated config
# because that will invalidate the matcher cache and force a
# re-read. We ideally want to update the cached matcher on the
# repo instance then flush the new config to disk once wdir is
# updated. But this requires massive rework to matcher() and its
# consumers.
if 'exp-sparse' in oldrequires and removing:
repo.requirements.discard('exp-sparse')
scmutil.writerequires(repo.vfs, repo.requirements)
elif 'exp-sparse' not in oldrequires:
repo.requirements.add('exp-sparse')
scmutil.writerequires(repo.vfs, repo.requirements)
try:
writeconfig(repo, includes, excludes, profiles)
return refreshwdir(repo, oldstatus, oldmatch, force=force)
except Exception:
if repo.requirements != oldrequires:
repo.requirements.clear()
repo.requirements |= oldrequires
scmutil.writerequires(repo.vfs, repo.requirements)
writeconfig(repo, oldincludes, oldexcludes, oldprofiles)
raise
def clearrules(repo, force=False):
"""Clears include/exclude rules from the sparse config.
The remaining sparse config only has profiles, if defined. The working
directory is refreshed, as needed.
"""
with repo.wlock():
raw = repo.vfs.tryread('sparse')
includes, excludes, profiles = parseconfig(repo.ui, raw)
if not includes and not excludes:
return
_updateconfigandrefreshwdir(repo, set(), set(), profiles, force=force)
def importfromfiles(repo, opts, paths, force=False):
"""Import sparse config rules from files.
The updated sparse config is written out and the working directory
is refreshed, as needed.
"""
with repo.wlock():
# read current configuration
raw = repo.vfs.tryread('sparse')
includes, excludes, profiles = parseconfig(repo.ui, raw)
aincludes, aexcludes, aprofiles = activeconfig(repo)
# Import rules on top; only take in rules that are not yet
# part of the active rules.
changed = False
for p in paths:
with util.posixfile(util.expandpath(p)) as fh:
raw = fh.read()
iincludes, iexcludes, iprofiles = parseconfig(repo.ui, raw)
oldsize = len(includes) + len(excludes) + len(profiles)
includes.update(iincludes - aincludes)
excludes.update(iexcludes - aexcludes)
profiles.update(iprofiles - aprofiles)
if len(includes) + len(excludes) + len(profiles) > oldsize:
changed = True
profilecount = includecount = excludecount = 0
fcounts = (0, 0, 0)
if changed:
profilecount = len(profiles - aprofiles)
includecount = len(includes - aincludes)
excludecount = len(excludes - aexcludes)
fcounts = map(len, _updateconfigandrefreshwdir(
repo, includes, excludes, profiles, force=force))
printchanges(repo.ui, opts, profilecount, includecount, excludecount,
*fcounts)
def updateconfig(repo, pats, opts, include=False, exclude=False, reset=False,
delete=False, enableprofile=False, disableprofile=False,
force=False, usereporootpaths=False):
"""Perform a sparse config update.
Only one of the actions may be performed.
The new config is written out and a working directory refresh is performed.
"""
with repo.wlock():
raw = repo.vfs.tryread('sparse')
oldinclude, oldexclude, oldprofiles = parseconfig(repo.ui, raw)
if reset:
newinclude = set()
newexclude = set()
newprofiles = set()
else:
newinclude = set(oldinclude)
newexclude = set(oldexclude)
newprofiles = set(oldprofiles)
if any(os.path.isabs(pat) for pat in pats):
raise error.Abort(_('paths cannot be absolute'))
if not usereporootpaths:
# let's treat paths as relative to cwd
root, cwd = repo.root, repo.getcwd()
abspats = []
for kindpat in pats:
kind, pat = matchmod._patsplit(kindpat, None)
if kind in matchmod.cwdrelativepatternkinds or kind is None:
ap = (kind + ':' if kind else '') +\
pathutil.canonpath(root, cwd, pat)
abspats.append(ap)
else:
abspats.append(kindpat)
pats = abspats
if include:
newinclude.update(pats)
elif exclude:
newexclude.update(pats)
elif enableprofile:
newprofiles.update(pats)
elif disableprofile:
newprofiles.difference_update(pats)
elif delete:
newinclude.difference_update(pats)
newexclude.difference_update(pats)
profilecount = (len(newprofiles - oldprofiles) -
len(oldprofiles - newprofiles))
includecount = (len(newinclude - oldinclude) -
len(oldinclude - newinclude))
excludecount = (len(newexclude - oldexclude) -
len(oldexclude - newexclude))
fcounts = map(len, _updateconfigandrefreshwdir(
repo, newinclude, newexclude, newprofiles, force=force,
removing=reset))
printchanges(repo.ui, opts, profilecount, includecount,
excludecount, *fcounts)
def printchanges(ui, opts, profilecount=0, includecount=0, excludecount=0,
added=0, dropped=0, conflicting=0):
"""Print output summarizing sparse config changes."""
with ui.formatter('sparse', opts) as fm:
fm.startitem()
fm.condwrite(ui.verbose, 'profiles_added', _('Profiles changed: %d\n'),
profilecount)
fm.condwrite(ui.verbose, 'include_rules_added',
_('Include rules changed: %d\n'), includecount)
fm.condwrite(ui.verbose, 'exclude_rules_added',
_('Exclude rules changed: %d\n'), excludecount)
# In 'plain' verbose mode, mergemod.applyupdates already outputs what
# files are added or removed outside of the templating formatter
# framework. No point in repeating ourselves in that case.
if not fm.isplain():
fm.condwrite(ui.verbose, 'files_added', _('Files added: %d\n'),
added)
fm.condwrite(ui.verbose, 'files_dropped', _('Files dropped: %d\n'),
dropped)
fm.condwrite(ui.verbose, 'files_conflicting',
_('Files conflicting: %d\n'), conflicting)