##// END OF EJS Templates
util: lower water mark when removing nodes after cost limit reached...
util: lower water mark when removing nodes after cost limit reached See the inline comment for the reasoning here. This is a pretty common strategy for garbage collectors, other cache-like primtives. The performance impact is substantial: $ hg perflrucachedict --size 4 --gets 1000000 --sets 1000000 --mixed 1000000 --costlimit 100 ! inserts w/ cost limit ! wall 1.659181 comb 1.650000 user 1.650000 sys 0.000000 (best of 7) ! wall 1.722122 comb 1.720000 user 1.720000 sys 0.000000 (best of 6) ! mixed w/ cost limit ! wall 1.139955 comb 1.140000 user 1.140000 sys 0.000000 (best of 9) ! wall 1.182513 comb 1.180000 user 1.180000 sys 0.000000 (best of 9) $ hg perflrucachedict --size 1000 --gets 1000000 --sets 1000000 --mixed 1000000 --costlimit 10000 ! inserts ! wall 0.679546 comb 0.680000 user 0.680000 sys 0.000000 (best of 15) ! sets ! wall 0.825147 comb 0.830000 user 0.830000 sys 0.000000 (best of 13) ! inserts w/ cost limit ! wall 25.105273 comb 25.080000 user 25.080000 sys 0.000000 (best of 3) ! wall 1.724397 comb 1.720000 user 1.720000 sys 0.000000 (best of 6) ! mixed ! wall 0.807096 comb 0.810000 user 0.810000 sys 0.000000 (best of 13) ! mixed w/ cost limit ! wall 12.104470 comb 12.070000 user 12.070000 sys 0.000000 (best of 3) ! wall 1.190563 comb 1.190000 user 1.190000 sys 0.000000 (best of 9) $ hg perflrucachedict --size 1000 --gets 1000000 --sets 1000000 --mixed 1000000 --costlimit 10000 --mixedgetfreq 90 ! inserts ! wall 0.711177 comb 0.710000 user 0.710000 sys 0.000000 (best of 14) ! sets ! wall 0.846992 comb 0.850000 user 0.850000 sys 0.000000 (best of 12) ! inserts w/ cost limit ! wall 25.963028 comb 25.960000 user 25.960000 sys 0.000000 (best of 3) ! wall 2.184311 comb 2.180000 user 2.180000 sys 0.000000 (best of 5) ! mixed ! wall 0.728256 comb 0.730000 user 0.730000 sys 0.000000 (best of 14) ! mixed w/ cost limit ! wall 3.174256 comb 3.170000 user 3.170000 sys 0.000000 (best of 4) ! wall 0.773186 comb 0.770000 user 0.770000 sys 0.000000 (best of 13) $ hg perflrucachedict --size 100000 --gets 1000000 --sets 1000000 --mixed 1000000 --mixedgetfreq 90 --costlimit 5000000 ! gets ! wall 1.191368 comb 1.190000 user 1.190000 sys 0.000000 (best of 9) ! wall 1.195304 comb 1.190000 user 1.190000 sys 0.000000 (best of 9) ! inserts ! wall 0.950995 comb 0.950000 user 0.950000 sys 0.000000 (best of 11) ! inserts w/ cost limit ! wall 1.589732 comb 1.590000 user 1.590000 sys 0.000000 (best of 7) ! sets ! wall 1.094941 comb 1.100000 user 1.090000 sys 0.010000 (best of 9) ! mixed ! wall 0.936420 comb 0.940000 user 0.930000 sys 0.010000 (best of 10) ! mixed w/ cost limit ! wall 0.882780 comb 0.870000 user 0.870000 sys 0.000000 (best of 11) This puts us ~2x slower than caches without cost accounting. And for read-heavy workloads (the prime use cases for caches), performance is nearly identical. In the worst case (pure write workloads with cost accounting enabled), we're looking at ~1.5us per insert on large caches. That seems "fast enough." Differential Revision: https://phab.mercurial-scm.org/D4505

File last commit:

r39597:d06834e0 default
r39606:f296c0b3 default
Show More
wireprotov2peer.py
293 lines | 9.0 KiB | text/x-python | PythonLexer
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 # wireprotov2peer.py - client side code for wire protocol version 2
#
# Copyright 2018 Gregory Szorc <gregory.szorc@gmail.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
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 import threading
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 from .i18n import _
from . import (
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739 encoding,
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 error,
wireprotoframing,
)
Gregory Szorc
wireprotov2peer: use our CBOR decoder...
r39481 from .utils import (
cborutil,
)
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def formatrichmessage(atoms):
"""Format an encoded message from the framing protocol."""
chunks = []
for atom in atoms:
msg = _(atom[b'msg'])
if b'args' in atom:
Gregory Szorc
wireprotov2peer: properly format errors...
r39522 msg = msg % tuple(atom[b'args'])
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743
chunks.append(msg)
return b''.join(chunks)
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738 class commandresponse(object):
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 """Represents the response to a command request.
Instances track the state of the command and hold its results.
An external entity is required to update the state of the object when
events occur.
"""
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738
def __init__(self, requestid, command):
self.requestid = requestid
self.command = command
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 # Whether all remote input related to this command has been
# received.
self._inputcomplete = False
# We have a lock that is acquired when important object state is
# mutated. This is to prevent race conditions between 1 thread
# sending us new data and another consuming it.
self._lock = threading.RLock()
# An event is set when state of the object changes. This event
# is waited on by the generator emitting objects.
self._serviceable = threading.Event()
self._pendingevents = []
self._decoder = cborutil.bufferingdecoder()
self._seeninitial = False
def _oninputcomplete(self):
with self._lock:
self._inputcomplete = True
self._serviceable.set()
def _onresponsedata(self, data):
available, readcount, wanted = self._decoder.decode(data)
if not available:
return
with self._lock:
for o in self._decoder.getavailable():
if not self._seeninitial:
self._handleinitial(o)
continue
self._pendingevents.append(o)
self._serviceable.set()
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 def _handleinitial(self, o):
self._seeninitial = True
if o[b'status'] == 'ok':
return
atoms = [{'msg': o[b'error'][b'message']}]
if b'args' in o[b'error']:
atoms[0]['args'] = o[b'error'][b'args']
raise error.RepoError(formatrichmessage(atoms))
def objects(self):
"""Obtained decoded objects from this response.
This is a generator of data structures that were decoded from the
command response.
Obtaining the next member of the generator may block due to waiting
on external data to become available.
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 If the server encountered an error in the middle of serving the data
or if another error occurred, an exception may be raised when
advancing the generator.
"""
while True:
# TODO this can infinite loop if self._inputcomplete is never
# set. We likely want to tie the lifetime of this object/state
# to that of the background thread receiving frames and updating
# our state.
self._serviceable.wait(1.0)
with self._lock:
self._serviceable.clear()
# Make copies because objects could be mutated during
# iteration.
stop = self._inputcomplete
pending = list(self._pendingevents)
self._pendingevents[:] = []
for o in pending:
yield o
if stop:
break
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 class clienthandler(object):
"""Object to handle higher-level client activities.
The ``clientreactor`` is used to hold low-level state about the frame-based
protocol, such as which requests and streams are active. This type is used
for higher-level operations, such as reading frames from a socket, exposing
and managing a higher-level primitive for representing command responses,
etc. This class is what peers should probably use to bridge wire activity
with the higher-level peer API.
"""
def __init__(self, ui, clientreactor):
self._ui = ui
self._reactor = clientreactor
self._requests = {}
self._futures = {}
self._responses = {}
def callcommand(self, command, args, f):
"""Register a request to call a command.
Returns an iterable of frames that should be sent over the wire.
"""
request, action, meta = self._reactor.callcommand(command, args)
if action != 'noop':
raise error.ProgrammingError('%s not yet supported' % action)
rid = request.requestid
self._requests[rid] = request
self._futures[rid] = f
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 # TODO we need some kind of lifetime on response instances otherwise
# objects() may deadlock.
Gregory Szorc
wireprotov2: establish a type for representing command response...
r37738 self._responses[rid] = commandresponse(rid, command)
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737
return iter(())
def flushcommands(self):
"""Flush all queued commands.
Returns an iterable of frames that should be sent over the wire.
"""
action, meta = self._reactor.flushcommands()
if action != 'sendframes':
raise error.ProgrammingError('%s not yet supported' % action)
return meta['framegen']
def readframe(self, fh):
"""Attempt to read and process a frame.
Returns None if no frame was read. Presumably this means EOF.
"""
frame = wireprotoframing.readframe(fh)
if frame is None:
# TODO tell reactor?
return
self._ui.note(_('received %r\n') % frame)
self._processframe(frame)
return True
def _processframe(self, frame):
"""Process a single read frame."""
action, meta = self._reactor.onframerecv(frame)
if action == 'error':
e = error.RepoError(meta['message'])
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 if frame.requestid in self._responses:
self._responses[frame.requestid]._oninputcomplete()
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 if frame.requestid in self._futures:
self._futures[frame.requestid].set_exception(e)
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 del self._futures[frame.requestid]
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 else:
raise e
Gregory Szorc
wireprotov2: implement commands as a generator of objects...
r39595 return
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 if frame.requestid not in self._requests:
raise error.ProgrammingError(
'received frame for unknown request; this is either a bug in '
'the clientreactor not screening for this or this instance was '
'never told about this request: %r' % frame)
response = self._responses[frame.requestid]
if action == 'responsedata':
Gregory Szorc
wireprotov2peer: report exceptions in frame handling against request future...
r39521 # Any failures processing this frame should bubble up to the
# future tracking the request.
try:
self._processresponsedata(frame, meta, response)
except BaseException as e:
self._futures[frame.requestid].set_exception(e)
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 del self._futures[frame.requestid]
response._oninputcomplete()
Gregory Szorc
wireprotov2: move response handling out of httppeer...
r37737 else:
raise error.ProgrammingError(
'unhandled action from clientreactor: %s' % action)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
Gregory Szorc
wireprotov2peer: split responsedata handling into separate function...
r39469 def _processresponsedata(self, frame, meta, response):
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 # This can raise. The caller can handle it.
response._onresponsedata(meta['data'])
Gregory Szorc
wireprotov2peer: split responsedata handling into separate function...
r39469
if meta['eos']:
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 response._oninputcomplete()
del self._requests[frame.requestid]
Gregory Szorc
wireprotov2peer: split responsedata handling into separate function...
r39469
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 # If the command has a decoder, we wait until all input has been
# received before resolving the future. Otherwise we resolve the
# future immediately.
if frame.requestid not in self._futures:
return
Gregory Szorc
wireprotov2peer: split responsedata handling into separate function...
r39469
Gregory Szorc
wireprotov2peer: stream decoded responses...
r39597 if response.command not in COMMAND_DECODERS:
self._futures[frame.requestid].set_result(response.objects())
del self._futures[frame.requestid]
elif response._inputcomplete:
decoded = COMMAND_DECODERS[response.command](response.objects())
self._futures[frame.requestid].set_result(decoded)
Gregory Szorc
wireprotov2peer: split responsedata handling into separate function...
r39469 del self._futures[frame.requestid]
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodebranchmap(objs):
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739 # Response should be a single CBOR map of branch name to array of nodes.
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 bm = next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
return {encoding.tolocal(k): v for k, v in bm.items()}
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodeheads(objs):
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739 # Array of node bytestrings.
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 return next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodeknown(objs):
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739 # Bytestring where each byte is a 0 or 1.
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 raw = next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
return [True if c == '1' else False for c in raw]
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodelistkeys(objs):
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739 # Map with bytestring keys and values.
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 return next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodelookup(objs):
return next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
Gregory Szorc
wireprotov2: change command response protocol to include a leading map...
r37743 def decodepushkey(objs):
return next(objs)
Gregory Szorc
wireprotov2: decode responses to their expected types...
r37739
COMMAND_DECODERS = {
'branchmap': decodebranchmap,
'heads': decodeheads,
'known': decodeknown,
'listkeys': decodelistkeys,
'lookup': decodelookup,
'pushkey': decodepushkey,
}