##// 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:

r37244:9bcf096a default
r37304:9bfcbe4f default
Show More
registrar.py
443 lines | 15.1 KiB | text/x-python | PythonLexer
# registrar.py - utilities to register function for specific purpose
#
# Copyright FUJIWARA Katsunori <foozy@lares.dti.ne.jp> and others
#
# 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
from . import (
configitems,
error,
pycompat,
util,
)
# unlike the other registered items, config options are neither functions or
# classes. Registering the option is just small function call.
#
# We still add the official API to the registrar module for consistency with
# the other items extensions want might to register.
configitem = configitems.getitemregister
class _funcregistrarbase(object):
"""Base of decorator to register a function for specific purpose
This decorator stores decorated functions into own dict 'table'.
The least derived class can be defined by overriding 'formatdoc',
for example::
class keyword(_funcregistrarbase):
_docformat = ":%s: %s"
This should be used as below:
keyword = registrar.keyword()
@keyword('bar')
def barfunc(*args, **kwargs):
'''Explanation of bar keyword ....
'''
pass
In this case:
- 'barfunc' is stored as 'bar' in '_table' of an instance 'keyword' above
- 'barfunc.__doc__' becomes ":bar: Explanation of bar keyword"
"""
def __init__(self, table=None):
if table is None:
self._table = {}
else:
self._table = table
def __call__(self, decl, *args, **kwargs):
return lambda func: self._doregister(func, decl, *args, **kwargs)
def _doregister(self, func, decl, *args, **kwargs):
name = self._getname(decl)
if name in self._table:
msg = 'duplicate registration for name: "%s"' % name
raise error.ProgrammingError(msg)
if func.__doc__ and not util.safehasattr(func, '_origdoc'):
doc = pycompat.sysbytes(func.__doc__).strip()
func._origdoc = doc
func.__doc__ = pycompat.sysstr(self._formatdoc(decl, doc))
self._table[name] = func
self._extrasetup(name, func, *args, **kwargs)
return func
def _parsefuncdecl(self, decl):
"""Parse function declaration and return the name of function in it
"""
i = decl.find('(')
if i >= 0:
return decl[:i]
else:
return decl
def _getname(self, decl):
"""Return the name of the registered function from decl
Derived class should override this, if it allows more
descriptive 'decl' string than just a name.
"""
return decl
_docformat = None
def _formatdoc(self, decl, doc):
"""Return formatted document of the registered function for help
'doc' is '__doc__.strip()' of the registered function.
"""
return self._docformat % (decl, doc)
def _extrasetup(self, name, func):
"""Execute exra setup for registered function, if needed
"""
class command(_funcregistrarbase):
"""Decorator to register a command function to table
This class receives a command table as its argument. The table should
be a dict.
The created object can be used as a decorator for adding commands to
that command table. This accepts multiple arguments to define a command.
The first argument is the command name (as bytes).
The `options` keyword argument is an iterable of tuples defining command
arguments. See ``mercurial.fancyopts.fancyopts()`` for the format of each
tuple.
The `synopsis` argument defines a short, one line summary of how to use the
command. This shows up in the help output.
There are three arguments that control what repository (if any) is found
and passed to the decorated function: `norepo`, `optionalrepo`, and
`inferrepo`.
The `norepo` argument defines whether the command does not require a
local repository. Most commands operate against a repository, thus the
default is False. When True, no repository will be passed.
The `optionalrepo` argument defines whether the command optionally requires
a local repository. If no repository can be found, None will be passed
to the decorated function.
The `inferrepo` argument defines whether to try to find a repository from
the command line arguments. If True, arguments will be examined for
potential repository locations. See ``findrepo()``. If a repository is
found, it will be used and passed to the decorated function.
There are three constants in the class which tells what type of the command
that is. That information will be helpful at various places. It will be also
be used to decide what level of access the command has on hidden commits.
The constants are:
`unrecoverablewrite` is for those write commands which can't be recovered
like push.
`recoverablewrite` is for write commands which can be recovered like commit.
`readonly` is for commands which are read only.
The signature of the decorated function looks like this:
def cmd(ui[, repo] [, <args>] [, <options>])
`repo` is required if `norepo` is False.
`<args>` are positional args (or `*args`) arguments, of non-option
arguments from the command line.
`<options>` are keyword arguments (or `**options`) of option arguments
from the command line.
See the WritingExtensions and MercurialApi documentation for more exhaustive
descriptions and examples.
"""
unrecoverablewrite = "unrecoverable"
recoverablewrite = "recoverable"
readonly = "readonly"
possiblecmdtypes = {unrecoverablewrite, recoverablewrite, readonly}
def _doregister(self, func, name, options=(), synopsis=None,
norepo=False, optionalrepo=False, inferrepo=False,
cmdtype=unrecoverablewrite):
if cmdtype not in self.possiblecmdtypes:
raise error.ProgrammingError("unknown cmdtype value '%s' for "
"'%s' command" % (cmdtype, name))
func.norepo = norepo
func.optionalrepo = optionalrepo
func.inferrepo = inferrepo
func.cmdtype = cmdtype
if synopsis:
self._table[name] = func, list(options), synopsis
else:
self._table[name] = func, list(options)
return func
class revsetpredicate(_funcregistrarbase):
"""Decorator to register revset predicate
Usage::
revsetpredicate = registrar.revsetpredicate()
@revsetpredicate('mypredicate(arg1, arg2[, arg3])')
def mypredicatefunc(repo, subset, x):
'''Explanation of this revset predicate ....
'''
pass
The first string argument is used also in online help.
Optional argument 'safe' indicates whether a predicate is safe for
DoS attack (False by default).
Optional argument 'takeorder' indicates whether a predicate function
takes ordering policy as the last argument.
Optional argument 'weight' indicates the estimated run-time cost, useful
for static optimization, default is 1. Higher weight means more expensive.
Usually, revsets that are fast and return only one revision has a weight of
0.5 (ex. a symbol); revsets with O(changelog) complexity and read only the
changelog have weight 10 (ex. author); revsets reading manifest deltas have
weight 30 (ex. adds); revset reading manifest contents have weight 100
(ex. contains). Note: those values are flexible. If the revset has a
same big-O time complexity as 'contains', but with a smaller constant, it
might have a weight of 90.
'revsetpredicate' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'revsetpredicate' is used for
decorating in extension.
Otherwise, explicit 'revset.loadpredicate()' is needed.
"""
_getname = _funcregistrarbase._parsefuncdecl
_docformat = "``%s``\n %s"
def _extrasetup(self, name, func, safe=False, takeorder=False, weight=1):
func._safe = safe
func._takeorder = takeorder
func._weight = weight
class filesetpredicate(_funcregistrarbase):
"""Decorator to register fileset predicate
Usage::
filesetpredicate = registrar.filesetpredicate()
@filesetpredicate('mypredicate()')
def mypredicatefunc(mctx, x):
'''Explanation of this fileset predicate ....
'''
pass
The first string argument is used also in online help.
Optional argument 'callstatus' indicates whether a predicate
implies 'matchctx.status()' at runtime or not (False, by
default).
Optional argument 'callexisting' indicates whether a predicate
implies 'matchctx.existing()' at runtime or not (False, by
default).
'filesetpredicate' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'filesetpredicate' is used for
decorating in extension.
Otherwise, explicit 'fileset.loadpredicate()' is needed.
"""
_getname = _funcregistrarbase._parsefuncdecl
_docformat = "``%s``\n %s"
def _extrasetup(self, name, func, callstatus=False, callexisting=False):
func._callstatus = callstatus
func._callexisting = callexisting
class _templateregistrarbase(_funcregistrarbase):
"""Base of decorator to register functions as template specific one
"""
_docformat = ":%s: %s"
class templatekeyword(_templateregistrarbase):
"""Decorator to register template keyword
Usage::
templatekeyword = registrar.templatekeyword()
# new API (since Mercurial 4.6)
@templatekeyword('mykeyword', requires={'repo', 'ctx'})
def mykeywordfunc(context, mapping):
'''Explanation of this template keyword ....
'''
pass
# old API
@templatekeyword('mykeyword')
def mykeywordfunc(repo, ctx, templ, cache, revcache, **args):
'''Explanation of this template keyword ....
'''
pass
The first string argument is used also in online help.
Optional argument 'requires' should be a collection of resource names
which the template keyword depends on. This also serves as a flag to
switch to the new API. If 'requires' is unspecified, all template
keywords and resources are expanded to the function arguments.
'templatekeyword' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'templatekeyword' is used for
decorating in extension.
Otherwise, explicit 'templatekw.loadkeyword()' is needed.
"""
def _extrasetup(self, name, func, requires=None):
func._requires = requires
class templatefilter(_templateregistrarbase):
"""Decorator to register template filer
Usage::
templatefilter = registrar.templatefilter()
@templatefilter('myfilter', intype=bytes)
def myfilterfunc(text):
'''Explanation of this template filter ....
'''
pass
The first string argument is used also in online help.
Optional argument 'intype' defines the type of the input argument,
which should be (bytes, int, templateutil.date, or None for any.)
'templatefilter' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'templatefilter' is used for
decorating in extension.
Otherwise, explicit 'templatefilters.loadkeyword()' is needed.
"""
def _extrasetup(self, name, func, intype=None):
func._intype = intype
class templatefunc(_templateregistrarbase):
"""Decorator to register template function
Usage::
templatefunc = registrar.templatefunc()
@templatefunc('myfunc(arg1, arg2[, arg3])', argspec='arg1 arg2 arg3')
def myfuncfunc(context, mapping, args):
'''Explanation of this template function ....
'''
pass
The first string argument is used also in online help.
If optional 'argspec' is defined, the function will receive 'args' as
a dict of named arguments. Otherwise 'args' is a list of positional
arguments.
'templatefunc' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'templatefunc' is used for
decorating in extension.
Otherwise, explicit 'templatefuncs.loadfunction()' is needed.
"""
_getname = _funcregistrarbase._parsefuncdecl
def _extrasetup(self, name, func, argspec=None):
func._argspec = argspec
class internalmerge(_funcregistrarbase):
"""Decorator to register in-process merge tool
Usage::
internalmerge = registrar.internalmerge()
@internalmerge('mymerge', internalmerge.mergeonly,
onfailure=None, precheck=None):
def mymergefunc(repo, mynode, orig, fcd, fco, fca,
toolconf, files, labels=None):
'''Explanation of this internal merge tool ....
'''
return 1, False # means "conflicted", "no deletion needed"
The first string argument is used to compose actual merge tool name,
":name" and "internal:name" (the latter is historical one).
The second argument is one of merge types below:
========== ======== ======== =========
merge type precheck premerge fullmerge
========== ======== ======== =========
nomerge x x x
mergeonly o x o
fullmerge o o o
========== ======== ======== =========
Optional argument 'onfailure' is the format of warning message
to be used at failure of merging (target filename is specified
at formatting). Or, None or so, if warning message should be
suppressed.
Optional argument 'precheck' is the function to be used
before actual invocation of internal merge tool itself.
It takes as same arguments as internal merge tool does, other than
'files' and 'labels'. If it returns false value, merging is aborted
immediately (and file is marked as "unresolved").
'internalmerge' instance in example above can be used to
decorate multiple functions.
Decorated functions are registered automatically at loading
extension, if an instance named as 'internalmerge' is used for
decorating in extension.
Otherwise, explicit 'filemerge.loadinternalmerge()' is needed.
"""
_docformat = "``:%s``\n %s"
# merge type definitions:
nomerge = None
mergeonly = 'mergeonly' # just the full merge, no premerge
fullmerge = 'fullmerge' # both premerge and merge
def _extrasetup(self, name, func, mergetype,
onfailure=None, precheck=None):
func.mergetype = mergetype
func.onfailure = onfailure
func.precheck = precheck