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
# formatter.py - generic output formatting for mercurial
#
# Copyright 2012 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.
"""Generic output formatting for Mercurial
The formatter provides API to show data in various ways. The following
functions should be used in place of ui.write():
- fm.write() for unconditional output
- fm.condwrite() to show some extra data conditionally in plain output
- fm.context() to provide changectx to template output
- fm.data() to provide extra data to JSON or template output
- fm.plain() to show raw text that isn't provided to JSON or template output
To show structured data (e.g. date tuples, dicts, lists), apply fm.format*()
beforehand so the data is converted to the appropriate data type. Use
fm.isplain() if you need to convert or format data conditionally which isn't
supported by the formatter API.
To build nested structure (i.e. a list of dicts), use fm.nested().
See also https://www.mercurial-scm.org/wiki/GenericTemplatingPlan
fm.condwrite() vs 'if cond:':
In most cases, use fm.condwrite() so users can selectively show the data
in template output. If it's costly to build data, use plain 'if cond:' with
fm.write().
fm.nested() vs fm.formatdict() (or fm.formatlist()):
fm.nested() should be used to form a tree structure (a list of dicts of
lists of dicts...) which can be accessed through template keywords, e.g.
"{foo % "{bar % {...}} {baz % {...}}"}". On the other hand, fm.formatdict()
exports a dict-type object to template, which can be accessed by e.g.
"{get(foo, key)}" function.
Doctest helper:
>>> def show(fn, verbose=False, **opts):
... import sys
... from . import ui as uimod
... ui = uimod.ui()
... ui.verbose = verbose
... ui.pushbuffer()
... try:
... return fn(ui, ui.formatter(pycompat.sysbytes(fn.__name__),
... pycompat.byteskwargs(opts)))
... finally:
... print(pycompat.sysstr(ui.popbuffer()), end='')
Basic example:
>>> def files(ui, fm):
... files = [(b'foo', 123, (0, 0)), (b'bar', 456, (1, 0))]
... for f in files:
... fm.startitem()
... fm.write(b'path', b'%s', f[0])
... fm.condwrite(ui.verbose, b'date', b' %s',
... fm.formatdate(f[2], b'%Y-%m-%d %H:%M:%S'))
... fm.data(size=f[1])
... fm.plain(b'\\n')
... fm.end()
>>> show(files)
foo
bar
>>> show(files, verbose=True)
foo 1970-01-01 00:00:00
bar 1970-01-01 00:00:01
>>> show(files, template=b'json')
[
{
"date": [0, 0],
"path": "foo",
"size": 123
},
{
"date": [1, 0],
"path": "bar",
"size": 456
}
]
>>> show(files, template=b'path: {path}\\ndate: {date|rfc3339date}\\n')
path: foo
date: 1970-01-01T00:00:00+00:00
path: bar
date: 1970-01-01T00:00:01+00:00
Nested example:
>>> def subrepos(ui, fm):
... fm.startitem()
... fm.write(b'reponame', b'[%s]\\n', b'baz')
... files(ui, fm.nested(b'files'))
... fm.end()
>>> show(subrepos)
[baz]
foo
bar
>>> show(subrepos, template=b'{reponame}: {join(files % "{path}", ", ")}\\n')
baz: foo, bar
"""
from __future__ import absolute_import, print_function
import collections
import contextlib
import itertools
import os
from .i18n import _
from .node import (
hex,
short,
)
from . import (
error,
pycompat,
templatefilters,
templatekw,
templater,
templateutil,
util,
)
from .utils import dateutil
pickle = util.pickle
class _nullconverter(object):
'''convert non-primitive data types to be processed by formatter'''
# set to True if context object should be stored as item
storecontext = False
@staticmethod
def formatdate(date, fmt):
'''convert date tuple to appropriate format'''
return date
@staticmethod
def formatdict(data, key, value, fmt, sep):
'''convert dict or key-value pairs to appropriate dict format'''
# use plain dict instead of util.sortdict so that data can be
# serialized as a builtin dict in pickle output
return dict(data)
@staticmethod
def formatlist(data, name, fmt, sep):
'''convert iterable to appropriate list format'''
return list(data)
class baseformatter(object):
def __init__(self, ui, topic, opts, converter):
self._ui = ui
self._topic = topic
self._style = opts.get("style")
self._template = opts.get("template")
self._converter = converter
self._item = None
# function to convert node to string suitable for this output
self.hexfunc = hex
def __enter__(self):
return self
def __exit__(self, exctype, excvalue, traceback):
if exctype is None:
self.end()
def _showitem(self):
'''show a formatted item once all data is collected'''
def startitem(self):
'''begin an item in the format list'''
if self._item is not None:
self._showitem()
self._item = {}
def formatdate(self, date, fmt='%a %b %d %H:%M:%S %Y %1%2'):
'''convert date tuple to appropriate format'''
return self._converter.formatdate(date, fmt)
def formatdict(self, data, key='key', value='value', fmt=None, sep=' '):
'''convert dict or key-value pairs to appropriate dict format'''
return self._converter.formatdict(data, key, value, fmt, sep)
def formatlist(self, data, name, fmt=None, sep=' '):
'''convert iterable to appropriate list format'''
# name is mandatory argument for now, but it could be optional if
# we have default template keyword, e.g. {item}
return self._converter.formatlist(data, name, fmt, sep)
def context(self, **ctxs):
'''insert context objects to be used to render template keywords'''
ctxs = pycompat.byteskwargs(ctxs)
assert all(k in {'ctx', 'fctx'} for k in ctxs)
if self._converter.storecontext:
self._item.update(ctxs)
def data(self, **data):
'''insert data into item that's not shown in default output'''
data = pycompat.byteskwargs(data)
self._item.update(data)
def write(self, fields, deftext, *fielddata, **opts):
'''do default text output while assigning data to item'''
fieldkeys = fields.split()
assert len(fieldkeys) == len(fielddata)
self._item.update(zip(fieldkeys, fielddata))
def condwrite(self, cond, fields, deftext, *fielddata, **opts):
'''do conditional write (primarily for plain formatter)'''
fieldkeys = fields.split()
assert len(fieldkeys) == len(fielddata)
self._item.update(zip(fieldkeys, fielddata))
def plain(self, text, **opts):
'''show raw text for non-templated mode'''
def isplain(self):
'''check for plain formatter usage'''
return False
def nested(self, field):
'''sub formatter to store nested data in the specified field'''
self._item[field] = data = []
return _nestedformatter(self._ui, self._converter, data)
def end(self):
'''end output for the formatter'''
if self._item is not None:
self._showitem()
def nullformatter(ui, topic):
'''formatter that prints nothing'''
return baseformatter(ui, topic, opts={}, converter=_nullconverter)
class _nestedformatter(baseformatter):
'''build sub items and store them in the parent formatter'''
def __init__(self, ui, converter, data):
baseformatter.__init__(self, ui, topic='', opts={}, converter=converter)
self._data = data
def _showitem(self):
self._data.append(self._item)
def _iteritems(data):
'''iterate key-value pairs in stable order'''
if isinstance(data, dict):
return sorted(data.iteritems())
return data
class _plainconverter(object):
'''convert non-primitive data types to text'''
storecontext = False
@staticmethod
def formatdate(date, fmt):
'''stringify date tuple in the given format'''
return dateutil.datestr(date, fmt)
@staticmethod
def formatdict(data, key, value, fmt, sep):
'''stringify key-value pairs separated by sep'''
prefmt = pycompat.identity
if fmt is None:
fmt = '%s=%s'
prefmt = pycompat.bytestr
return sep.join(fmt % (prefmt(k), prefmt(v))
for k, v in _iteritems(data))
@staticmethod
def formatlist(data, name, fmt, sep):
'''stringify iterable separated by sep'''
prefmt = pycompat.identity
if fmt is None:
fmt = '%s'
prefmt = pycompat.bytestr
return sep.join(fmt % prefmt(e) for e in data)
class plainformatter(baseformatter):
'''the default text output scheme'''
def __init__(self, ui, out, topic, opts):
baseformatter.__init__(self, ui, topic, opts, _plainconverter)
if ui.debugflag:
self.hexfunc = hex
else:
self.hexfunc = short
if ui is out:
self._write = ui.write
else:
self._write = lambda s, **opts: out.write(s)
def startitem(self):
pass
def data(self, **data):
pass
def write(self, fields, deftext, *fielddata, **opts):
self._write(deftext % fielddata, **opts)
def condwrite(self, cond, fields, deftext, *fielddata, **opts):
'''do conditional write'''
if cond:
self._write(deftext % fielddata, **opts)
def plain(self, text, **opts):
self._write(text, **opts)
def isplain(self):
return True
def nested(self, field):
# nested data will be directly written to ui
return self
def end(self):
pass
class debugformatter(baseformatter):
def __init__(self, ui, out, topic, opts):
baseformatter.__init__(self, ui, topic, opts, _nullconverter)
self._out = out
self._out.write("%s = [\n" % self._topic)
def _showitem(self):
self._out.write(' %s,\n' % pycompat.byterepr(self._item))
def end(self):
baseformatter.end(self)
self._out.write("]\n")
class pickleformatter(baseformatter):
def __init__(self, ui, out, topic, opts):
baseformatter.__init__(self, ui, topic, opts, _nullconverter)
self._out = out
self._data = []
def _showitem(self):
self._data.append(self._item)
def end(self):
baseformatter.end(self)
self._out.write(pickle.dumps(self._data))
class jsonformatter(baseformatter):
def __init__(self, ui, out, topic, opts):
baseformatter.__init__(self, ui, topic, opts, _nullconverter)
self._out = out
self._out.write("[")
self._first = True
def _showitem(self):
if self._first:
self._first = False
else:
self._out.write(",")
self._out.write("\n {\n")
first = True
for k, v in sorted(self._item.items()):
if first:
first = False
else:
self._out.write(",\n")
u = templatefilters.json(v, paranoid=False)
self._out.write(' "%s": %s' % (k, u))
self._out.write("\n }")
def end(self):
baseformatter.end(self)
self._out.write("\n]\n")
class _templateconverter(object):
'''convert non-primitive data types to be processed by templater'''
storecontext = True
@staticmethod
def formatdate(date, fmt):
'''return date tuple'''
return date
@staticmethod
def formatdict(data, key, value, fmt, sep):
'''build object that can be evaluated as either plain string or dict'''
data = util.sortdict(_iteritems(data))
def f():
yield _plainconverter.formatdict(data, key, value, fmt, sep)
return templateutil.hybriddict(data, key=key, value=value, fmt=fmt,
gen=f)
@staticmethod
def formatlist(data, name, fmt, sep):
'''build object that can be evaluated as either plain string or list'''
data = list(data)
def f():
yield _plainconverter.formatlist(data, name, fmt, sep)
return templateutil.hybridlist(data, name=name, fmt=fmt, gen=f)
class templateformatter(baseformatter):
def __init__(self, ui, out, topic, opts):
baseformatter.__init__(self, ui, topic, opts, _templateconverter)
self._out = out
spec = lookuptemplate(ui, topic, opts.get('template', ''))
self._tref = spec.ref
self._t = loadtemplater(ui, spec, defaults=templatekw.keywords,
resources=templateresources(ui),
cache=templatekw.defaulttempl)
self._parts = templatepartsmap(spec, self._t,
['docheader', 'docfooter', 'separator'])
self._counter = itertools.count()
self._renderitem('docheader', {})
def _showitem(self):
item = self._item.copy()
item['index'] = index = next(self._counter)
if index > 0:
self._renderitem('separator', {})
self._renderitem(self._tref, item)
def _renderitem(self, part, item):
if part not in self._parts:
return
ref = self._parts[part]
self._out.write(self._t.render(ref, item))
def end(self):
baseformatter.end(self)
self._renderitem('docfooter', {})
templatespec = collections.namedtuple(r'templatespec',
r'ref tmpl mapfile')
def lookuptemplate(ui, topic, tmpl):
"""Find the template matching the given -T/--template spec 'tmpl'
'tmpl' can be any of the following:
- a literal template (e.g. '{rev}')
- a map-file name or path (e.g. 'changelog')
- a reference to [templates] in config file
- a path to raw template file
A map file defines a stand-alone template environment. If a map file
selected, all templates defined in the file will be loaded, and the
template matching the given topic will be rendered. Aliases won't be
loaded from user config, but from the map file.
If no map file selected, all templates in [templates] section will be
available as well as aliases in [templatealias].
"""
# looks like a literal template?
if '{' in tmpl:
return templatespec('', tmpl, None)
# perhaps a stock style?
if not os.path.split(tmpl)[0]:
mapname = (templater.templatepath('map-cmdline.' + tmpl)
or templater.templatepath(tmpl))
if mapname and os.path.isfile(mapname):
return templatespec(topic, None, mapname)
# perhaps it's a reference to [templates]
if ui.config('templates', tmpl):
return templatespec(tmpl, None, None)
if tmpl == 'list':
ui.write(_("available styles: %s\n") % templater.stylelist())
raise error.Abort(_("specify a template"))
# perhaps it's a path to a map or a template
if ('/' in tmpl or '\\' in tmpl) and os.path.isfile(tmpl):
# is it a mapfile for a style?
if os.path.basename(tmpl).startswith("map-"):
return templatespec(topic, None, os.path.realpath(tmpl))
with util.posixfile(tmpl, 'rb') as f:
tmpl = f.read()
return templatespec('', tmpl, None)
# constant string?
return templatespec('', tmpl, None)
def templatepartsmap(spec, t, partnames):
"""Create a mapping of {part: ref}"""
partsmap = {spec.ref: spec.ref} # initial ref must exist in t
if spec.mapfile:
partsmap.update((p, p) for p in partnames if p in t)
elif spec.ref:
for part in partnames:
ref = '%s:%s' % (spec.ref, part) # select config sub-section
if ref in t:
partsmap[part] = ref
return partsmap
def loadtemplater(ui, spec, defaults=None, resources=None, cache=None):
"""Create a templater from either a literal template or loading from
a map file"""
assert not (spec.tmpl and spec.mapfile)
if spec.mapfile:
frommapfile = templater.templater.frommapfile
return frommapfile(spec.mapfile, defaults=defaults, resources=resources,
cache=cache)
return maketemplater(ui, spec.tmpl, defaults=defaults, resources=resources,
cache=cache)
def maketemplater(ui, tmpl, defaults=None, resources=None, cache=None):
"""Create a templater from a string template 'tmpl'"""
aliases = ui.configitems('templatealias')
t = templater.templater(defaults=defaults, resources=resources,
cache=cache, aliases=aliases)
t.cache.update((k, templater.unquotestring(v))
for k, v in ui.configitems('templates'))
if tmpl:
t.cache[''] = tmpl
return t
class templateresources(templater.resourcemapper):
"""Resource mapper designed for the default templatekw and function"""
def __init__(self, ui, repo=None):
self._resmap = {
'cache': {}, # for templatekw/funcs to store reusable data
'repo': repo,
'ui': ui,
}
def availablekeys(self, context, mapping):
return {k for k, g in self._gettermap.iteritems()
if g(self, context, mapping, k) is not None}
def knownkeys(self):
return self._knownkeys
def lookup(self, context, mapping, key):
get = self._gettermap.get(key)
if not get:
return None
return get(self, context, mapping, key)
def populatemap(self, context, origmapping, newmapping):
mapping = {}
if self._hasctx(newmapping):
mapping['revcache'] = {} # per-ctx cache
if (('node' in origmapping or self._hasctx(origmapping))
and ('node' in newmapping or self._hasctx(newmapping))):
orignode = templateutil.runsymbol(context, origmapping, 'node')
mapping['originalnode'] = orignode
return mapping
def _getsome(self, context, mapping, key):
v = mapping.get(key)
if v is not None:
return v
return self._resmap.get(key)
def _hasctx(self, mapping):
return 'ctx' in mapping or 'fctx' in mapping
def _getctx(self, context, mapping, key):
ctx = mapping.get('ctx')
if ctx is not None:
return ctx
fctx = mapping.get('fctx')
if fctx is not None:
return fctx.changectx()
def _getrepo(self, context, mapping, key):
ctx = self._getctx(context, mapping, 'ctx')
if ctx is not None:
return ctx.repo()
return self._getsome(context, mapping, key)
_gettermap = {
'cache': _getsome,
'ctx': _getctx,
'fctx': _getsome,
'repo': _getrepo,
'revcache': _getsome,
'ui': _getsome,
}
_knownkeys = set(_gettermap.keys())
def formatter(ui, out, topic, opts):
template = opts.get("template", "")
if template == "json":
return jsonformatter(ui, out, topic, opts)
elif template == "pickle":
return pickleformatter(ui, out, topic, opts)
elif template == "debug":
return debugformatter(ui, out, topic, opts)
elif template != "":
return templateformatter(ui, out, topic, opts)
# developer config: ui.formatdebug
elif ui.configbool('ui', 'formatdebug'):
return debugformatter(ui, out, topic, opts)
# deprecated config: ui.formatjson
elif ui.configbool('ui', 'formatjson'):
return jsonformatter(ui, out, topic, opts)
return plainformatter(ui, out, topic, opts)
@contextlib.contextmanager
def openformatter(ui, filename, topic, opts):
"""Create a formatter that writes outputs to the specified file
Must be invoked using the 'with' statement.
"""
with util.posixfile(filename, 'wb') as out:
with formatter(ui, out, topic, opts) as fm:
yield fm
@contextlib.contextmanager
def _neverending(fm):
yield fm
def maybereopen(fm, filename, opts):
"""Create a formatter backed by file if filename specified, else return
the given formatter
Must be invoked using the 'with' statement. This will never call fm.end()
of the given formatter.
"""
if filename:
return openformatter(fm._ui, filename, fm._topic, opts)
else:
return _neverending(fm)