##// 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:

r36441:2831d918 default
r37304:9bfcbe4f default
Show More
progress.py
302 lines | 10.9 KiB | text/x-python | PythonLexer
# progress.py progress bars related code
#
# Copyright (C) 2010 Augie Fackler <durin42@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
import errno
import threading
import time
from .i18n import _
from . import encoding
def spacejoin(*args):
return ' '.join(s for s in args if s)
def shouldprint(ui):
return not (ui.quiet or ui.plain('progress')) and (
ui._isatty(ui.ferr) or ui.configbool('progress', 'assume-tty'))
def fmtremaining(seconds):
"""format a number of remaining seconds in human readable way
This will properly display seconds, minutes, hours, days if needed"""
if seconds < 60:
# i18n: format XX seconds as "XXs"
return _("%02ds") % (seconds)
minutes = seconds // 60
if minutes < 60:
seconds -= minutes * 60
# i18n: format X minutes and YY seconds as "XmYYs"
return _("%dm%02ds") % (minutes, seconds)
# we're going to ignore seconds in this case
minutes += 1
hours = minutes // 60
minutes -= hours * 60
if hours < 30:
# i18n: format X hours and YY minutes as "XhYYm"
return _("%dh%02dm") % (hours, minutes)
# we're going to ignore minutes in this case
hours += 1
days = hours // 24
hours -= days * 24
if days < 15:
# i18n: format X days and YY hours as "XdYYh"
return _("%dd%02dh") % (days, hours)
# we're going to ignore hours in this case
days += 1
weeks = days // 7
days -= weeks * 7
if weeks < 55:
# i18n: format X weeks and YY days as "XwYYd"
return _("%dw%02dd") % (weeks, days)
# we're going to ignore days and treat a year as 52 weeks
weeks += 1
years = weeks // 52
weeks -= years * 52
# i18n: format X years and YY weeks as "XyYYw"
return _("%dy%02dw") % (years, weeks)
# file_write() and file_flush() of Python 2 do not restart on EINTR if
# the file is attached to a "slow" device (e.g. a terminal) and raise
# IOError. We cannot know how many bytes would be written by file_write(),
# but a progress text is known to be short enough to be written by a
# single write() syscall, so we can just retry file_write() with the whole
# text. (issue5532)
#
# This should be a short-term workaround. We'll need to fix every occurrence
# of write() to a terminal or pipe.
def _eintrretry(func, *args):
while True:
try:
return func(*args)
except IOError as err:
if err.errno == errno.EINTR:
continue
raise
class progbar(object):
def __init__(self, ui):
self.ui = ui
self._refreshlock = threading.Lock()
self.resetstate()
def resetstate(self):
self.topics = []
self.topicstates = {}
self.starttimes = {}
self.startvals = {}
self.printed = False
self.lastprint = time.time() + float(self.ui.config(
'progress', 'delay'))
self.curtopic = None
self.lasttopic = None
self.indetcount = 0
self.refresh = float(self.ui.config(
'progress', 'refresh'))
self.changedelay = max(3 * self.refresh,
float(self.ui.config(
'progress', 'changedelay')))
self.order = self.ui.configlist('progress', 'format')
self.estimateinterval = self.ui.configwith(
float, 'progress', 'estimateinterval')
def show(self, now, topic, pos, item, unit, total):
if not shouldprint(self.ui):
return
termwidth = self.width()
self.printed = True
head = ''
needprogress = False
tail = ''
for indicator in self.order:
add = ''
if indicator == 'topic':
add = topic
elif indicator == 'number':
if total:
add = b'%*d/%d' % (len(str(total)), pos, total)
else:
add = b'%d' % pos
elif indicator.startswith('item') and item:
slice = 'end'
if '-' in indicator:
wid = int(indicator.split('-')[1])
elif '+' in indicator:
slice = 'beginning'
wid = int(indicator.split('+')[1])
else:
wid = 20
if slice == 'end':
add = encoding.trim(item, wid, leftside=True)
else:
add = encoding.trim(item, wid)
add += (wid - encoding.colwidth(add)) * ' '
elif indicator == 'bar':
add = ''
needprogress = True
elif indicator == 'unit' and unit:
add = unit
elif indicator == 'estimate':
add = self.estimate(topic, pos, total, now)
elif indicator == 'speed':
add = self.speed(topic, pos, unit, now)
if not needprogress:
head = spacejoin(head, add)
else:
tail = spacejoin(tail, add)
if needprogress:
used = 0
if head:
used += encoding.colwidth(head) + 1
if tail:
used += encoding.colwidth(tail) + 1
progwidth = termwidth - used - 3
if total and pos <= total:
amt = pos * progwidth // total
bar = '=' * (amt - 1)
if amt > 0:
bar += '>'
bar += ' ' * (progwidth - amt)
else:
progwidth -= 3
self.indetcount += 1
# mod the count by twice the width so we can make the
# cursor bounce between the right and left sides
amt = self.indetcount % (2 * progwidth)
amt -= progwidth
bar = (' ' * int(progwidth - abs(amt)) + '<=>' +
' ' * int(abs(amt)))
prog = ''.join(('[', bar, ']'))
out = spacejoin(head, prog, tail)
else:
out = spacejoin(head, tail)
self._writeerr('\r' + encoding.trim(out, termwidth))
self.lasttopic = topic
self._flusherr()
def clear(self):
if not self.printed or not self.lastprint or not shouldprint(self.ui):
return
self._writeerr('\r%s\r' % (' ' * self.width()))
if self.printed:
# force immediate re-paint of progress bar
self.lastprint = 0
def complete(self):
if not shouldprint(self.ui):
return
if self.ui.configbool('progress', 'clear-complete'):
self.clear()
else:
self._writeerr('\n')
self._flusherr()
def _flusherr(self):
_eintrretry(self.ui.ferr.flush)
def _writeerr(self, msg):
_eintrretry(self.ui.ferr.write, msg)
def width(self):
tw = self.ui.termwidth()
return min(int(self.ui.config('progress', 'width', default=tw)), tw)
def estimate(self, topic, pos, total, now):
if total is None:
return ''
initialpos = self.startvals[topic]
target = total - initialpos
delta = pos - initialpos
if delta > 0:
elapsed = now - self.starttimes[topic]
seconds = (elapsed * (target - delta)) // delta + 1
return fmtremaining(seconds)
return ''
def speed(self, topic, pos, unit, now):
initialpos = self.startvals[topic]
delta = pos - initialpos
elapsed = now - self.starttimes[topic]
if elapsed > 0:
return _('%d %s/sec') % (delta / elapsed, unit)
return ''
def _oktoprint(self, now):
'''Check if conditions are met to print - e.g. changedelay elapsed'''
if (self.lasttopic is None # first time we printed
# not a topic change
or self.curtopic == self.lasttopic
# it's been long enough we should print anyway
or now - self.lastprint >= self.changedelay):
return True
else:
return False
def _calibrateestimate(self, topic, now, pos):
'''Adjust starttimes and startvals for topic so ETA works better
If progress is non-linear (ex. get much slower in the last minute),
it's more friendly to only use a recent time span for ETA and speed
calculation.
[======================================> ]
^^^^^^^
estimateinterval, only use this for estimation
'''
interval = self.estimateinterval
if interval <= 0:
return
elapsed = now - self.starttimes[topic]
if elapsed > interval:
delta = pos - self.startvals[topic]
newdelta = delta * interval / elapsed
# If a stall happens temporarily, ETA could change dramatically
# frequently. This is to avoid such dramatical change and make ETA
# smoother.
if newdelta < 0.1:
return
self.startvals[topic] = pos - newdelta
self.starttimes[topic] = now - interval
def progress(self, topic, pos, item='', unit='', total=None):
now = time.time()
self._refreshlock.acquire()
try:
if pos is None:
self.starttimes.pop(topic, None)
self.startvals.pop(topic, None)
self.topicstates.pop(topic, None)
# reset the progress bar if this is the outermost topic
if self.topics and self.topics[0] == topic and self.printed:
self.complete()
self.resetstate()
# truncate the list of topics assuming all topics within
# this one are also closed
if topic in self.topics:
self.topics = self.topics[:self.topics.index(topic)]
# reset the last topic to the one we just unwound to,
# so that higher-level topics will be stickier than
# lower-level topics
if self.topics:
self.lasttopic = self.topics[-1]
else:
self.lasttopic = None
else:
if topic not in self.topics:
self.starttimes[topic] = now
self.startvals[topic] = pos
self.topics.append(topic)
self.topicstates[topic] = pos, item, unit, total
self.curtopic = topic
self._calibrateestimate(topic, now, pos)
if now - self.lastprint >= self.refresh and self.topics:
if self._oktoprint(now):
self.lastprint = now
self.show(now, topic, *self.topicstates[topic])
finally:
self._refreshlock.release()