修改Python的pyxmpp2中的主循環(huán)使其提高性能
引子
之前clubot使用的pyxmpp2的默認(rèn)mainloop也就是一個(gè)poll的主循環(huán),但是clubot上線后資源占用非常厲害,使用strace跟蹤發(fā)現(xiàn)clubot在不停的poll,查看pyxmpp2代碼發(fā)現(xiàn)pyxmpp2的poll在使用超時(shí)阻塞時(shí)使用最小超時(shí)時(shí)間,而最小超時(shí)時(shí)間一直是0,所以會(huì)變成一個(gè)沒(méi)有超時(shí)的非阻塞poll很浪費(fèi)資源,不打算更改庫(kù)代碼,所以自己仿照poll的mainloop寫(xiě)了一個(gè)更加高效的epoll的mainloop
實(shí)現(xiàn)
#!/usr/bin/env python
# -*- coding:utf-8 -*-
#
# Author : cold
# E-mail : wh_linux@126.com
# Date : 13/01/06 10:41:31
# Desc : Clubot epoll mainloop
#
from __future__ import absolute_import, division
import select
from pyxmpp2.mainloop.interfaces import HandlerReady, PrepareAgain
from pyxmpp2.mainloop.base import MainLoopBase
from plugin.util import get_logger
class EpollMainLoop(MainLoopBase):
""" Main event loop based on the epoll() syscall on Linux system """
READ_ONLY = (select.EPOLLIN | select.EPOLLPRI | select.EPOLLHUP |
select.EPOLLERR |select.EPOLLET)
READ_WRITE = READ_ONLY | select.EPOLLOUT
def __init__(self, settings = None, handlers= None):
self.epoll = select.epoll()
self._handlers = {}
self._unprepared_handlers = {}
self._timeout = None
self._exists_fd = {}
self.logger = get_logger()
MainLoopBase.__init__(self, settings, handlers)
return
def _add_io_handler(self, handler):
self._unprepared_handlers[handler] = None
self._configure_io_handler(handler)
def _configure_io_handler(self, handler):
if self.check_events():
return
if handler in self._unprepared_handlers:
old_fileno = self._unprepared_handlers[handler]
prepared = self._prepare_io_handler(handler)
else:
old_fileno = None
prepared = True
fileno = handler.fileno()
if old_fileno is not None and fileno != old_fileno:
del self._handlers[old_fileno]
self._exists.pop(old_fileno, None)
self.epoll.unregister(old_fileno)
if not prepared:
self._unprepared_handlers[handler] = fileno
if not fileno:
return
self._handlers[fileno] = handler
events = 0
if handler.is_readable():
events |= self.READ_ONLY
if handler.is_writable():
events |= self.READ_WRITE
if events:
if fileno in self._exists_fd:
self.epoll.modify(fileno, events)
else:
self._exists_fd.update({fileno:1})
self.epoll.register(fileno, events)
def _prepare_io_handler(self, handler):
ret = handler.prepare()
if isinstance(ret, HandlerReady):
del self._unprepared_handlers[handler]
prepared = True
elif isinstance(ret, PrepareAgain):
if ret.timeout is not None:
if self._timeout is not None:
self._timeout = min(self._timeout, ret.timeout)
else:
self._timeout = ret.timeout
prepared = False
else:
raise TypeError("Unexpected result from prepare()")
return prepared
def _remove_io_handler(self, handler):
if handler in self._unprepared_handlers:
old_fileno = self._unprepared_handlers[handler]
del self._unprepared_handlers[handler]
else:
old_fileno = handler.fileno()
if old_fileno is not None:
try:
del self._handlers[old_fileno]
self._exists.pop(old_fileno, None)
self.epoll.unregister(old_fileno)
except KeyError:
pass
def loop_iteration(self, timeout = 60):
next_timeout, sources_handled = self._call_timeout_handlers()
if self.check_events():
return
if self._quit:
return sources_handled
for handler in list(self._unprepared_handlers):
self._configure_io_handler(handler)
if self._timeout is not None:
timeout = min(timeout, self._timeout)
if next_timeout is not None:
timeout = min(next_timeout, timeout)
if timeout == 0:
timeout += 1 # 帶有超時(shí)的非阻塞,解約資源
events = self.epoll.poll(timeout)
for fd, flag in events:
if flag & (select.EPOLLIN | select.EPOLLPRI | select.EPOLLET):
self._handlers[fd].handle_read()
if flag & (select.EPOLLOUT|select.EPOLLET):
self._handlers[fd].handle_write()
if flag & (select.EPOLLERR | select.EPOLLET):
self._handlers[fd].handle_err()
if flag & (select.EPOLLHUP | select.EPOLLET):
self._handlers[fd].handle_hup()
#if flag & select.EPOLLNVAL:
#self._handlers[fd].handle_nval()
sources_handled += 1
self._configure_io_handler(self._handlers[fd])
return sources_handled
使用
如何使用新的mainloop?只需在實(shí)例化Client時(shí)傳入
mainloop = EpollMainLoop(settings) client = Client(my_jid, [self, version_provider], settings, mainloop)
這樣就會(huì)使用epoll作為mainloop
注意
epoll僅僅在Linux下支持
相關(guān)文章
Python?pycryptodome庫(kù)實(shí)現(xiàn)RSA加密解密消息
本文為大家介紹了如何在?Python?中使用?RSA?公鑰加密技術(shù)來(lái)加密和解密消息,并使用?pycryptodome?庫(kù)進(jìn)行實(shí)現(xiàn),有需要的朋友可以借鑒參考下,希望能夠有所幫助2024-02-02
教你怎么用PyCharm為同一服務(wù)器配置多個(gè)python解釋器
當(dāng)我們?cè)诜?wù)器上創(chuàng)建了多個(gè)虛擬環(huán)境時(shí),也可以在 PyCharm 中配置這些虛擬環(huán)境,方便不同的項(xiàng)目使用不同的環(huán)境,然而按照網(wǎng)上教程添加多個(gè)python解釋器后,PyCharm會(huì)自動(dòng)幫我們創(chuàng)建多個(gè)重復(fù)的服務(wù)器,本文主要給出該問(wèn)題的解決方法,同時(shí)也對(duì)添加解釋器做一個(gè)詳細(xì)的講解2021-05-05
python+requests接口壓力測(cè)試500次,查看響應(yīng)時(shí)間的實(shí)例
這篇文章主要介紹了python+requests接口壓力測(cè)試500次,查看響應(yīng)時(shí)間的實(shí)例,具有很好的參考價(jià)值,希望對(duì)大家有所幫助。一起跟隨小編過(guò)來(lái)看看吧2020-04-04
Python定時(shí)庫(kù)APScheduler的原理以及用法示例
APScheduler的全稱(chēng)是Advanced Python Scheduler,它是一個(gè)輕量級(jí)的 Python 定時(shí)任務(wù)調(diào)度框架,下面這篇文章主要給大家介紹了關(guān)于Python定時(shí)庫(kù)APScheduler的原理以及用法的相關(guān)資料,需要的朋友可以參考下2021-12-12
Python文件基本操作open函數(shù)應(yīng)用與示例詳解
這篇文章主要為大家介紹了Python文件基本操作open函數(shù)應(yīng)用與示例詳解,有需要的朋友可以借鑒參考下,希望能夠有所幫助,祝大家多多進(jìn)步,早日升職加薪2023-12-12
python利用winreg生成桌面路徑及實(shí)現(xiàn)掃描二維碼圖片返回相關(guān)信息
這篇文章主要介紹了python生成桌面路徑及實(shí)現(xiàn)掃描二維碼圖片返回相關(guān)信息,winreg是python的一個(gè)標(biāo)準(zhǔn)庫(kù),用來(lái)對(duì)windows注冊(cè)表的操作,更多相關(guān)內(nèi)容需要的小伙伴可以參考一下2022-06-06
Python虛擬環(huán)境virtualenv的安裝與使用詳解
virtualenv可以用來(lái)管理互不干擾的獨(dú)立python虛擬環(huán)境,在有些場(chǎng)景下非常有用,下面這篇文章主要給大家介紹了Python虛擬環(huán)境virtualenv安裝與使用的相關(guān)資料,需要的朋友可以參考借鑒,下面來(lái)一起看看吧。2017-05-05
pyspark 隨機(jī)森林的實(shí)現(xiàn)
這篇文章主要介紹了pyspark 隨機(jī)森林的實(shí)現(xiàn),文中通過(guò)示例代碼介紹的非常詳細(xì),對(duì)大家的學(xué)習(xí)或者工作具有一定的參考學(xué)習(xí)價(jià)值,需要的朋友們下面隨著小編來(lái)一起學(xué)習(xí)學(xué)習(xí)吧2020-04-04

