tornado 数据库初始化

起因: 在使用tornado构建的web服务中，我们常常需要对数据库进行访问，如何数据连接才是最为友好的方式，我们一般写法可能是这样的

db.py

class DB(object):
    def __init__(self):
        self.mysql_db = MySQLDatabase(host=mysql_conf['host'], user=mysql_conf['username'], passwd=mysql_conf['password'], database='test')
        self.cache_redis = Redis(host=redis_conf['host'], port=redis_conf['port'], db=redis_conf['db'])     

然后

import tornado.ioloop
import tornado.web
from db import DB

db = DB()  # ***注意这里***
class MainHandler(tornado.web.RequestHandler):
    def get(self):
        db.mysql_db.excute()
        ...

application = tornado.web.Application([
    (r"/", MainHandler),
])

if __name__ == "__main__":
    print 'starting....'
    application.listen(8090)
    tornado.ioloop.IOLoop.instance().start()

这个写法一般情况下并不会出问题，因为tornado是基于epoll模型的，整个tornado是单线程的，逐个的处理每一个收到的Event，不会出现对数据库连接的并发访问，也就不存在线程安全问题

但是db作为一个模块中的变量暴露在外部显得非常突兀，一种更内聚的做法，可能是这样

from db import DB
class MainHandler(tornado.web.RequestHandler):
    def initialize(self):  
        self.db = DB() # ***留意这里***

    def get(self):
        self.db.mysql_db.excute()
        ...

application = tornado.web.Application([
    (r"/", MainHandler),
])

if __name__ == "__main__":
    print 'starting....'
    application.listen(8090)
    tornado.ioloop.IOLoop.instance().start()

initialize(self) 是 RequestHandler 中的预留方法，供子类覆盖，用于执行数据库等初始化动作, 只会在对应的handler被初始化时，调用有且一次

    def initialize(self):
        """Hook for subclass initialization.

        A dictionary passed as the third argument of a url spec will be
        supplied as keyword arguments to initialize().

        Example::

            class ProfileHandler(RequestHandler):
                def initialize(self, database):
                    self.database = database

                def get(self, username):
                    ...

            app = Application([
                (r'/user/(.*)', ProfileHandler, dict(database=database)),
                ])
        """
        pass

在上面的例子中，显然DB类被反复执行(有多个handler的情况下)，我们可以把db.py在改造下，把DB类改造成单例模式

def singleton(cls, *args, **kw):
    instances = {}

    def _singleton():
        key = str(cls) + str(os.getpid())
        if key not in instances:
            instances[key] = cls(*args, **kw)
        return instances[key]
    return _singleton

class DB(object):
    def __init__(self):
        self.mysql_db = MySQLDatabase(host=mysql_conf['host'], user=mysql_conf['username'], passwd=mysql_conf['password'], database='test')
        self.cache_redis = Redis(host=redis_conf['host'], port=redis_conf['port'], db=redis_conf['db'])  

当然也可以是用官方推荐的做法

# encoding=utf-8
# !/usr/bin/python
# std lib
import sys
import tornado.ioloop
import tornado.web

# our own libs
from db import DB


class MainHandler(tornado.web.RequestHandler):
    def initialize(self, db):
        '''
            覆盖父类的initialize方法
        '''
        self.db = db

    def get(self):
        # use self.db to access db
        pass

if __name__ == "__main__":
    print 'starting....'
    port = 8888
    db = DB()
    application = tornado.web.Application([
        (r"/", MainHandler, dict(db=db)),
    ])
    if len(sys.argv) > 1:
        port = int(sys.argv[1])
    application.listen(port)
    tornado.ioloop.IOLoop.instance().start()

参考资料：
http://www.tornadoweb.org/en/stable/web.html#tornado.web.RequestHandler.initialize