如何在asyncio中重用套接字到服务器?而不是为每个查询创建新的连接?
这是我的代码;
async def lookup(server, port, query, sema):
async with sema as sema:
try:
reader, writer = await asyncio.open_connection(server, port)
except:
return {}
writer.write(query.encode("ISO-8859-1"))
await writer.drain()
data = b""
while True:
d = await reader.read(4096)
if not d:
break
data += d
writer.close()
data = data.decode("ISO-8859-1")
return data
答案 0 :(得分:2)
您只需调用一次asyncio.open_connection(server, port)
协程,并继续使用读取器和写入器(当然,前提是服务器不只是在其末端关闭连接)。
我会在一个单独的异步上下文管理器对象中为您的连接执行此操作,并使用connection pool来管理连接,以便您可以为许多并发任务创建和重用套接字连接。通过使用(异步)上下文管理器,Python确保在完成代码后通知连接,因此可以将连接释放回池中:
import asyncio
import contextlib
from collections import OrderedDict
from types import TracebackType
from typing import Any, List, Optional, Tuple, Type
try: # Python 3.7
base = contextlib.AbstractAsyncContextManager
except AttributeError:
base = object # type: ignore
Server = str
Port = int
Host = Tuple[Server, Port]
class ConnectionPool(base):
def __init__(
self,
max_connections: int = 1000,
loop: Optional[asyncio.AbstractEventLoop] = None,
):
self.max_connections = max_connections
self._loop = loop or asyncio.get_event_loop()
self._connections: OrderedDict[Host, List["Connection"]] = OrderedDict()
self._semaphore = asyncio.Semaphore(max_connections)
async def connect(self, server: Server, port: Port) -> "Connection":
host = (server, port)
# enforce the connection limit, releasing connections notifies
# the semaphore to release here
await self._semaphore.acquire()
connections = self._connections.setdefault(host, [])
# find an un-used connection for this host
connection = next((conn for conn in connections if not conn.in_use), None)
if connection is None:
# disconnect the least-recently-used un-used connection to make space
# for a new connection. There will be at least one.
for conns_per_host in reversed(self._connections.values()):
for conn in conns_per_host:
if not conn.in_use:
await conn.close()
break
reader, writer = await asyncio.open_connection(server, port)
connection = Connection(self, host, reader, writer)
connections.append(connection)
connection.in_use = True
# move current host to the front as most-recently used
self._connections.move_to_end(host, False)
return connection
async def close(self):
"""Close all connections"""
connections = [c for cs in self._connections.values() for c in cs]
self._connections = OrderedDict()
for connection in connections:
await connection.close()
def _remove(self, connection):
conns_for_host = self._connections.get(connection._host)
if not conns_for_host:
return
conns_for_host[:] = [c for c in conns_for_host if c != connection]
def _notify_release(self):
self._semaphore.release()
async def __aenter__(self) -> "ConnectionPool":
return self
async def __aexit__(
self,
exc_type: Optional[Type[BaseException]],
exc: Optional[BaseException],
tb: Optional[TracebackType],
) -> None:
await self.close()
def __del__(self) -> None:
connections = [repr(c) for cs in self._connections.values() for c in cs]
if not connections:
return
context = {
"pool": self,
"connections": connections,
"message": "Unclosed connection pool",
}
self._loop.call_exception_handler(context)
class Connection(base):
def __init__(
self,
pool: ConnectionPool,
host: Host,
reader: asyncio.StreamReader,
writer: asyncio.StreamWriter,
):
self._host = host
self._pool = pool
self._reader = reader
self._writer = writer
self._closed = False
self.in_use = False
def __repr__(self):
host = f"{self._host[0]}:{self._host[1]}"
return f"Connection<{host}>"
@property
def closed(self):
return self._closed
def release(self) -> None:
self.in_use = False
self._pool._notify_release()
async def close(self) -> None:
if self._closed:
return
self._closed = True
self._writer.close()
self._pool._remove(self)
try:
await self._writer.wait_closed()
except AttributeError: # wait_closed is new in 3.7
pass
def __getattr__(self, name: str) -> Any:
"""All unknown attributes are delegated to the reader and writer"""
if self._closed or not self.in_use:
raise ValueError("Can't use a closed or unacquired connection")
if hasattr(self._reader, name):
return getattr(self._reader, name)
return getattr(self._writer, name)
async def __aenter__(self) -> "Connection":
if self._closed or not self.in_use:
raise ValueError("Can't use a closed or unacquired connection")
return self
async def __aexit__(
self,
exc_type: Optional[Type[BaseException]],
exc: Optional[BaseException],
tb: Optional[TracebackType],
) -> None:
self.release()
def __del__(self) -> None:
if self._closed:
return
context = {"connection": self, "message": "Unclosed connection"}
self._pool._loop.call_exception_handler(context)
然后将一个池对象传递给您的查询协程;连接对象为读取器和写入器部分生成了代理:
async def lookup(pool, server, port, query):
try:
conn = await pool.connect(server, port)
except (ValueError, OSError):
return {}
async with conn:
conn.write(query.encode("ISO-8859-1"))
await conn.drain()
data = b""
while True:
d = await conn.read(4096)
if not d:
break
data += d
data = data.decode("ISO-8859-1")
return data
请注意,standard WHOIS protocol (RFC 3912 or predecessors指出在每次查询后连接都是关闭的。如果要通过端口43连接到标准WHOIS服务,则没有必要重新使用套接字。
在这种情况下,发生的情况是读者将达到EOF(reader.at_eof()
是正确的),并且任何进一步的读取尝试将仅不返回任何内容(reader.read(...)
将始终返回空的{{1 }}值)。直到套接字连接在超时后被远程端终止后,写入写入器才不会出错。您可以将所有所需内容写入连接,但是WHOIS服务器将仅忽略查询。
答案 1 :(得分:1)
您可以通过将读写器对存储到全局词典中来创建连接缓存。
# at top-level
connections = {}
然后在lookup
中,用首先检查字典的代码替换对open_connection
的调用:
if (server, port) in connections:
reader, writer = connections[server, port]
else:
reader, writer = await asyncio.open_connection(server, port)
connections[server, port] = reader, writer