class twisted.web.http.HTTPChannel(basic.LineReceiver, policies.TimeoutMixin): (source)
Known subclasses: twisted.web.proxy.Proxy, twisted.web.proxy.ReverseProxy
Implements interfaces: twisted.internet.interfaces.IConsumer, twisted.internet.interfaces.IPushProducer, twisted.internet.interfaces.ITransport
A receiver for HTTP requests.
The HTTPChannel provides interfaces.ITransport and interfaces.IConsumer to the Request objects it creates. It also implements interfaces.IPushProducer to self.transport, allowing the transport to pause it.
| Instance Variable | MAX_LENGTH | Maximum length for initial request line and each line from the header. |
| Instance Variable | maxHeaders | Maximum number of headers allowed per request. |
| Instance Variable | totalHeadersSize | Maximum bytes for request line plus all headers from the request. |
| Instance Variable | abortTimeout | The number of seconds to wait after we attempt to shut the transport down cleanly to give up and forcibly terminate it. This is only used when we time a connection out, to prevent errors causing the FD to get leaked. If this is None, we will wait forever. |
| Instance Variable | length | Undocumented |
| Instance Variable | persistent | Undocumented |
| Method | __init__ | Undocumented |
| Instance Variable | requests | Undocumented |
| Method | connectionMade | Called when a connection is made. |
| Method | lineReceived | Called for each line from request until the end of headers when it enters binary mode. |
| Method | headerReceived | Do pre-processing (for content-length) and store this header away. Enforce the per-request header limit. |
| Method | allContentReceived | Undocumented |
| Method | dataReceived | Data was received from the network. Process it. |
| Method | rawDataReceived | Override this for when raw data is received. |
| Method | allHeadersReceived | Undocumented |
| Method | checkPersistence | Check if the channel should close or not. |
| Method | requestDone | Called by first request in queue when it is done. |
| Method | timeoutConnection | Called when the connection times out. |
| Method | forceAbortClient | No summary |
| Method | connectionLost | Called when the connection is shut down. |
| Method | isSecure | Return True if this channel is using a secure transport. |
| Method | writeHeaders | Called by Request objects to write a complete set of HTTP headers to a transport. |
| Method | write | Called by Request objects to write response data. |
| Method | writeSequence | Write a list of strings to the HTTP response. |
| Method | getPeer | Get the remote address of this connection. |
| Method | getHost | Get the local address of this connection. |
| Method | loseConnection | Closes the connection. Will write any data that is pending to be sent on the network, but if this response has not yet been written to the network will not write anything. |
| Method | registerProducer | Register to receive data from a producer. |
| Method | unregisterProducer | Stop consuming data from a producer, without disconnecting. |
| Method | stopProducing | Stop producing data. |
| Method | pauseProducing | Pause producing data. |
| Method | resumeProducing | Resume producing data. |
| Instance Variable | _transferDecoder | None or a decoder instance if the request body uses the chunked Transfer-Encoding. |
| Instance Variable | _receivedHeaderSize | Bytes received so far for the header. |
| Instance Variable | _handlingRequest | Whether a request is currently being processed. |
| Instance Variable | _dataBuffer | Any data that has been received from the connection while processing an outstanding request. |
| Instance Variable | _networkProducer | Either the transport, if it provides interfaces.IPushProducer, or a null implementation of interfaces.IPushProducer. Used to attempt to prevent the transport from producing excess data when we're responding to a request. |
| Instance Variable | _requestProducer | If the Request object or anything it calls registers itself as an interfaces.IProducer, it will be stored here. This is used to create a producing pipeline: pause/resume producing methods will be propagated from the transport, through the HTTPChannel instance, to the c{_requestProducer}.The reason we proxy through the producing methods rather than the old behaviour (where we literally just set the |
| Instance Variable | _requestProducerStreaming | A boolean that tracks whether the producer on the Request side of this channel has registered itself as a interfaces.IPushProducer or an interfaces.IPullProducer. |
| Instance Variable | _waitingForTransport | A boolean that tracks whether the transport has asked us to stop producing. This is used to keep track of what we're waiting for: if the transport has asked us to stop producing then we don't want to unpause the transport until it asks us to produce again. |
| Instance Variable | _abortingCall | The twisted.internet.base.DelayedCall that will be used to forcibly close the transport if it doesn't close cleanly. |
| Instance Variable | _optimisticEagerReadSize | When a resource takes a long time to answer a request (via twisted.web.server.NOT_DONE_YET, hopefully one day by a Deferred), we would like to be able to let that resource know about the underlying transport disappearing as promptly as possible, via Request.notifyFinish, and therefore via self.requests[...].connectionLost() on this HTTPChannel.However, in order to simplify application logic, we implement head-of-line blocking, and do not relay pipelined requests to the application until the previous request has been answered. This means that said application cannot dispose of any entity-body that comes in from those subsequent requests, which may be arbitrarily large, and it may need to be buffered in memory. To implement this tradeoff between prompt notification when possible (in the most frequent case of non-pipelined requests) and correct behavior when not (say, if a client sends a very long-running GET request followed by a PUT request with a very large body) we will continue reading pipelined requests into
This behavior has been in place since Twisted 17.9.0 . |
| Instance Variable | __header | Undocumented |
| Instance Variable | __first_line | Undocumented |
| Class Variable | __content | Undocumented |
| Instance Variable | _savedTimeOut | Undocumented |
| Instance Variable | _receivedHeaderCount | Undocumented |
| Class Variable | _log | Undocumented |
| Instance Variable | _command | Undocumented |
| Instance Variable | _path | Undocumented |
| Instance Variable | _version | Undocumented |
| Method | _finishRequestBody | Undocumented |
| Method | _maybeChooseTransferDecoder | If the provided header is content-length or transfer-encoding, choose the appropriate decoder if any. |
| Method | _send100Continue | Sends a 100 Continue response, used to signal to clients that further processing will be performed. |
| Method | _respondToBadRequestAndDisconnect | This is a quick and dirty way of responding to bad requests. |
Inherited from LineReceiver:
| Class Variable | delimiter | The line-ending delimiter to use. By default this is b'\r\n'. |
| Instance Variable | line_mode | Undocumented |
| Method | clearLineBuffer | Clear buffered data. |
| Method | setLineMode | Sets the line-mode of this receiver. |
| Method | setRawMode | Sets the raw mode of this receiver. Further data received will be sent to rawDataReceived rather than lineReceived. |
| Method | sendLine | Sends a line to the other end of the connection. |
| Method | lineLengthExceeded | Called when the maximum line length has been reached. Override if it needs to be dealt with in some special way. |
| Instance Variable | _buffer | Undocumented |
| Instance Variable | _busyReceiving | Undocumented |
Inherited from Protocol (via LineReceiver):
| Class Variable | factory | Undocumented |
| Method | logPrefix | Return a prefix matching the class name, to identify log messages related to this protocol instance. |
Inherited from BaseProtocol (via LineReceiver, Protocol):
| Instance Variable | connected | Undocumented |
| Instance Variable | transport | Undocumented |
| Method | makeConnection | Make a connection to a transport and a server. |
Inherited from _PauseableMixin (via LineReceiver):
| Instance Variable | paused | Undocumented |
Inherited from TimeoutMixin:
| Instance Variable | timeOut | The number of seconds after which to timeout the connection. |
| Method | callLater | Wrapper around reactor.callLater for test purpose. |
| Method | resetTimeout | Reset the timeout count down. |
| Method | setTimeout | Change the timeout period |
| Instance Variable | __timeoutCall | Undocumented |
| Method | __timedOut | Undocumented |
None or a decoder instance if the request body uses the chunked Transfer-Encoding._ChunkedTransferDecoder)
int)
interfaces.IPushProducer, or a null implementation of interfaces.IPushProducer. Used to attempt to prevent the transport from producing excess data when we're responding to a request.interfaces.IPushProducer)
Request object or anything it calls registers itself as an interfaces.IProducer, it will be stored here. This is used to create a producing pipeline: pause/resume producing methods will be propagated from the transport, through the HTTPChannel instance, to the c{_requestProducer}.The reason we proxy through the producing methods rather than the old behaviour (where we literally just set the Request object as the producer on the transport) is because we want to be able to exert backpressure on the client to prevent it from sending in arbitrarily many requests without ever reading responses. Essentially, if the client never reads our responses we will eventually stop reading its requests.
interfaces.IPushProducer)
Request side of this channel has registered itself as a interfaces.IPushProducer or an interfaces.IPullProducer.bool or None)
bool)
twisted.internet.base.DelayedCall that will be used to forcibly close the transport if it doesn't close cleanly.twisted.internet.base.DelayedCall)
twisted.web.server.NOT_DONE_YET, hopefully one day by a Deferred), we would like to be able to let that resource know about the underlying transport disappearing as promptly as possible, via Request.notifyFinish, and therefore via self.requests[...].connectionLost() on this HTTPChannel.However, in order to simplify application logic, we implement head-of-line blocking, and do not relay pipelined requests to the application until the previous request has been answered. This means that said application cannot dispose of any entity-body that comes in from those subsequent requests, which may be arbitrarily large, and it may need to be buffered in memory.
To implement this tradeoff between prompt notification when possible (in the most frequent case of non-pipelined requests) and correct behavior when not (say, if a client sends a very long-running GET request followed by a PUT request with a very large body) we will continue reading pipelined requests into self._dataBuffer up to a given limit.
_optimisticEagerReadSize is the number of bytes we will accept from the client and buffer before pausing the transport.
This behavior has been in place since Twisted 17.9.0 .
int)
Called when a connection is made.
This may be considered the initializer of the protocol, because it is called when the connection is completed. For clients, this is called once the connection to the server has been established; for servers, this is called after an accept() call stops blocking and a socket has been received. If you need to send any greeting or initial message, do it here.
Called for each line from request until the end of headers when it enters binary mode.
Do pre-processing (for content-length) and store this header away. Enforce the per-request header limit.
| Parameters | line | A line from the header section of a request, excluding the line delimiter. (type: bytes) |
| Returns | A flag indicating whether the header was valid. (type: bool) | |
Check if the channel should close or not.
| Parameters | request | The request most recently received over this channel against which checks will be made to determine if this connection can remain open after a matching response is returned. |
| version | The version of the request. (type: bytes) | |
| Returns | A flag which, if True, indicates that this connection may remain open to receive another request; if False, the connection must be closed in order to indicate the completion of the response to request. (type: bool) | |
Called when the connection times out.
Override to define behavior other than dropping the connection.
Called if abortTimeout seconds have passed since the timeout fired, and the connection still hasn't gone away. This can really only happen on extremely bad connections or when clients are maliciously attempting to keep connections open.
Called when the connection is shut down.
Clear any circular references here, and any external references to this Protocol. The connection has been closed.
| Parameters | reason | Undocumented (type: twisted.python.failure.Failure) |
Return True if this channel is using a secure transport.
Normally this method returns True if this instance is using a transport that implements interfaces.ISSLTransport.
| Returns | True if this request is secure (type: bool) | |
Called by Request objects to write a complete set of HTTP headers to a transport.
| Parameters | version | The HTTP version in use. (type: bytes) |
| code | The HTTP status code to write. (type: bytes) | |
| reason | The HTTP reason phrase to write. (type: bytes) | |
| headers | The headers to write to the transport. (type: twisted.web.http_headers.Headers) |
Closes the connection. Will write any data that is pending to be sent on the network, but if this response has not yet been written to the network will not write anything.
| Returns | None | |
Register to receive data from a producer.
This sets self to be a consumer for a producer. When this object runs out of data (as when a send(2) call on a socket succeeds in moving the last data from a userspace buffer into a kernelspace buffer), it will ask the producer to resumeProducing().
For IPullProducer providers, resumeProducing will be called once each time data is required.
For IPushProducer providers, pauseProducing will be called whenever the write buffer fills up and resumeProducing will only be called when it empties.
| Parameters | producer | The IProducer that will be producing data. (type: IProducer provider) |
| streaming | True if producer provides IPushProducer, False if producer provides IPullProducer. (type: bool) | |
| Returns | None | |
| Raises | RuntimeError | If a producer is already registered. |
Stop consuming data from a producer, without disconnecting.
| Returns | None | |
Stop producing data.
The HTTPChannel doesn't *actually* implement this, beacuse the assumption is that it will only be called just before loseConnection is called. There's nothing sensible we can do other than call loseConnection anyway.
Pause producing data.
This will be called by the transport when the send buffers have been filled up. We want to simultaneously pause the producing Request object and also pause our transport.
The logic behind pausing the transport is specifically to avoid issues like https://twistedmatrix.com/trac/ticket/8868. In this case, our inability to send does not prevent us handling more requests, which means we increasingly queue up more responses in our send buffer without end. The easiest way to handle this is to ensure that if we are unable to send our responses, we will not read further data from the connection until the client pulls some data out. This is a bit of a blunt instrument, but it's ok.
Note that this potentially interacts with timeout handling in a positive way. Once the transport is paused the client may run into a timeout which will cause us to tear the connection down. That's a good thing!
Resume producing data.
This will be called by the transport when the send buffer has dropped enough to actually send more data. When this happens we can unpause any outstanding Request producers we have, and also unpause our transport.
Sends a 100 Continue response, used to signal to clients that further processing will be performed.
This is a quick and dirty way of responding to bad requests.
As described by HTTP standard we should be patient and accept the whole request from the client before sending a polite bad request response, even in the case when clients send tons of data.