Status

Current State: Implemented
Authors: qingfeng(wlliqipeng),dinglei
Shepherds: zhengdong
Mailing List discussion: users@rocketmq.apache.org;dev@rocketmq.apache.org
Pull Request:https://github.com/apache/rocketmq-client-python/pull/1
Released: <4.3.2>

Background & Motivation

What do we need to do

 Now, applications developed by python can not send or consume messages through RocketMQ. So we want to provide new client which can be used by  that applications.We will add a new module  called rocketmq-client-python.

Goals

What problem is this proposal designed to solve?

   We plan to provide the same functions as the Python client. For example, producers can be sent in two ways, synchronous and asynchronous. For consumers, Python clients also offer two modes of push and pull.We are also concerned about availability, reliability and delay.

To what degree should we solve the problem?

Support all rocketmq features，such as broadcast/cluster model, concurrency/orderly publish/subscribe, timed/delay msg, consumer status query and so on.
Support across platform,and all features are supported on both windows and linux system.
Low latency,publish latency < 2ms, subscribe latency < 10ms .
Fault recovery capability.Based on nameServer snapshot and network disaster recovery strategy, no real-time impact on publish/subscribe when anyone of broker or nameSrv was broken.

Non-Goals

What problem is this proposal NOT designed to solve?

Rocketmq-client-Python does not support other protocols except rocketmq protocol, such as JMS and AMQP. Are there any limits of this proposal? Rocketmq-client-python needs to rely on boost python, rocketmq cpp client library(librocketmq.so) and so.

Changes

We will provide a new module named rocketmq-client-python. Read below sections to get more details about the python client for RocketMQ

Architecture

 Rocketmq-client-python  is based on the encapsulation of the C interface , which is provided by librocketmq.so or rocketmq.a . so it is logically divided into two  layers: API layer encapsulated by boost python, and C library which is alse divided into C API layer ,message layer, protocol layer and transport layer.

Interface Design

1 Python Interface

BOOST_PYTHON_MODULE (librocketmqclientpython) {
/*
   class_<CMessage>("CMessage");
   class_<CMessageExt>("CMessageExt");
   class_<CProducer>("CProducer");
   class_<CPushConsumer>("CPushConsumer");
*/
   enum_<CStatus>("CStatus")
           .value("OK", OK)
           .value("NULL_POINTER", NULL_POINTER);

   enum_<CSendStatus>("CSendStatus")
           .value("E_SEND_OK", E_SEND_OK)
           .value("E_SEND_FLUSH_DISK_TIMEOUT", E_SEND_FLUSH_DISK_TIMEOUT)
           .value("E_SEND_FLUSH_SLAVE_TIMEOUT", E_SEND_FLUSH_SLAVE_TIMEOUT)
           .value("E_SEND_SLAVE_NOT_AVAILABLE", E_SEND_SLAVE_NOT_AVAILABLE);

   enum_<CConsumeStatus>("CConsumeStatus")
           .value("E_CONSUME_SUCCESS", E_CONSUME_SUCCESS)
           .value("E_RECONSUME_LATER", E_RECONSUME_LATER);

   class_<PySendResult>("SendResult")
           .def_readonly("offset", &PySendResult::offset, "offset")
                   //.def_readonly("msgId", &PySendResult::msgId, "msgId")
           .def_readonly("sendStatus", &PySendResult::sendStatus, "sendStatus")
           .def("GetMsgId", &PySendResult::GetMsgId);
   class_<PyMessageExt>("CMessageExt");

   //For Message
   def("CreateMessage", PyCreateMessage, return_value_policy<return_opaque_pointer>());
   def("DestroyMessage", PyDestroyMessage);
   def("SetMessageTopic", PySetMessageTopic);
   def("SetMessageTags", PySetMessageTags);
   def("SetMessageKeys", PySetMessageKeys);
   def("SetMessageBody", PySetMessageBody);
   def("SetByteMessageBody", PySetByteMessageBody);
   def("SetMessageProperty", PySetMessageProperty);

   //For MessageExt
   def("GetMessageTopic", PyGetMessageTopic);
   def("GetMessageTags", PyGetMessageTags);
   def("GetMessageKeys", PyGetMessageKeys);
   def("GetMessageBody", PyGetMessageBody);
   def("GetMessageProperty", PyGetMessageProperty);
   def("GetMessageId", PyGetMessageId);

   //For producer
   def("CreateProducer", PyCreateProducer, return_value_policy<return_opaque_pointer>());
   def("DestroyProducer", PyDestroyProducer);
   def("StartProducer", PyStartProducer);
   def("ShutdownProducer", PyShutdownProducer);
   def("SetProducerNameServerAddress", PySetProducerNameServerAddress);
   def("SendMessageSync", PySendMessageSync);

   //For Consumer
   def("CreatePushConsumer", PyCreatePushConsumer, return_value_policy<return_opaque_pointer>());
   def("DestroyPushConsumer", PyDestroyPushConsumer);
   def("StartPushConsumer", PyStartPushConsumer);
   def("ShutdownPushConsumer", PyShutdownPushConsumer);
   def("SetPushConsumerNameServerAddress", PySetPushConsumerNameServerAddress);
   def("SetPushConsumerThreadCount", PySetPushConsumerThreadCount);
   def("SetPushConsumerMessageBatchMaxSize", PySetPushConsumerMessageBatchMaxSize);
   def("Subscribe", PySubscribe);
   def("RegisterMessageCallback", PyRegisterMessageCallback);

   //For Version
   def("GetVersion", PyGetVersion);
}

Compatibility, Deprecation, and Migration Plan

Are backward and forward compatibility taken into consideration? Yes，all interfaces are backward compatible Are there deprecated APIs? No, all APIs are new.

How do we do migration?

All the interfaces are newly developed, and new access applications are directly used without migration problems. If your system is currently using other python clients, you need to modify the code according to the current specifications.

Implementation Outline

We will implement the proposed changes by 2 phases.
API layer, message layer, protocol layer and transport layer
Phase Development of basic future
Implementation of sending message
Implementation of Consuming message by push model
Phase Development of other futures
Implementation of orderly message and transaction message and so.