You are viewing a plain text version of this content. The canonical link for it is here.
Posted to submarine-dev@hadoop.apache.org by "Liu Xun (Jira)" <ji...@apache.org> on 2019/10/12 06:04:00 UTC

[jira] [Created] (SUBMARINE-235) [Interpreter] Submarine Python Interpreter

Liu Xun created SUBMARINE-235:
---------------------------------

             Summary: [Interpreter] Submarine Python Interpreter
                 Key: SUBMARINE-235
                 URL: https://issues.apache.org/jira/browse/SUBMARINE-235
             Project: Hadoop Submarine
          Issue Type: Sub-task
          Components: Submarine Interpreter
            Reporter: Liu Xun
            Assignee: Liu Xun
             Fix For: 0.4.0


Submarine introduced the JAR package of Apache Zeppelin's Python interpreter in our submodule through POM import. 
Due to the differences between Zeppelin's way of calling the interpreter and the interface and Submarine, Submarine needs Zeppelin's The interpreter calls the interface class and performs a layer of encapsulation.
Zeppelin RemoteInterpreterServer.java Is the startup class of the interpreter process
Launch interpreter in Zeppelin,The main() method function in RemoteInterpreterServer is started from ZeppelinServer via the interpreter.sh script.
We need to develop our own startup class SubmarineInterpreterProcess instead of using Zeppelin's RemoteInterpreterServer class.
Zeppelin starts all of the more than twenty interpreters via the interpreter.sh script, making this startup script more complex, and ZeppelinServer launches  RemoteInterpreterServer.java via the interpreter.sh script. The parameter passing needs to be passed from JAVA to SHELL script and then passed to the JAVA execution process. The code is scattered in 3 places and it is difficult to maintain. So our startup method needs to be redesigned.

Design Doc: https://docs.google.com/document/d/1LJc5hRTWfIs6K5f7vMgGC_FUizeQf-qqiWICqO8SoR8/edit#



--
This message was sent by Atlassian Jira
(v8.3.4#803005)