RE: [Jython-users] precompiled code

[Humbel O. <Otm...@bi...>]

[ Prathibha ]

> I currently use python code(jython actually) which I
> retrieve from the DB to compile and interpret it at
> runtime.  Will using precompiled python scripts in the
> DB improve performance significantly? I am asking this
> as I am also looking to caching these precompiled
> scripts to improve performance as well.

That's exactly the way we do it:
 - the database contains the .py code (as clob)
 - we use a runtime cache containing precompiled code=20
   for the most used scripts

This improves performance considerably (I'm sorry I can't
give you a percentage - I simply don't know).

Below I try to sketch how we do the precompiling.=20
ATTN: This may not be state of the art.
I would be really glad if someone knows a better(*)
way to do it.

Server code:
  ByteArrayOutputStream outputStream =3D=20
	new ByteArrayOutputStream();
  ByteArrayInputStream sourceCode =3D=20
	new ByteArrayInputStream( scriptData );
  String uniqueID =3D "org.python.pycode._py" + scriptID;
  SimpleNode node =3D=20
	parser.parse( sourceCode, "exec", uniqueID, null );
  Module.compile( node, outputStream , name, filename,=20
			  true, false, false, null );
  byte[] byteCode =3D outputStream.toByteArray();

This byte array can be transferred from server to client.
On the client we make runnable code out of it:

  byte[] byteCode =3D getByteCodeFromServer( scriptID);
  String uniqueID =3D "org.python.pycode._py" + scriptID;
  PyCode currentCode =3D=20
	BytecodeLoader.makeCode( uniqueID, byteCode );

and run it as follows:

  PyObject result =3D Py.runCode( currentCode, locals, globals );

Hopefully this gives you some ideas.
I also would also appreciate it if someone knows a better(*)
way to precompile if everything takes place in the same JVM.

Best wishes, and thanks in advance.
Oti.

Legend:
-------
(*) better:	more upwards compatible,
			more performant
			more elegant
			....