[Jython-dev] Missing a method for resetting the PySystemState

[Ivan P. <ipo...@ma...>]

Hello,

We develop an open source project, using Jython as a script
interpretation tool. Our application uses a pool of PythonInterpreter
objects to facilitate multi-user work in web application.

After upgrading to Jython version 2.7 we faced the following critical
problem. In version 2.5, in order to create new PythonInterpreter each
time we needed one more we used the following:

    new PythonInterpreter(null, new PySystemState())

Due to changes in version 2.7 we had to change to

    new PythonInterpreter(null, Py.getSystemState())

as previous method failed to work properly (random packages were
reported as missing).

Now the problem is that each time we change a Jython script file it is
not enough to clear a pool of PythonInterpreters anymore to get things
work the new way. We need to restart the whole Java application because
PySystemState returned by Py.getSystemState() apparently caches the
previous version of compiled byte code. Restarting the application takes
up to several minutes, but we have to change the scripts 'on the fly'.

If we had a proper method of resetting PySystemState it would have
solved our problem. But we don't have one, e. g. see an open issue
http://bugs.jython.org/issue2053. Could you comment, if you are going to
solve the problem.  Maybe we could help you to solve it.

Regards,

Ivan Ponomarev, PhD


Course-IT LLC.