jython-users

[Jython-users] Multiple Jython interpreters in different Java threads

[Barnabas W. <bar...@ms...>]

I'm trying to add Python scripting capabilities to an existing Java
application.  I would need to be able to execute scripts fully
independently in different threads.  I've attempted this by creating
an instance of PythonInterpreter in each of the threads.  This approach 
appears to work, but not perfectly.  For example, all interpreters 
share the same IO streams -- calling setOut(), setErr(), etc. on any 
of the interpreters will redirect *all* of the interpreters to the 
new stream.  This particular problem is not a complete show stopper, 
but it makes me worry about how independent the interpreters really 
are.  Have my spiciest behaved properly only by accident?  Are there 
any other volatile data structures shared by the interpreters?  Is 
there a recommended way to create independent interpreters that I've
overlooked?  (Aside from setting up independent classloaders in a 
sandbox for each thread, which I'd like to avoid)

Any help or suggestions would be appreciated...
Barnabas

Re: [Jython-users] Multiple Jython interpreters in different Java threads

[Mats W. <ma...@la...>]

At 12:09 PM 9/28/2001 -0400, Barnabas Wolf wrote:
 >I'm trying to add Python scripting capabilities to an existing Java
 >application.  I would need to be able to execute scripts fully
 >independently in different threads.  I've attempted this by creating
 >an instance of PythonInterpreter in each of the threads.  This approach
 >appears to work, but not perfectly.  For example, all interpreters
 >share the same IO streams -- calling setOut(), setErr(), etc. on any
 >of the interpreters will redirect *all* of the interpreters to the
 >new stream.  This particular problem is not a complete show stopper,
 >but it makes me worry about how independent the interpreters really
 >are.

That's the way threads are supposed to work:  separate
threads of control within a single execution context,
which means they have a lot of resources in common,
including open files.

Mats