#8 bug in use with ipython

[Jan Strube]

Hi Koji,

I have written a library that automatically instantiates an IAnalysisFactory for interactive use.
When importing the module in ipython, it bails out. Running the import a second time works, so I assume it's a threading problem again.
Are threads actually used in batch mode ?
There's no gui that could block I/O.

Please see the attached script.
start ipython
from paidaUtilities import paida

You can see that importing works the second time around.
This is python2.4, on Scientific Linux 3. paida from cvs head.

[Jan Strube]

module that instantiates a couple of factories

[Koji Kishimoto]

Logged In: YES
user_id=734761
Originator: NO

Sorry for my late reply.

After some investigations, the problem is due to the deep part of current PAIDA architecture. As you pointed out, PAIDA uses (and has to use) "threading" module even in the batch mode and the problem is (partly) related to it.

This will be fixed PAIDA4 which is planned to use separate processes instead of threads...

[Jan Strube]

Logged In: YES
user_id=1366327
Originator: YES

Hi Koji,
Thanks for the reply. Very interesting approach to use separate processes rather than threads. How do they communicate?
I think threads would be good enough, if they only get started when actually invoking the plotter.
Anyways, it's not a big bug, so it doesn't bother me too much.

My Analysis is starting to move into the "hot phase", so PAIDA will get some more serious testing.
Keep rocking.

Cheers,
Jan

[Koji Kishimoto]

Logged In: YES
user_id=734761
Originator: NO

In PAIDA4, they will communicate with normal and wide used internet protocols like TCP/IP.
If we use PAIDA with another application that uses thread like ipython and wxPython, we always have some problems due to it. It is because the applications (including PAIDA) that use threading module tend to forget to pay attention to each other. This invokes the problems like main loop process inside GUI and timing treatment.

Anyway, this way will boost PAIDA to new level. Please continue to help PAIDA development.