Problems with pudb and IPython 0.12 about pudb

[inducer]

Hi Aaron, I'm not sure how pudb factors into any of this. Wrong bug tracker? :) If I missed something, feel free to reopen. Also, happy new year! Andreas

[asmeurer]

Sorry if it wasn't clear. I got this using pudb ("if I type ! to go to IPython..."). Or do you think it's an IPython bug?

I don't have the permissions to reopen issues.

[inducer]

Ah, never mind. Still sleepy apparently. Can you check if 14c076b fixes this? (Just adds a catch-and-ignore for this specific error.)

[asmeurer]

Well, yeah, it fixes the problem. I'm not so sure that it's the correct way, though...

Perhaps we should ask the @ipython guys to take a look.

[inducer]

Agree.d

[asmeurer]

I asked about it on their mailing list: http://mail.scipy.org/pipermail/ipython-dev/2012-January/008530.html.

[asmeurer]

See that discussion for the solution.

[inducer]

I'm not sure I fully understand what a solution might be. The following questions remain:

• What about backward compatibility? Can we detect at what version this changed, and transparently work with both the pre- and post-change versions? If so, how?
• Do they really want a module in user_module? I.e.

>>> from types import ModuleType
>>> x = ModuleType("yo")
>>> x.__dict__["zzz"] = 1
>>> ipy_thing.user_module = x

[takluyver]

Hi, I made the change in IPython - apologies for breaking the API. We did try to see what projects might have been affected, but I didn't know about PuDB.

• Detecting the change: It changed only for version 0.12. Three options spring to mind: 1) catch the error setting user_global_ns and set user_module instead, 2) check if the shell object has the user_module attribute, and 3) check the IPython version number.
• AFAIK, it doesn't strictly have to be a module - in certain situations, we set it to a custom object. Its __dict__ attribute is taken as the global namespace. Have a look at the code that initialises it: https://github.com/ipython/ipython/blob/master/IPython/core/interactiveshell.py#L973

[inducer]

Hi Thomas,

thanks for your input. I've implemented what I think is a fix in 07bfee0. Does that look reasonable to you? Aaron, can you please try this? Looking forward to hearing from both of you. Thanks.

Andreas

[takluyver]

I think that will work (though I haven't tried it).

[asmeurer]

Yes, it fixes it.

By the way, for some reason, pudb has gotten painfully slow all of a sudden. Every operation, from starting up to stepping one line, takes several seconds to several minutes to complete. It's nothing from PuDB, as the same commit in the git history that used to work fine has the problem. Any idea what could be going on? I don't think it's urwid related, as just moving around with the arrow keys works fine.

[asmeurer]

I'm closing the issue.

[inducer]

No idea about the slowness. If it persists, please open a new bug.