Repository navigation
list comprehensions don't see local variables in pdb in python3 #65360
Description
Activity
kay commented
on Apr 5, 2014 kaymannequinMannequinAuthorUsing generators in pdb are very handy but since Python3 they don't work properly. For example:
import pdb def foo(): items = [1, 2, 3] limit = 5 pdb.set_trace() foo()
in pdb prompt the following fails:
(pdf) all(x < limit for x in items)
*** NameError: global name 'items' is not definedI can express that in a lambda expression (i.e. pass items as an argument) but this seems unnecessary and very inelegant.
Your failure appears to be not pasted from an interpreter session; the actual failure is:
(Pdb) all(x < limit for x in items)
*** NameError: global name 'limit' is not defined(i.e. "limit" is not found, not "items"). This actually does not work in Python 2 either. What did work in Python 2 and doesn't work in 3, is using a list comprehension like
all([x < limit for x in items])
This is because list comprehensions are now implemented with their own function object like generator expressions have always been. To make the code in pdb compile "as if" it was put in the debugged function will be as good as impossible, so I'm closing this as won't fix.
JurkoGospodneti commented
on Apr 10, 2014 JurkoGospodnetimannequinMannequinJust ran into this problem and it's sooooo uncomfortable
researching dynamic structures at run-time using PDB
without this. :-(As a workaround, you can use a trick similar to this one
to 'import' your locals into the list comprehension body:[l'r' for l in (locals(),) for x in l['some_local']]
assuming 'r' & 'some_local' are two local variables in
your surrounding scope.Ugly, but at least it can be made/forced to work if needed...
Best regards,
Jurko Gospodnetićxdegaye commented
on Apr 12, 2014 xdegayemannequinMannequinFWIW this generator expression can be evaluated with the 'interact' command:
--Return-- > /test/comprehension.py(5)foo()->None -> pdb.set_trace() (Pdb) interact *interactive* >>> all(x < limit for x in items) True >>> (Pdb)
xdegaye commented
on Apr 12, 2014 xdegayemannequinMannequinThe runcode() method of InteractiveInterpreter in code.py uses the 'self.locals' dictionary as the 'globals' parameter of the invoked exec() function. And the do_interact() method of Pdb instantiates InteractiveInterpreter with 'locals' as a merge of the current frame's locals and globals dictionary. This explains why the interact command of pdb evaluates sucessfully the generator expression: the generator function object is evaluated by the interpreter in a frame where 'locals' is NULL (see fast_function() in ceval.c) and 'globals' includes now the debugged frame locals dictionary.
So a fix for this problem is to have the default() method of pdb be implemented in the same manner as do_interact() and the runcode() method of InteractiveInterpreter. The attached patch does this.
JurkoGospodneti commented
on Apr 12, 2014 JurkoGospodnetimannequinMannequinThanks for looking into this Xavier.
Could someone reopen this issue so it can gets looked at again?
xdegaye commented
on Apr 13, 2014 xdegayemannequinMannequinThe NameError exception occuring on a generator expression referencing a local variable when the generator is called within exec() is the object of multiple entries in the bug tracker, see bpo-13557.
msg 149096 in this issue suggests using exec(code, locals()) to fix the problem. It seems that what does currently the do_interact() method, and what is proposed in the patch is the recommended practice to handle this problem.
xdegaye commented
on Apr 14, 2014 xdegayemannequinMannequinInterestingly, the interact command was added at changeset:
$ hg log -v --pager cat -r $(hg blame Lib/pdb.py | grep do_interact | awk -F : '{print $1}') changeset: 66691:c2578a68879d user: Georg Brandl <georg@python.org> date: Sat Dec 04 11:20:26 2010 +0000 files: Doc/library/pdb.rst Lib/pdb.py Misc/NEWS description: Add the "interact" pdb command from pdb++.
And the included documentation, added at that time, states (emphasis added by me):
Start an interative interpreter (using the code module) whose global namespace contains all the ***(global and local)*** names found in the current scope.I can provide a test case for the patch when this issue is re-opened, unless someone else is willing to do it.
xdegaye commented
on May 20, 2014 xdegayemannequinMannequinThe patch fails to invoke exec() with the locals argument set to the current frame locals.
The attached patch fixes this, and test_pdb runs now fine with it.blueyed commented
on Aug 1, 2019 blueyedmannequinMannequinGeorg, please consider re-opening.
This patch here appears to fix this after all.will this be fixed at any point? its really a nuisance when debugging
This came up in https://discuss.python.org/t/inlining-comprehensions-what-are-the-compatibility-constraints/22527/25
Reopening.
Reacted by Matt Fisher and Tomasz SakrejdaFixed by #111094
Reacted by Carl Meyer and Tomasz Sakrejda
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
bugs.python.org fields: