pdbr vs ipdb

[iloveitaly]

Are there any features missing from ipdb? This seems like a full replacement for ipdb (which is awesome). If that's the case, hopefully we can get the ipython documentation updated to include this package as an alternative.

[cansarigol]

Hi @iloveitaly I haven't checked it before but at least that issue is known. it would be great if pdbr was placed as an alternative.

[ssbarnea]

I used ipdb for quite some time and I just tried to use pdbr, hoping that I would be able to get some (rich) color when printing various data structures.

Still, first thing that I observe is that my breakpoints() no longer work. I did use the same method I used to enable ipdb, via export PYTHONBREAKPOINT=pdbr.set_trace.

I also observed that installing pip install pdbr[ipython] did downgrade ipython to 7.34 from 8.4.0,... not sure if that was really the best idea. I am using py311, so I am wondering if there are some problems between python-ipython-pdbr it that were not yet fixed yet.

[cansarigol]

Hi, actually there was a problem with a version of ipython but now it is working. I removed the specific version definition now and you can see v0.7.2.

For breakpoint, I couldn't reproduce it. Do you have any sample?

[ssbarnea]

Thanks, I updated both pdbr and ipython and gave it another try, but apparently pdbr did not want to debug. Later I created minimal python test file with just a breakpoint in it and found the culprit:

Traceback (most recent call last):
  File "/Users/ssbarnea/c/a/ansible-lint/z.py", line 3, in <module>
    breakpoint()
  File "/Users/ssbarnea/.pyenv/versions/3.11-dev/lib/python3.11/site-packages/pdbr/__main__.py", line 13, in set_trace
    pdb_cls = _pdbr_cls(context=context, show_layouts=show_layouts)
              ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/ssbarnea/.pyenv/versions/3.11-dev/lib/python3.11/site-packages/pdbr/utils.py", line 92, in _pdbr_cls
    return klass()
           ^^^^^^^
  File "/Users/ssbarnea/.pyenv/versions/3.11-dev/lib/python3.11/site-packages/pdbr/_pdbr.py", line 72, in __init__
    super().__init__(**init_kwargs)
  File "/Users/ssbarnea/.pyenv/versions/3.11-dev/lib/python3.11/site-packages/IPython/terminal/debugger.py", line 29, in __init__
    self.pt_init(pt_session_options)
  File "/Users/ssbarnea/.pyenv/versions/3.11-dev/lib/python3.11/site-packages/pdbr/_pdbr.py", line 86, in pt_init
    func_args = inspect.getargspec(super().pt_init).args
                ^^^^^^^^^^^^^^^^^^
AttributeError: module 'inspect' has no attribute 'getargspec'. Did you mean: 'getargs'?

That is not the first project that is hit by this, it is specific to py311. I think you should enable py311 pipeline as soon as possible, take a look at https://github.com/ansible/ansible-lint/blob/main/.github/workflows/tox.yml#L144 for how to make use of it in GHA.

[ssbarnea]

It seems that pdbr could use some love because while having an attempt to contribute a py311 change I faces too many other issues. It uses nox which is uncommon, installed it but when I run nox -e test-3.10, I seen that it failed with two tests on macos:

FAILED tests/test_magic.py::TestPdbrChildProcess::test_time - pexpect.exceptions.TIMEOUT: Timeout exceeded.
FAILED tests/test_magic.py::test_expr_questionmark_pinfo - AssertionError: untagged = 'Signature: foo(arg)\nDocstring: Foo docstring\nFile:      /private/var/folders/3...

Sadly I do not have time to fix all the bugs and look at the history of merged PRs, i see that:

• changes are merged while all windows pipelines report broken (for long time)
• there is no macos pipeline

If some of these fixes are restored I would not mind to contribute back few for py311 but for the moment, I am sticking to ipbr, which seems reliable. Yep, keeping CI/CD green takes a lot of time, sometimes more than anything else.

[cansarigol]

Hi, @ssbarnea thanks for having a look. I fixed the win test problems (it was related pexpect package, I tried winpexpect but it didn't work. So I excluded magic tests for windows,) and added macos tests in the action. Feel free to add your MR.

[cansarigol]

one more failed test still exists. It is related to win32 and python3.10. trying to fix it right now