* Added more restrictions for using pandas for numpy_based frameworks:
* pandas uses numpy rec-arrays under the hood
* numpy doesn't work with arrays that are stored not on CPU
* as a result, for arrays/tensors that are stored on "gpu," "cuda," "mps," etc. we cannot use pandas
(cherry picked from commit 47c4f83242374448333d508e9f575472b05dc1cc)
IJ-CR-164759
GitOrigin-RevId: 5e83e920f6da55fe88bce0543c2e3b979b71702f
* Add PYTHON314_PREFIXES to CompatibilityVisitor
* Patch typeshed, add t-string related stubs
* Infer `string.templatelib.Template` type for t-strings instead of plain `str`
(cherry picked from commit 0e913910ab9e0dca4052856b0585ce66265291c0)
GitOrigin-RevId: f03a68ae9aee458a74eccc19c38d8d3e4dc7491a
1. Do MultiProcessDebugger disconnect method synchronized.
2. Remove `getProcessHandler().waitFor()` from `PyDebugProcess#detachDebuggedProcess` (deadlock).
3. Add `kwargs` for `pydev_monkey#create_warn_multiproc` (new args in Python 3.13).
4. Ignore `BrokenPipeError` in `NetCommand#send` (could happen after debugging stops).
5. Handle `FrameLocalsProxy` (a new class for frame locals in Python 3.13)
6. Handle exceptions in all callbacks functions (PEP669)
7. Handle new bytecode instruction `CALL_KW` (Python 3.13 smart stem into)
Merge-request: IJ-MR-159198
Merged-by: Egor Eliseev <Egor.Eliseev@jetbrains.com>
(cherry picked from commit bae3df0980c231aea2ea4518ac0eed42959c989d)
GitOrigin-RevId: 66f05ba56b7592e6206d0caa897f1e098ea94852
Do not attempt to access __class__ from a partially initialized module
Merge-request: IJ-MR-158047
Merged-by: Egor Eliseev <Egor.Eliseev@jetbrains.com>
(cherry picked from commit 0af9e6d3d6973625ec92a4792c73ab3edf8e861f)
GitOrigin-RevId: 175df3429efe7dd9bc4471df408ceb0b94feda52
We use bundled venv to create virtual envs because system python might not have one. However, if it does, it must be compatible with the one we bundle. A new version was released 31.mar.2025. When installed, it breaks our bundled one.
So, we bundled a version. It doesn't work with Py2 though, dus we now check language level.
(cherry picked from commit cc7ff994bf32ef4333908321c6bb189a0e9ca9e5)
IJ-CR-159120
GitOrigin-RevId: 5b40ec5f1e8537ca3337013453c8e6e86b606589
New Tensorflow Typeshed has updated stubs for the keras module, as well as for other Tensorflow modules. So PY-65966 should not be reproducible anymore
(cherry picked from commit 929e1da40c82bf6dd26efd227c0421087df3b75c)
IJ-MR-157295
GitOrigin-RevId: 4bf2396c3cd62aea0339db51fbec058f1d765a05
`use_readline` is no longer used by IPython,
by passing this argument we trigger a warning
from `traitlets` (IPython dependency)
this warning will soon be an error, removing it
should not change the behavior as it is ignored
since IPython 6 and we don't touch the code
supporting previous IPython versions
(cherry picked from commit 121038c4598093177690eb7dc3df16c7325de801)
IJ-MR-158254
GitOrigin-RevId: 869181cec6941d385f1eee79a6e604cd5a1d9e93
instead of disabling each test function/class
manually and using `try/except` we can utilize
`pytest.importorskip()`
(cherry picked from commit bf37db563ec1db5572c7d6aff688d3d83f2eeea8)
IJ-MR-158254
GitOrigin-RevId: 7436944669ef202afa658c98379f8bbb91e65fcc
- PyCharm still supports Python 2.7
- if you install latest `tox` on a modern Python
and run `tox -e py27` it will fail as under the
hood `tox` uses `virtualenv` version
incompatible with Python 2.7
- to prevent this from happening we need to
pin `virtualenv` version
- it works on CI without this change because for
each Python version we spawn a separate
Docker container and install `tox` inside so
the `tox` version always match the Python one
see the relevant `tox` docs https://tox.wiki/en/latest/faq.html#testing-end-of-life-python-versions
(cherry picked from commit f4d7ee543cf427f54788a03697d4a5abcd7af0cf)
IJ-MR-158254
GitOrigin-RevId: a29090bcf5a8343215f0a4cf7e515231c8bd5f62
* Changed "Missing" to "None Count" for clarity
* Fixed types for the __compute_sliced_data function
* Fixed a case where there is no visualization for the first column because of warnings
(cherry picked from commit 2e4f700fba4f6578c3f91eac9fa0756e94e71310)
GitOrigin-RevId: e31cd31bb3e718377e7a8957fb71ccd11d0c18a6
* Removed redundant check for the pandas version at the scripts' beginning
* Removed redundant check for the Python version when we specify the 'display.max_colwidth' option, as it only depends on the pandas version
(cherry picked from commit 04311fee7879b528fb3db9df025e17ca28647277)
GitOrigin-RevId: 2288ebde5622958e508e666b4d1bf7284b003564
