AttributeError: module 'matplotlib' has no attribute '__version_info__'

[scheemik]

I ran into the following error when importing matplotlib (using version 3.4.3) in a Jupyter Notebook:

AttributeError: module 'matplotlib' has no attribute '__version_info__'

TLDR

I was able to work around this issue by specifying matplotlib-inline to be "<0.2.0" (essentially fixing it to "0.1.7").

It seems like this is a specific version incompatibility between matplotlib and matplotlib-inline which is only an issue when using maptlotlib "<3.5". However, I wanted to document what I found in case anyone else runs into similar issues.

Steps to reproduce

Create an environment with Python 3.9, matplotlib version 3.4.3, and matplotlib-inline version "<=0.2.0" then use that environment to run the following code in a Jupyter Notebook:

import matplotlib.pyplot as plt

plt.show()

This results in an error. Here is the full output:

AttributeError                            Traceback (most recent call last)
Cell In[2], line 1
----> 1 import matplotlib.pyplot as plt
      3 plt.show()

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib/pyplot.py:2500
   2498     dict.__setitem__(rcParams, \"backend\", rcsetup._auto_backend_sentinel)
   2499 # Set up the backend.
-> 2500 switch_backend(rcParams[\"backend\"])
   2502 # Just to be safe.  Interactive mode can be turned on without
   2503 # calling `plt.ion()` so register it again here.
   2504 # This is safe because multiple calls to `install_repl_displayhook`
   2505 # are no-ops and the registered function respect `mpl.is_interactive()`
   2506 # to determine if they should trigger a draw.
   2507 install_repl_displayhook()

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib/pyplot.py:277, in switch_backend(newbackend)
    270 # Backends are implemented as modules, but \"inherit\" default method
    271 # implementations from backend_bases._Backend.  This is achieved by
    272 # creating a \"class\" that inherits from backend_bases._Backend and whose
    273 # body is filled with the module's globals.
    275 backend_name = cbook._backend_module_name(newbackend)
--> 277 class backend_mod(matplotlib.backend_bases._Backend):
    278     locals().update(vars(importlib.import_module(backend_name)))
    280 required_framework = _get_required_interactive_framework(backend_mod)

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib/pyplot.py:278, in switch_backend.<locals>.backend_mod()
    277 class backend_mod(matplotlib.backend_bases._Backend):
--> 278     locals().update(vars(importlib.import_module(backend_name)))

File ~/miniconda3/envs/to_delete0/lib/python3.9/importlib/__init__.py:127, in import_module(name, package)
    125             break
    126         level += 1
--> 127 return _bootstrap._gcd_import(name[level:], package, level)

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib_inline/__init__.py:1
----> 1 from . import backend_inline, config  # noqa
      3 __version__ = \"0.2.1\"
      5 # we can't ''.join(...) otherwise finding the version number at build time requires
      6 # import which introduces IPython and matplotlib at build time, and thus circular
      7 # dependencies.

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib_inline/backend_inline.py:236
    231                 ip.events.unregister(\"post_run_cell\", configure_once)
    233             ip.events.register(\"post_run_cell\", configure_once)
--> 236 _enable_matplotlib_integration()
    239 def _fetch_figure_metadata(fig):
    240     \"\"\"Get some metadata to help with displaying a figure.\"\"\"

File ~/miniconda3/envs/to_delete0/lib/python3.9/site-packages/matplotlib_inline/backend_inline.py:215, in _enable_matplotlib_integration()
    211 ip = get_ipython()
    213 import matplotlib
--> 215 if matplotlib.__version_info__ >= (3, 10):
    216     backend = matplotlib.get_backend(auto_select=False)
    217 else:

AttributeError: module 'matplotlib' has no attribute '__version_info__'"

The output indicates that the issue is in the backend_inline module of matplotlib_inline. In that file, it calls for matplotlib.__version_info__. However, in matplotlib version 3.4.3, that variable doesn't exist. The variable __version_info__ seems to be introduced in the next release, matplotlib version 3.5.

From testing environments with different versions of matplotlib_inline, I found that this error is due to the introduction of the call to matplotlib.__version_info__ in version 0.2.0 of matplotlib_inline. I took a look at the .toml file for this project, but didn't make a pull request because I'm unfamiliar with how it is handling version dependencies. Is there a way to specify that matplotlib_inline version ">=0.2.0" is incompatible with matplotlib version "<=3.4.3"?