Properly handle missing attributes in query/eval strings

[alexmojaki]

Consider this script:

import pandas as pd

pd.eval("pd.thing")

Currently it raises an error like this:

  File "/home/alex/work/pandas/pandas/core/computation/expr.py", line 640, in visit_Attribute
    raise ValueError(f"Invalid Attribute context {ctx.__name__}")
AttributeError: 'Load' object has no attribute '__name__'

Adding __class__ to that line changes the error to the more sensible:

ValueError: Invalid Attribute context Load

Re-raising the original error gives what's really needed:

  File "/home/alex/work/pandas/pandas/core/computation/expr.py", line 631, in visit_Attribute
    v = getattr(resolved, attr)
  File "/home/alex/work/pandas/pandas/__init__.py", line 260, in __getattr__
    raise AttributeError(f"module 'pandas' has no attribute '{name}'")
AttributeError: module 'pandas' has no attribute 'thing'

[alexmojaki]

I don't know what the failure "codecov/patch — 0% of diff hit (target 50%)" means. Am I supposed to write tests for this change?

[jreback]

always start with a test

[jreback]

can you add a whatsnew note for 1.1. in the conversion bug fix section

[jreback]

can you add a whatsnew note for 1.1, in the Other bug fix section

[jreback]

thanks @alexmojaki will merge on green.

[alexmojaki]

Presumably that docs failure is for unrelated reasons?

no theme named 'pandas_sphinx_theme' found (missing theme.conf?)

[simonjayhawkins]

Presumably that docs failure is for unrelated reasons?

no theme named 'pandas_sphinx_theme' found (missing theme.conf?)

I think fixed by #32840. can you merge master

[alexmojaki]

1 failed test in 1 job in 1 build: numpy array values are different (2.0 %)

Again, unrelated.

[WillAyd]

Can you merge master again and re-push? Should be fixed on master

[alexmojaki]

The last build succeeded 5 hours ago, but github hasn't been updated. I've never seen that before.

[WillAyd]

Thanks @alexmojaki