fix: VarLookupDict __iter__ + DataFrame dtype compat (fixes #259, #232)

[kouyouqi123]

Fixes #259 — VarLookupDict missing __iter__

Added __iter__, __len__, and keys() to VarLookupDict to prevent CPython crash during traceback formatting.

Fixes #232 — DataFrame .dtype access in _eval_factor

Changed result.dtype to np.asarray(result).dtype so the error message works with both numpy arrays and pandas DataFrames.

[bashtage]

This is the type of change that is considered hard in patsy. Could you explain more as to why this is experienced when it does not come up in standard CPython?

[bashtage]

Also, does this happen in standard CPython, only not get tested in patsy?

[codecov-commenter]

Codecov Report

❌ Patch coverage is 27.27273% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 98.20%. Comparing base (331879f) to head (128b51d).
⚠️ Report is 60 commits behind head on master.

Files with missing lines	Patch %	Lines
patsy/eval.py	27.27%	8 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #260      +/-   ##
==========================================
+ Coverage   97.60%   98.20%   +0.59%     
==========================================
  Files          30       30              
  Lines        3096     3114      +18     
  Branches      591      679      +88     
==========================================
+ Hits         3022     3058      +36     
+ Misses         39       35       -4     
+ Partials       35       21      -14     

Flag	Coverage Δ
unittests	97.81% <27.27%> (+0.20%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[bashtage]

I think I understand this issues now. Shouldn't the fix either go into the notebook environment or into CPython directly? These appear to be fully upstream of patsy.

The hesitation is around the general treatment of patsy in being in maintenance mode. This has been interpreted to mean fixes for change in CPython where essential for it to run.

This said, statsmodels need to get out the version where formulaic is available.

[bashtage]

For this to have a realistic prospect (given the status of the bug and the maintenance mode of patsy), we will need 100% coverage. Ideally you could add a single test that triggers the error using the same API that caused this issue in the first place, and we will see that everything that has been added will be hit.