Fix bug in mixture logprob inference with None indices

[asifzubair]

Description

Previously, calculating the log-probability (logp) of an AdvancedSubTensor that included a None index would cause the system to crash with an unhandled internal AttributeError.

This change resolves the bug by adding a guard within the logp rewriting logic in mixture.py. Instead of crashing, the system now correctly identifies the ambiguous operation and raises a controlled RuntimeError.

A new test, test_advanced_subtensor_none_and_integer, has been added to test_mixture.py to validate this fix and prevent regressions.

Reproducible example from the ticket now executes:

>>> import numpy as np
... import pymc as pm
... 
... 
... obs = np.random.default_rng().normal(size=(7, 4))
... with pm.Model():
...    inds = np.arange(obs.shape[1])
...    a = pm.Normal("a", shape=10)
...    b = pm.Deterministic("b", a[None, inds])
...    c = pm.Normal("c", mu=b, sigma=1, observed=obs)
...    pm.sample()
...    
Initializing NUTS using jitter+adapt_diag...
Multiprocess sampling (3 chains in 3 jobs)
NUTS: [a]
                                                                                                                                  
  Progress                                   Draws   Divergences   Step size   Grad evals   Sampling Speed   Elapsed   Remaining  
 ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────── 
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   2000    0             0.871       3            792.47 draws/s   0:00:02   0:00:00    
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   2000    0             0.819       3            807.38 draws/s   0:00:02   0:00:00    
  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━   2000    0             0.951       3            783.45 draws/s   0:00:02   0:00:00    
                                                                                                                                  
Sampling 3 chains for 1_000 tune and 1_000 draw iterations (3_000 + 3_000 draws total) took 3 seconds.
We recommend running at least 4 chains for robust computation of convergence diagnostics
Inference data with groups:
	> posterior
	> sample_stats
	> observed_data

Related Issue

• Closes BUG: AdvancedSubTensor with None and integer indices raises a logprob error instead of silently failing #7762
• Related to #

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a relevant logical change

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pymc--7877.org.readthedocs.build/en/7877/

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 92.94%. Comparing base (58b49f2) to head (d25ce81).
⚠️ Report is 21 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7877   +/-   ##
=======================================
  Coverage   92.94%   92.94%           
=======================================
  Files         116      116           
  Lines       18845    18845           
=======================================
  Hits        17516    17516           
  Misses       1329     1329           

Files with missing lines	Coverage Δ
pymc/logprob/mixture.py	95.70% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[asifzubair]

Hi @ricardoV94 , Thank you for your comments. I've addressed them now

However, I did a rebase and push (as recommended) but it seems the rebase has added new commits on my PR ? Would you have a suggestion on how to handle this ? I can start a new PR in the worst case. Thank you 🙏

EDIT: I had to troubleshoot a bit, but now the PR looks to be in good state. Sorry for the alarm. :)

[ricardoV94]

Looks good just a small cleanup suggestion on the test

[ricardoV94]

Thanks @asifzubair