fix: release build failures from plot API and autofit bugs

[Jammy2211]

Overview

The Apr 10 release build has 345 failures (up from 237 on Apr 9). The spike is almost entirely caused by simulator scripts crashing due to plot API changes, which cascades into ~220+ downstream failures. There are also genuine bugs in PyAutoFit's interpolator and grid search result, plus workspace scripts referencing the removed UltraNest search.

Fixing the library bugs (PyAutoArray zoom_array, PyAutoFit interpolator/grid_search) and updating workspace scripts should bring failures down dramatically.

Plan

• Fix zoom_array in PyAutoArray to handle plain numpy arrays that have a .mask attribute but no .is_all_false
• Fix PyAutoFit interpolator.__getitem__ to guard against empty instances list
• Fix PyAutoFit grid_search_result.log_evidences() to handle None log_evidence values
• Update all workspace simulator scripts using the removed aplt.Output API to use plot_array's current parameter signature
• Update autofit_workspace scripts that reference the removed af.UltraNest
• Re-trigger release build and verify failure count drops

Detailed implementation plan

Affected Repositories

• PyAutoArray (primary — zoom_array fix)
• PyAutoFit (interpolator + grid_search fixes)
• autolens_workspace (simulator scripts, guide scripts)
• autogalaxy_workspace (simulator scripts)
• autofit_workspace (UltraNest refs, graphical model dataset paths)

Work Classification

Both (library fixes first, then workspace fixes)

Branch Survey

Repository	Current Branch	Dirty?
PyAutoArray	main	minor untracked test files
PyAutoFit	main	clean
PyAutoGalaxy	main	minor untracked
PyAutoLens	main	clean
autolens_workspace	main	dataset changes (expected)
autofit_workspace	main	dataset changes (expected)
autogalaxy_workspace	main	dataset changes (expected)

Suggested branch: feature/release-fixes-apr-2026
Worktree root: ~/Code/PyAutoLabs-wt/release-fixes-apr-2026/

Implementation Steps

1. PyAutoArray — autoarray/plot/utils.py:108
   Change: hasattr(array, "mask") and not array.mask.is_all_false
   To: hasattr(array, "mask") and hasattr(array.mask, "is_all_false") and not array.mask.is_all_false

2. PyAutoFit — autofit/interpolator/abstract.py:98
   Guard self.instances[0] against empty list — raise informative error or return early

3. PyAutoFit — autofit/non_linear/grid/grid_search/result.py:306
   Handle None log_evidence values in the list comprehension (skip or replace with NaN)

4. autolens_workspace / autogalaxy_workspace simulator scripts
   Replace aplt.plot_array(array=X, output=aplt.Output(path=P, format=F))
   With aplt.plot_array(array=X, output_path=P, output_format=F)
   Affects ~14 active scripts

5. autofit_workspace scripts

   • Remove/comment UltraNest references in scripts/cookbooks/search.py and related
   • Fix graphical model tutorial dataset paths

Key Files

• autoarray/plot/utils.py — zoom_array type check fix
• autofit/interpolator/abstract.py — empty instances guard
• autofit/non_linear/grid/grid_search/result.py — None log_evidence handling
• Workspace simulator scripts across autolens/autogalaxy/autofit workspaces

Original Prompt

Click to expand starting prompt

Release Fix: 2026-04-12. Based on release run Apr 10 (run 24226814280). 345 failures, 808 passes, 90 skipped. Priority 1: Simulator Plot API Bugs (~200+ cascading failures) from zoom_array and aplt.Output removal. Priority 2: Source Code Bugs (60 failures) including UltraNest removal, interpolator IndexError, grid_search TypeError. Priority 3: Environment Issues (6, non-blocking).

[Jammy2211]

Library PRs Created — 2026-04-12

PRs

• fix: guard zoom_array against non-autoarray mask objects #270
• fix: guard interpolator and grid search against edge cases PyAutoFit#1201

Status

Workspace changes needed. Next: /start_workspace

API Changes

• PyAutoArray: internal fix to zoom_array — no public API change
• PyAutoFit: GridSearchResult.log_evidences() now handles None values; interpolator.__getitem__ gives clear error on empty instances

[Jammy2211]

All PRs Created — 2026-04-12

Library PRs

• fix: guard zoom_array against non-autoarray mask objects #270 — zoom_array guard
• fix: guard interpolator and grid search against edge cases PyAutoFit#1201 — interpolator + grid search fixes

Build & Workspace PRs

• fix: correct env var name and skip removed search packages PyAutoBuild#44 — fix env var name, add UltraNest/PySwarms to no_run
• fix: unset PYAUTO_SMALL_DATASETS for howtolens and guides autolens_workspace#52 — unset SMALL_DATASETS for howtolens/guides
• fix: unset PYAUTO_SMALL_DATASETS for howtogalaxy and guides autogalaxy_workspace#26 — unset SMALL_DATASETS for howtogalaxy/guides

Summary

Root causes identified:

1. zoom_array crashed on plain numpy arrays → ~200 cascading simulator failures
2. PYAUTO_WORKSPACE_SMALL_DATASETS in PyAutoBuild didn't match library's PYAUTO_SMALL_DATASETS
3. Scripts loading committed 100x100 FITS with 15x15 mask cap → shape mismatch
4. UltraNest/PySwarms scripts not in no_run.yaml after removal from PyAutoFit

Next

Merge all 5 PRs, then re-trigger the release build.

[Jammy2211]

Shipped — 2026-04-12

PRs

• fix: guard zoom_array against non-autoarray mask objects #270
• fix: guard interpolator and grid search against edge cases PyAutoFit#1201
• fix: correct env var name and skip removed search packages PyAutoBuild#44
• fix: unset PYAUTO_SMALL_DATASETS for howtolens and guides autolens_workspace#52
• fix: unset PYAUTO_SMALL_DATASETS for howtogalaxy and guides autogalaxy_workspace#26

Summary

All 5 PRs merged. Release build re-triggered (run 24312783694, skip_release=true) to verify fixes pass. Root causes: zoom_array crash on plain numpy arrays (~200 cascading failures), mismatched env var name in PyAutoBuild, missing SMALL_DATASETS overrides for tutorial/guide scripts, and removed search packages not in no_run.yaml.

Session Notes

• The env var mismatch (PYAUTO_WORKSPACE_SMALL_DATASETS vs PYAUTO_SMALL_DATASETS) was a subtle bug — the workspace-local env_vars.yaml files had the correct name, but PyAutoBuild's fallback config had the wrong one.
• The shape mismatch fix is a workaround (unsetting the env var for affected paths). The proper fix is migrating ~184 scripts to the should_simulate pattern, which would let SMALL_DATASETS work everywhere.
• ~37 remaining failures are already marked NEEDS_FIX/SLOW in workspace no_run.yaml files.