refactor: one shared colour-norm helper (fixes inversion ignoring log10_min_value)

[Jammy2211]

Summary

Closes #488.

Building the matplotlib colour norm from use_log10 / vmin / vmax was written out three times — plot/array.py, plot/inversion.py, and (in PyAutoGalaxy) util/plot_utils.py::norm_from. The copies had diverged, and one divergence was a live behaviour bug: plot_inversion_reconstruction hardcoded the 1e-4 log floor and never consulted autonerves config, so a user who changed visualize.general.general.log10_min_value had it honoured on array plots and silently ignored on inversion plots.

norm_from now lives once in autoarray/plot/utils.py and both plot functions call it. The config read lands as _conf_log10_min_value, alongside the other _conf_* readers. Deduplicating is how the divergence stops recurring; fixing the floor is the point.

The companion PyAutoGalaxy PR makes its norm_from a delegate. Library-first: this merges first.

API Changes

New public helper autoarray.plot.utils.norm_from(array=None, use_log10=False, vmin=None, vmax=None) — the single implementation behind every PyAuto colour scale. Nothing is removed and no existing signature changes; plot_array and plot_inversion_reconstruction are unchanged from a caller's point of view.

Two deliberate behaviour changes, both on plot_inversion_reconstruction only (plot_array and the autogalaxy GUI path are untouched):

1. The configured log10_min_value is now honoured, and values are clipped to it before vmax is derived. Inversion plots differ when the configured floor is not 1e-4. This is the fix the task exists for.
2. A degenerate vmax <= vmin is now widened however it arose, not only in the derived branch — LogNorm(vmin=10, vmax=1) is unusable whoever supplied the numbers. Inversion plots differ when a caller passes vmax <= vmin explicitly.

A third difference was deliberately not merged into one behaviour: array.py derived vmax from the image, inversion.py from pixel_values. The helper's array argument means "the values being coloured", so the two call sites legitimately pass different data. array=None takes the same fallback as an all-NaN array — exactly what inversion's separate else branch did — so nothing changed there. All three decisions are recorded in the helper's docstring, per the task's requirement that they be decided rather than assumed.

Test Plan

• test_autoarray/plot — 17 passed (11 new)
• test_autoarray full suite — 1157 passed, 55 skipped
• test_autogalaxy full suite against this branch — 1119 passed, 1 skipped
• test_autogalaxy/gui/test_plot_norm.py (pins PyAutoGalaxy#586's behaviour) — 9 passed, file untouched
• New tests verified to have teeth: reverting plot_inversion_reconstruction to the hardcoded floor makes test__plot_inversion_reconstruction fail with assert 0.0001 == 0.003

The new tests cover the configured floor at both call sites end-to-end (a spy wrapping the shared helper records the norm each plot function actually applied), plus explicit limits, derived vmax, clipping to the floor, the degenerate range, all-NaN values, and array=None.

Two caveats, stated rather than glossed:

• 8 inversion/inversion/test_{abstract,factory}.py sparse-operator tests fail in the authoring environment. They fail identically on unmodified main (verified by stashing) and are unrelated to plotting.
• Ran on Python 3.12 only — the authoring session had no 3.13 environment. CI covers 3.13.

Full API Changes (for automation & release notes)

Added

• autoarray.plot.utils.norm_from(array=None, use_log10=False, vmin=None, vmax=None) — returns a LogNorm, a Normalize, or None. The single colour-norm implementation; its docstring records the three divergence decisions.
• autoarray.plot.utils._conf_log10_min_value() — reads visualize.general.general.log10_min_value, falling back to 1.0e-4.

Removed

• Nothing. The inline colour-norm blocks in autoarray/plot/array.py and autoarray/plot/inversion.py are replaced by calls to the helper; from matplotlib.colors import LogNorm, Normalize drops out of inversion.py's module imports (both remain lazily imported inside the helper).

Migration

• None required for callers of plot_array / plot_inversion_reconstruction.
• Downstream code building its own colour norm should call autoarray.plot.utils.norm_from rather than reimplementing the block — that reimplementation is what this PR exists to stop.

Generated by the PyAutoLabs agent workflow.

---

Generated by Claude Code