[Jammy2211]

Summary

Point-source critical curves were rendering as 4 disconnected arcs (fit_image_with_points.png, on PyAutoEyes). Both VisualizerPoint.visualize and PlotterPoint's fallback computed them on Grid2D.from_extent(extent=dataset.extent_from()). That grid has two problems:

• Too small: it covered only the multiple images' bounding box plus 0.1". Critical curves often extend beyond the images, so the contour was clipped. When the box lies entirely inside the curve, there is no curve at all.
• Transposed: extent_from() returns [y_min, y_max, x_min, x_max], but from_extent takes (x0, x1, y0, y1).

The new critical_curve_grid_from(dataset) builds a square grid centred on the images, with half-width 2× their half-span (+0.1"), in x-first extent order. On the visualization dataset, the 4 clipped arcs are now one closed loop of radius 1.6".

The same grid is passed to the point visualizer's tracer and galaxy plots, so those now show the full critical curves too.

Weak lensing (added at the human's request): VisualizerWeak had the same transposed extent_from() → from_extent grid. It now uses the same helper with scale=1.0: a square grid covering the source-galaxy field, in (y, x) order. The helper lives in the shared autolens/lens/plot/critical_curves.py.

Closes #783.

API Changes

• New autolens.lens.plot.critical_curves.critical_curve_grid_from(dataset, shape_native=(100, 100), scale=2.0, buffer=0.1) (any dataset with extent_from).
• The point visualizer and plotter (scale 2) and the weak visualizer (scale 1) compute critical curves on this grid. PointDataset.extent_from() is unchanged.
  See full details below.

Test Plan

• pytest test_autolens/: 846 passed, 1 xfailed.
• New weak test: a field elongated in y and offset from the origin gets a grid centred on it in (y, x) that covers it. Control: the old grid was transposed.
• New test: with images inside a 1" box and an Einstein radius of 1.6", the grid yields one closed tangential line of radius 1.6. Control: the old grid returned no curve for this case.
• New test: the grid is centred on the images in (y, x) and is square, which catches any axis swap.
• Re-rendered both point domains in autolens_visualization: closed loop through the images.

Readiness

Heart readiness was YELLOW (score 85) at ship time:

• manifest drift: workspace checkouts (manifest ↔ disk) — 1 mismatch(es) vs PyAutoMind/repos.yaml
• release validation stale: source moved since rehearsal (PyAutoArray)

The human explicitly acknowledged this on 2026-10-21. Neither reason involves this change.

Full API Changes (for automation & release notes)

Added

• autolens.lens.plot.critical_curves.critical_curve_grid_from(dataset, shape_native=(100, 100), scale=2.0, buffer=0.1) -> Grid2D

Changed

• VisualizerPoint.visualize / PlotterPoint.fit_point critical-curve grid: was the images' bounding box with a transposed extent; it is now critical_curve_grid_from(dataset).
• VisualizerWeak.visualize tracer/galaxies grid: was a transposed extent_from() box; it is now critical_curve_grid_from(dataset, scale=1.0).

Migration

• None.

Generated by the PyAutoLabs agent workflow.

🤖 Generated with Claude Code

[Jammy2211]

Visualization PR: PyAutoLabs/autolens_visualization#10