feat: tangential/cross shear profile + Kaiser-Squires map (weak series step 6)

[Jammy2211]

Overview

Step 6 of the weak-lensing series (PyAutoMind/z_features/weak_shear.md): add the two canonical cluster weak-lensing diagnostics missing from autolens/weak/plot/ — the azimuthally averaged tangential/cross shear radial profile (THE standard observable in cluster weak lensing; the cross component is the B-mode null test) and a Kaiser-Squires convergence-map reconstruction. Science anchors: Oguri et al. 2012 (SGAS), Medezinski et al. 2016 (A2744). Independent of the other queued series steps; step 7 (real-data A2744) will reuse both plotters.

Plan

• Add shear-profile math + plotters to autolens/weak/plot/: shear_tangential_cross_from(shear_yx, centre) and plot_shear_profile (binned γ_t/γ_x vs radius with error bars).
• Add convergence_via_kaiser_squires_from (grid the irregular shear field, FFT KS inversion) + plot_convergence_map.
• Re-export both plotters into aplt alongside the existing nine weak helpers.
• NumPy-only unit tests against the SIS analytic profile γ_t(r) = θ_E/2r, γ_x = 0.
• Workspace follow-up: demo both in scripts/weak/fit.py (profile of data vs model) and simulator.py (KS map), navigator catalogue regeneration if summaries change.

Detailed implementation plan

Affected Repositories

• PyAutoLens (primary — library)
• autolens_workspace (follow-up — extend scripts/weak/{simulator,fit}.py demos)

Branch Survey

Repository	Current Branch	Dirty?
./PyAutoLens	main (post #580)	clean
./autolens_workspace	main (post #241)	pre-existing regenerated dataset files only

Suggested branch: feature/weak-viz-profiles
Worktree: ~/Code/PyAutoLabs-wt/weak-viz-profiles/
Note: autolens_workspace remains claimed by lenstool-example; the workspace demo uses the parallel-worktree pattern (fourth use; scripts/weak/ has zero overlap).

Implementation Steps

1. autolens/weak/plot/shear_profile_plots.py:
   • shear_tangential_cross_from(shear_yx, centre=(0.0, 0.0)) → (gamma_t, gamma_x, radii) per galaxy. Use ONLY the public accessors (.ellipticities for |γ|, .phis for the shear angle in degrees, .grid for positions) — never index the raw [γ₂, γ₁] storage (encapsulation rule from series step 2). γ_t = −Re[γ e^{−2iφ}], γ_x = −Im[γ e^{−2iφ}] with φ the position angle about centre.
   • plot_shear_profile(obj, centre=, bins=, ax=None, output_*=) accepting a WeakDataset or FitWeak (plot data profile; if a fit, overlay the model-shear profile). Radial bins: plain mean + standard error per bin; γ_x series expected consistent with zero.
2. autolens/weak/plot/convergence_plots.py:
   • convergence_via_kaiser_squires_from(shear_yx, shape_native=(50, 50), smoothing_sigma_pixels=1.0) — bin the irregular (γ1, γ2) onto a regular grid over extent_from(), FFT inversion κ̂ = [(k1²−k2²)γ̂1 + 2k1k2 γ̂2]/k², optional Gaussian smoothing, return aa.Array2D.
   • plot_convergence_map(...) imshow-style with galaxy positions overlaid.
3. Re-export both plot functions in autolens/plot/__init__.py (weak block, lines ~75-87).
4. Tests test_autolens/weak/test_shear_profile.py + test_convergence.py (NumPy-only):
   • Noise-free SimulatorShearYX SIS dataset: binned γ_t matches θ_E/2r within bin tolerance; γ_x ≈ 0 (atol ~1e-10 noise-free).
   • KS map: peak within one grid pixel of the lens centre; κ positive at centre; B-mode (imaginary residual) negligible.
   • Direct module imports in tests (the al.plot pytest recursion gotcha).
5. Full test_autolens/ suite.

Workspace follow-up (second PR, library-first gate)

1. scripts/weak/fit.py: add a __Shear Profile__ section plotting data-vs-model γ_t/γ_x profiles.
2. scripts/weak/simulator.py: add a __Convergence Map__ section with the KS reconstruction.
3. Regenerate navigator catalogue (PyAutoBuild/autobuild/regenerate_navigator.py autolens) — Contents lines change.

Key Files

• autolens/weak/plot/{weak_dataset_plots,fit_weak_plots}.py — existing helpers + save/subplot conventions to match.
• autogalaxy/util/shear_field.py — ShearYX2DIrregular public accessors.
• test_autolens/weak/test_fit.py — SIS fixture pattern to reuse.

Autonomy

--auto continuation (launch 2026-07-09), effective level supervised (header supervised, feature cap supervised): plan-to-issue, mechanical stretches proceed, ship sign-off parks as a batched question here, merge stays human. Brain Feature Agent failed to classify this prompt (no repos resolved, suggested research); classification emulated inline per WORKFLOW.md — recorded for post-hoc review.

Original Prompt

Click to expand starting prompt

Weak lensing shear profile and convergence map visualization. Add the two canonical cluster weak-lensing diagnostics missing from autolens/weak/plot: (1) an azimuthally averaged tangential and cross shear radial profile helper (binned gamma_t / gamma_x vs radius with error bars, cross-shear as the standard B-mode null test) computed about a chosen centre from a WeakDataset or FitWeak; (2) a Kaiser-Squires style convergence map reconstruction plotted from the shear field. Re-export into aplt alongside the existing nine quiver helpers and demo both in autolens_workspace/scripts/weak (extend simulator.py / fit.py outputs). Tangential shear profiles are THE standard observable in cluster weak lensing (Oguri 2012 SGAS; Medezinski 2016 A2744).

[Jammy2211]

Scope addition (user direction, 2026-07-09): the workspace half of this task also adopts the standard al.util.dataset.should_simulate auto-simulation pattern (the imaging-style __Dataset Auto-Simulation__ section, pointed at scripts/weak/simulator.py) in scripts/weak/fit.py and migrates scripts/weak/modeling.py's plain exists() check to it. Prompt 9 (PYAUTO_SMALL_DATASETS) updated to reference this rather than duplicate it.

[Jammy2211]

Ship sign-off requested (supervised checkpoint — run parked here)

What was done. Both halves implemented, validated and committed locally on feature/weak-viz-profiles (nothing pushed). Library commit b508ffa8: shear_profile_plots.py (per-galaxy γ_t/γ_x + binned profiles + plot_shear_profile with model overlay for fits) and convergence_plots.py (Kaiser-Squires E/B-mode inversion + plot_convergence_map), both re-exported into aplt, plus 6 new unit tests. Workspace commit d6090b72: __Shear Profile__ section in fit.py, __Convergence Map__ section in simulator.py, the should_simulate auto-simulation adoption in fit.py + modeling.py (user-directed scope addition), navigator catalogue regenerated in-branch (so the staleness CI leg passes first time, unlike PR #241).

Validation.

• PyAutoLens full suite: 352 passed (38 weak, 6 new).
• Convention checks: SIS field is pure-tangential (γ_t = |γ| to 1e-6, γ_x = 0); binned γ_t matches the SIS analytic θ_E/2r within bin discretisation; KS map peaks at the lens centre with E ≫ B.
• Axis-parity check (the SIS tests are rotation-invariant and blind to mirror errors): a q=0.5 lens at 45° reconstructs with KS major axis 47.1° — added as a regression test.
• End-to-end: simulator.py and fit.py ran in the worktree; figures attached to the PR after push. fit.py auto-simulated the dataset from scratch after rm -rf (proving the should_simulate path); χ² = 437.510 for 400 DoF — bit-identical to the series' known value.
• Dataset regeneration churn (dataset.json) was restored, not committed.

Heart: YELLOW, reason set unchanged from the list you acknowledged at the weak-modeling sign-off earlier today (stale 2026-07-08 workspace-validation report, stale parked scripts, assistant version pin, install verification, release validation, old Mind PR).

The question. Approve shipping? On approval: push both branches, open the PyAutoLens PR (pending-release) and the autolens_workspace PR (library-first gate), append the autonomy-log row — same shape as #580/#241. Amendments welcome here; any session resumes from active.md.

[Jammy2211]

Ship approved — PRs open: #582 (library, pending-release) and PyAutoLabs/autolens_workspace#244 (workspace, merge after the library PR). Watching CI; will merge library-first on green and close this issue.

[Jammy2211]

Shipped ✅

Both PRs merged library-first with all CI green on the first run (the in-branch catalogue regeneration avoided #241's staleness red):

• feat: tangential/cross shear profiles + Kaiser-Squires map (weak series step 6) #582 — shear profiles + Kaiser-Squires (352 tests, incl. the axis-parity regression test)
• feat: weak shear-profile + convergence-map demos; should_simulate adoption autolens_workspace#244 — fit/simulator demos + should_simulate adoption across the weak scripts

The weak series now has its full core visualization suite (11 aplt helpers). Continuing the home straight: step 5 (likelihood_function.py) next. Closing.