Thread styling kwargs through plotly visualization backends

[igerber]

Summary

• Thread marker, markersize, linewidth, capsize, ci_linewidth, and show_grid kwargs from public plot functions through to their plotly renderers
• Add _mpl_marker_to_plotly_symbol() helper in _common.py for matplotlib-to-plotly marker symbol conversion
• Convert plot_bacon scatter markersize (area in points²) to plotly diameter (px) via sqrt conversion
• linewidth/capsize not passed to event study plotly renderers (plotly uses CI bands, not error bars)
• 13 new tests verifying kwargs reach plotly trace properties
• Remove resolved TODO item

Methodology references

• N/A — no methodology changes, visualization only

Validation

• Tests added/updated: tests/test_visualization_plotly.py (13 new tests)
• All 114 visualization tests pass (33 plotly + 81 matplotlib)
• Formatting (black) and linting (ruff) clean

Security / privacy

• Confirm no secrets/PII in this PR: Yes

---

Generated with Claude Code

[github-actions]

Overall Assessment

✅ Looks good

Executive Summary

• Cross-checking the touched areas against the Methodology Registry found no estimator, weighting, identification, variance, or default-behavior change. This PR is limited to visualization plumbing for Bacon, HonestDiD sensitivity/event-study, PowerAnalysis, and PreTrendsPower.
• The main plotly threading work is correct for ci_linewidth, most marker/markersize paths, and linewidth/show_grid on the power plots.
• plot_event_study() and plot_honest_event_study() still leave some public styling kwargs silently ignored on plotly, so the old styling-gap TODO is not fully resolved.
• The new marker helper silently coerces unsupported matplotlib markers to "circle" and uses the same hollow-marker recipe for line-only symbols like "+"/"x", which does not preserve matplotlib semantics on plotly.
• New tests cover happy-path supported kwargs, but not the remaining ignored/rejected-kwargs paths or line-only reference-marker behavior.
• Plotly is not installed in this review environment, so plotly-specific rendering comments are from static review rather than execution.

Methodology

• Severity: P3-informational. Impact: No methodology defect. The affected registry sections are docs/methodology/REGISTRY.md:L1533-L1718, and the touched public APIs are diff_diff/visualization/_diagnostic.py:L12-L79, diff_diff/visualization/_event_study.py:L29-L120, and diff_diff/visualization/_power.py:L13-L86. The diff changes only plotly trace/layout styling, not estimator equations or inference logic. Concrete fix: None.

Code Quality

• Severity: P2. Impact: plot_event_study() still advertises public linewidth and capsize controls in diff_diff/visualization/_event_study.py:L29-L104, but the plotly branch drops both when dispatching in diff_diff/visualization/_event_study.py:L266-L282. plot_honest_event_study() likewise accepts capsize yet omits it on the plotly path in diff_diff/visualization/_event_study.py:L728-L856. Non-default values therefore remain silent no-ops on plotly. Concrete fix: Either implement plotly equivalents, or explicitly reject/warn on non-default linewidth/capsize when backend="plotly".

• Severity: P2. Impact: _mpl_marker_to_plotly_symbol() falls back to "circle" for any unrecognized marker in diff_diff/visualization/_common.py:L305-L319, and the new tests codify that behavior in tests/test_visualization_plotly.py:L333-L357. That silently changes valid matplotlib marker requests instead of honoring or rejecting them. The reference-marker traces in diff_diff/visualization/_event_study.py:L527-L542 and diff_diff/visualization/_event_study.py:L1065-L1078 also assume a filled marker with outline, which does not preserve line-only symbols such as "+" and "x". Concrete fix: Validate the supported marker subset up front, raise on unsupported markers, and special-case line-only symbols in the reference traces instead of using the hollow filled-marker recipe.

Performance

• Severity: P3-informational. Impact: No meaningful performance regression is apparent; the added work is limited to kwarg forwarding, a small dictionary lookup, and axis layout updates in diff_diff/visualization/_common.py:L286-L319, diff_diff/visualization/_diagnostic.py:L698-L775, and diff_diff/visualization/_power.py:L284-L344. Concrete fix: None.

Maintainability

• Severity: P3-informational. Impact: Centralizing marker translation in diff_diff/visualization/_common.py:L286-L319 is a maintainability improvement. No separate maintainability defect stood out beyond the API-contract gaps above. Concrete fix: None.

Tech Debt

• Severity: P2. Impact: The PR removes the tracked TODO for plotly styling kwargs, but the event-study/honest-event-study gaps above remain. After merge, that limitation would no longer be tracked in the current TODO section at TODO.md:L70-L77. Concrete fix: Keep the TODO until all public styling kwargs are either supported or explicitly rejected on plotly, or replace it with a narrower TODO covering the remaining event-study limitations.

Security

• Severity: P3-informational. Impact: No new file/network/process behavior, no secrets handling, and no obvious injection surface in the touched files. Concrete fix: None.

Documentation/Tests

• Severity: P3. Impact: The new tests in tests/test_visualization_plotly.py:L363-L522 cover happy-path supported kwargs, but they do not cover unsupported markers, line-only reference markers, or explicit support/rejection of non-default event-study linewidth/capsize. Those are the remaining contract edges in the new code. Concrete fix: Add plotly tests for marker="+" and marker="x" with a reference period, for an unsupported matplotlib marker (expecting error/warning), and for non-default event-study linewidth/capsize on plotly (expecting either support or explicit rejection).