fix(ci): green main — pyvis.shiny unit guard + de-flake leverage/simulation e2e

[razinkele]

Why

main CI has been red on every push since ~2026-07-01. Two independent causes:

1. Unit-test jobs (py3.11 + py3.12) fail deterministically at collection with
   ModuleNotFoundError: No module named 'pyvis.shiny' on tests/test_cld.py and
   tests/test_cld_filter.py. Those modules import sespy.modules.cld_visualization /
   analysis_* at top level, which import the pyvis fork at load. The unit jobs install
   stock pyvis from PyPI (by design — fork tests belong in the conda jobs), so collection
   crashes. The two files simply missed the importorskip guard that test_network /
   test_report already use.
2. Playwright e2e job is flaky — test_leverage_e2e hard-asserted it caught a
   transient "Computing…" caption (racy), and separately read window.pyvisNetworks[...]
   after a fixed 2.5s sleep (crashed on undefined.nodes when the network hadn't
   registered). test_simulation_e2e had the same arbitrary-timeout-then-assert pattern.

What

• test_cld.py / test_cld_filter.py: add module-level pytest.importorskip("pyvis.shiny").
  Skips cleanly in the stock-pyvis unit jobs; still runs for real in the conda full-app/e2e jobs.
• test_simulation_e2e.py: replace wait_for_timeout→assert with wait_for_selector on the
  plot images (and run button), matching the Monte-Carlo poll already in the file.
• test_leverage_e2e.py: downgrade the transient-caption sighting to best-effort (log on miss;
  the load-bearing 95% CI-column assertion is kept and already polled), and replace the pyvis
  network sleep with wait_for_function waits.

Verification (local)

• pyvis guard: reproduced stock-pyvis env by blocking pyvis.shiny → 2 skipped, not errored;
  fork present → 5 passed; ruff clean.
• simulation e2e: 3/3 green (was failing on a slower machine).
• leverage e2e: 4/4 green (was ~1/2; caught + fixed a second race during verification).
• full unit suite (fork env): 474 passed, 1 skipped.

Not reproducible locally: the exact CI unit env (ubuntu + stock pyvis) and the fast-runner e2e
timing — this PR exists to confirm those go green.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Tests
  • Improved end-to-end test reliability by waiting for screens and results to load before checking them.
  • Made some test suites skip automatically when an optional visualization dependency isn’t available, reducing false failures in limited environments.
  • Loosened one transient loading-state check so tests focus on the final rendered result instead of brief intermediate states.

[coderabbitai]

📝 Walkthrough

Walkthrough

Test modules test_cld.py and test_cld_filter.py now skip entirely when the pyvis.shiny dependency is unavailable. E2E tests for Leverage Points and Dynamic Simulation replace fixed timeouts with polling/selector-based waits and relax a transient "Computing" caption assertion to a warning.

Changes

Test Robustness Updates

Layer / File(s)	Summary
pyvis.shiny import guards tests/test_cld.py, tests/test_cld_filter.py	Adds module-level pytest.importorskip("pyvis.shiny") with explanatory comments so these modules skip when the pyvis fork is missing.
Leverage Points E2E polling tests/test_leverage_e2e.py	Replaces fixed wait_for_timeout with polling for nav button activation and pyvis network/node readiness; makes the "Computing" caption check best-effort with a warning instead of a hard failure.
Simulation E2E selector waits tests/test_simulation_e2e.py	Replaces fixed timeouts and JS DOM checks with wait_for_selector for trajectory, final-state, and Monte Carlo histogram plot images.

Estimated code review effort: 2 (Simple) | ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately summarizes the CI-focused pyvis.shiny test guard and e2e flake fixes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/ci-unit-pyvis-shiny-guard

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

tests/test_leverage_e2e.py (1)

17-27: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value

Minor redundancy: double-checking nav activation.

wait_for_function (Line 18-23) already asserts the exact same condition (.sespy-nav-btn.active has exactly one match equal to sespy_nav_leverage) that the follow-up eval_on_selector_all + assert (Line 24-27) re-verifies. Not incorrect, just a redundant round trip.

♻️ Optional simplification

-        await page.wait_for_function(
-            "() => { const a = Array.from(document.querySelectorAll("
-            "'.sespy-nav-btn.active')).map(e => e.id);"
-            " return a.length === 1 && a[0] === 'sespy_nav_leverage'; }",
-            timeout=15000,
-        )
-        nav_active = await page.eval_on_selector_all(
-            ".sespy-nav-btn.active", "els => els.map(e => e.id)"
-        )
-        assert nav_active == ["sespy_nav_leverage"], f"unexpected: {nav_active}"
+        nav_active = await page.wait_for_function(
+            "() => { const a = Array.from(document.querySelectorAll("
+            "'.sespy-nav-btn.active')).map(e => e.id);"
+            " return (a.length === 1 && a[0] === 'sespy_nav_leverage') ? a : null; }",
+            timeout=15000,
+        )
+        assert (await nav_active.json_value()) == ["sespy_nav_leverage"]

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/test_leverage_e2e.py` around lines 17 - 27, The nav activation check in
the leverage e2e test is duplicated: wait_for_function already verifies that
.sespy-nav-btn.active contains exactly sespy_nav_leverage, so the follow-up
eval_on_selector_all and assert in the same test are redundant. Simplify the
test by keeping the wait_for_function in the leverage flow and removing the
extra selector readback/assert, or otherwise consolidate the check into a single
validation using the existing page wait logic.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@tests/test_leverage_e2e.py`:
- Around line 17-27: The nav activation check in the leverage e2e test is
duplicated: wait_for_function already verifies that .sespy-nav-btn.active
contains exactly sespy_nav_leverage, so the follow-up eval_on_selector_all and
assert in the same test are redundant. Simplify the test by keeping the
wait_for_function in the leverage flow and removing the extra selector
readback/assert, or otherwise consolidate the check into a single validation
using the existing page wait logic.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: dd351d48-0603-4a48-86f0-8cbef7a4ed1c

📥 Commits

Reviewing files that changed from the base of the PR and between 57ae262 and 55e0650.

📒 Files selected for processing (4)

• tests/test_cld.py
• tests/test_cld_filter.py
• tests/test_leverage_e2e.py
• tests/test_simulation_e2e.py