fix(research): survive the #137 config layout, and anchor the flag rewrite

[TimeToBuildBob]

Two defects in scripts/patch_research_edition_config.py, both introduced by the interaction with ActivityWatch/aw-watcher-window#137. Neither is visible today, and both fire on the next submodule bump.

#137 itself is right — moving research knobs out of default_config stops them being persisted into the config of every user who is not in a study. It just removes the two TOML table headers this script anchors on.

1. Injection breaks on the next submodule bump

patch_config() requires exactly one [aw-watcher-window.research_category_map] and one [aw-watcher-window.research_app_category_map] header. Past #137 neither exists, so the script aborts and the Research Edition build fails.

Verified against both revisions of aw_watcher_window/config.py:

submodule sha	category header	app header
a297914 (#136 merge — the current pin)	1	1
a7690ac (aw-watcher-window master, #137)	0	0

So the build works today purely because the pin has not moved. Anything that advances it — including a routine dependabot submodule bump — breaks it. The fail-closed message would then misdiagnose the cause as "submodule predates #136", which is the opposite of what happened.

2. The flag rewrite could silently disable the Research Edition

#137 documents the release-time rewrite in a comment containing the literal text:

#   sed -i 's/^research_enabled = false$/research_enabled = true/'

That comment sits above the real flag. The existing unanchored text.replace(enabled_line, ..., 1) therefore rewrites the comment and leaves research_enabled = false.

That failure has no error and no failed step. It ships a signed Research Edition installer with the privacy filter off — to study participants. This is the one I'd want a second pair of eyes on.

Fix

• Both layouts handled. Post-Wakatime supports browsers #137 the maps are injected into research_defaults with unprefixed headers — that template is parsed standalone and merged into the [aw-watcher-window] section key by key, so a prefixed header would create a nested aw-watcher-window key that nothing reads. Pre-Wakatime supports browsers #137 the existing prefixed headers are used, unchanged.
• Flag rewrite is line-anchored via re.MULTILINE, matching the contract Wakatime supports browsers #137's own comment documents. It refuses to proceed on anything other than exactly one match rather than guessing.
• The pre-Add API endpoint for removing events #136 guard now keys on the runtime lookup config.get("research_app_category_map" instead of a table header. That tests the capability that actually matters — whether the watcher can consume an app map — and survives further reshuffling of the config templates.

Verification

Against the two real config.py revisions, not just synthetic fixtures — both yield research_enabled = true, 570 category patterns, 72 app entries, and output that parses as valid Python and valid TOML.

Test suite rewritten to cover both layouts (7 tests, passing), including a regression test asserting the sed-example comment survives while the real flag flips. The previous fixtures omitted the runtime lookup that real config.py always has, which is why they did not catch this.

Context

This is on the critical path for a Lund University study build promised this week. Related: aw-watcher-window#136, #137, and #1397.

[greptile-apps]

Greptile Summary

The PR updates the Research Edition configuration patcher to support both watcher configuration layouts while safely targeting the effective feature flag.

• Anchors the research_enabled rewrite and fails closed on missing or ambiguous matches.
• Injects category maps into either the legacy prefixed tables or the post-Wakatime supports browsers #137 standalone defaults template.
• Replaces layout-dependent app-map validation with a runtime-lookup capability marker.
• Expands tests across both layouts, TOML/Python parsing, comment preservation, and failure paths.

Confidence Score: 5/5

The PR appears safe to merge, with no concrete blocking or non-blocking defects identified in the changed paths.

The patcher validates the unique effective flag and required app-map capability before handling each known configuration layout, while the expanded tests cover successful output and relevant failure modes.

Important Files Changed

Filename	Overview
scripts/patch_research_edition_config.py	Adds anchored flag replacement, runtime capability validation, and layout-aware map injection without an established actionable defect.
scripts/tests/test_patch_research_edition_config.py	Reworks fixtures and assertions to validate both layouts, parsed output, comment preservation, and fail-closed behavior.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Research Edition build] --> B[Load watcher config.py]
  B --> C{Exactly one anchored disabled flag?}
  C -->|No| X[Fail build]
  C -->|Yes| D{App-map runtime lookup present?}
  D -->|No| X
  D -->|Yes| E{research_defaults template present?}
  E -->|Yes| F[Enable flag and inject unprefixed maps]
  E -->|No| G[Validate legacy prefixed tables]
  G -->|Invalid| X
  G -->|Valid| H[Enable flag and populate prefixed maps]
  F --> I[Write patched watcher config]
  H --> I

Loading

_{Reviews (1): Last reviewed commit: "fix(research): survive the #137 config l..." | Re-trigger Greptile}