Increase test coverage to prevent regressions

[jkeresman01]

Title: Increase test coverage to prevent regressions

Summary

Recent refactors (e.g., dependency picker ↔ UI adapter) exposed gaps in test coverage that allowed breaking changes to be delivered. Expand automated tests to cover core modules and integration paths to prevent similar regressions.

Scope

• Add/extend unit tests across modules
• Add light integration “smoke” tests for UI wiring
• Keep the Arrange–Act–Assert convention
• Reuse the existing Plenary/Busted runner and tests/minimal_init.lua

Target Areas

1. algo/hashset.lua

   • Already covered; add edge cases (nil inputs, boolean keys, large inserts, union with empty set).

2. telescope/telescope.lua (dependency picker)

   • Flattens groups correctly (empty groups, missing values).
   • Entry formatting ([Group] Name) and ordering.
   • Selection handler calls record_selection once per confirm.
   • No duplicates when the same item is confirmed multiple times.

3. ui/deps.lua (adapter used by layout)

   • create_button maps <CR> to open picker and then calls update_display.
   • create_display initializes buffer and first render.
   • update_display renders “no items” vs list state.
   • (Stub/mocking) Verify it reads from telescope.selected_dependencies.

4. ui/layout.lua

   • “Builds” end-to-end without throwing (smoke test).
   • Left/right panels both mount; focus maps exist.

5. metadata/metadata.lua

   • Successful decode: calls callbacks with table and sets loaded.
   • Failed decode: sets error, calls callbacks with error.
   • Subsequent fetch_metadata returns cached data without issuing a job.

6. utils/

   • url: urlencode and encode_query (reserved chars, spaces, unicode).
   • http: download_file success vs non-zero exit (mock Job).
   • file: unzip triggers on_done and removes file (mock Job, stub os.remove).
   • window: get_winid for {winid} vs {popup.winid}; safe_close no-op on invalid.

Proposed Plan

• Create new specs in tests/ mirroring module paths:

  tests/
  ├── algo/hashset_spec.lua
  ├── telescope/telescope_spec.lua
  ├── ui/deps_spec.lua
  ├── ui/layout_smoke_spec.lua
  ├── metadata/metadata_spec.lua
  └── utils/
      ├── url_spec.lua
      ├── http_spec.lua
      ├── file_spec.lua
      └── window_spec.lua
  

• Use Plenary stubs/mocks:
  • Stub require("plenary.job").new to inject fake on_exit, result, stderr_result.
  • For UI smoke tests, construct minimal instances and assert no exceptions and expected side-effects (buffer lines, keymaps exist).
• Keep tests headless; avoid real network/file system effects.

CI Integration

• Reuse/extend existing test workflow:
  • Ensure plenary is installed (vendor path) in CI.
  • Run: nvim --headless -u tests/minimal_init.lua -c "PlenaryBustedDirectory tests" +q

Acceptance Criteria

• Specs exist for all modules listed in Target Areas.
• Tests follow Arrange–Act–Assert and run headlessly via Plenary/Busted.
• ui/layout.lua smoke test mounts the layout without runtime errors.
• ui/deps.lua adapter tests verify create_button, create_display, update_display behavior and interaction with the picker state.
• metadata.lua tests cover success, parse failure, and cache hit paths.
• utils tests cover happy paths and error branches.
• CI test job passes on push and pull_request to main.

Notes

• Where direct mocking is awkward, wrap external calls (e.g., Job/new) behind tiny local helpers to make stubbing easier in tests.
• Keep tests deterministic and avoid timing flakiness by calling scheduled callbacks synchronously where practical.