test: improve test coverage from 43% — analysis and roadmap

[junhoyeo]

Current State

Coverage badge:

Overall: 43% (4,784 covered / 11,125 coverable lines)

The coverage was stuck at 0% due to a CI bug (wrong tarpaulin JSON filename), fixed in c187ac1. Now that it's reporting correctly, here's what's actually uncovered.

0% Coverage — Completely Untested (2,534 lines)

TUI Rendering (~2,000 lines)

These are ratatui rendering functions that require a terminal — inherently difficult to unit test.

File	Coverable Lines
tui/ui/stats.rs	422
tui/ui/overview.rs	223
tui/ui/models.rs	190
tui/ui/footer.rs	190
tui/mod.rs	148
tui/ui/daily.rs	133
tui/ui/bar_chart.rs	117
tui/ui/widgets.rs	113
tui/ui/dialog/source_picker.rs	162
tui/ui/dialog/group_by_picker.rs	84
tui/ui/mod.rs	67
tui/ui/header.rs	52
tui/ui/spinner.rs	32
tui/config.rs	32
tui/ui/dialog/overlay.rs	14

Network/Pricing (~211 lines)

Requires HTTP mocking to test without real network calls.

File	Coverable Lines
pricing/openrouter.rs	126
pricing/cache.rs	44
pricing/litellm.rs	41

Session Parsers (~116 lines)

Pure parsing logic with no external deps — easiest wins.

File	Coverable Lines
sessions/amp.rs	66
sessions/kilo.rs	50

Low Coverage (<40%) — High-Value Targets

File	Coverage	Covered/Total	Notes
main.rs	10.7%	150/1403	CLI dispatch — tested via integration tests but tarpaulin can't trace subprocess
cursor.rs	12%	64/535	Cursor API sync — needs HTTP mocking
auth.rs	27.9%	36/129	Device auth flow — needs API mocking
commands/wrapped.rs	30.4%	270/889	Image generation — heavy external deps
sessions/droid.rs	34%	34/100	Droid parser — pure parsing, easy to test
sessions/synthetic.rs	35%	36/103	Synthetic message handling

Well-Tested Core (>80%)

File	Coverage
aggregator.rs	98.4%
provider_identity.rs	97.9%
sessions/qwen.rs	98.1%
sessions/gemini.rs	96.6%
sessions/openclaw.rs	96.1%
scanner.rs	95.3%
sessions/claudecode.rs	94.6%
sessions/pi.rs	94.6%
sessions/roocode.rs	93.6%
sessions/crush.rs	93.3%
message_cache.rs	90.1%
pricing/lookup.rs	86.2%
sessions/codex.rs	85.2%

Recommended Priority

Quick wins (pure parsing, no deps)

1. sessions/amp.rs — 66 lines, 0% → similar pattern to existing Pi/Gemini tests
2. sessions/kilo.rs — 50 lines, 0% → same
3. sessions/droid.rs — 100 lines, 34% → add more fixture cases
4. sessions/synthetic.rs — 103 lines, 35% → add edge cases

Medium effort (HTTP mocking)

5. pricing/litellm.rs + pricing/openrouter.rs — mock HTTP responses
6. pricing/cache.rs — filesystem mocking
7. cursor.rs — API sync mocking

Hard / low ROI

8. TUI rendering code — consider snapshot testing with ratatui::backend::TestBackend
9. main.rs — already covered by integration tests, tarpaulin limitation
10. commands/wrapped.rs — requires image generation deps

Target

Realistic near-term target: 60% by covering the quick wins.
Stretch goal: 70% with HTTP mocking for pricing/cursor.