test(#115): unit tests for _parse_log_summary, _age_str, and cmd_cost

[CoreyRDean]

Closes #115

Motivation

`_parse_log_summary` is the core log-parsing function that drives both `clauck history` and `clauck trace`. It landed in #93 with zero unit test coverage. `cmd_cost`'s log-glob + regex parse loop is pure IO + computation — entirely testable with temp files. `_age_str` is a pure function called from both commands.

If any of these silently mis-parse (missed exit code format, wrong cost regex, wrong invocation_id composition), `clauck history` and `clauck cost` would display incorrect data with no test catching the regression.

Before / After

Before: python3 -m pytest tests/ runs 177 tests. _parse_log_summary, _age_str, and cmd_cost are entirely dark.

After: 26 new tests in tests/test_clauck_history_cost.py. python3 -m pytest tests/ runs 203 tests, all green.

What's covered

_parse_log_summary (11 tests):

• All fields populated: name, ts_str, pid, exit_code, cost_usd, terminal_reason
• invocation_id composed as {ts}-{pid}
• No exit_code line → running=True, exit_code=None
• Nonzero exit code extracted correctly
• Cost extracted from "total_cost_usd":N.NNNN line
• No cost line → cost_usd=None
• terminal_reason extracted from JSON-embedded line
• Malformed filename (no timestamp) → returns None
• Empty file → dict returned with running=True, all fields None
• path field is the original path object
• Hyphenated job name parsed correctly

_age_str (6 tests):

• Seconds boundary (< 60s)
• Minutes boundary (>= 60s, < 3600s)
• Hours boundary (>= 3600s, < 86400s)
• Days boundary (>= 86400s)
• Exactly 60 seconds → "1m ago"
• Exactly 1 hour → "1h ago"

cmd_cost (9 tests):

• No log files → prints "no log data found"
• Single run: correct spend shown in table
• Multiple jobs sorted by spend descending
• Total row present when multiple jobs
• --name filter: only named job shown, others excluded
• Failure run (terminal_reason ≠ completed) counted in fails column
• DAG orchestration log (matching -dag- pattern) excluded
• --days filter: old log (mtime 60 days ago) excluded from 30-day window
• --all-time: same old log appears when filter bypassed

Cost-to-Value

282-line test file. Zero production code changes. Zero new dependencies. CI adds ~0.1s.

Confidence

• 203/203 pass locally
• Tests import lib/clauck via SourceFileLoader (same pattern as other clauck test files)
• cmd_cost tests patch JOBS_DIR with a tempfile.TemporaryDirectory — fully isolated from ~/.clauck

Validation Steps

git fetch origin '#115-log-parsing-tests'
git checkout '#115-log-parsing-tests'
python3 -m pytest tests/test_clauck_history_cost.py -v
# expect: 26 passed
python3 -m pytest tests/ -q
# expect: 203 passed

[CoreyRDean]

🤖 Autonomous Agent Disclosure

This PR was produced entirely without human oversight by an autonomous agent (Model: claude-sonnet-4-6).

Readiness: Draft. Implementation is complete — 26/26 tests pass, 203/203 full suite passes.

Known limitations: None identified. All tests are isolated from `~/.clauck` via `tempfile.TemporaryDirectory` and `JOBS_DIR` patching.

This PR must remain in draft status until a human reviewer explicitly promotes it to ready-for-review.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 625e935838

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Check fails column value explicitly

self.assertIn("1", out) can pass even when failure accounting is broken, because cmd_cost output already contains 1 in other fields (for example the runs count for a single job). In that scenario this test would still be green and would not catch a regression in the fails column logic, so it should assert the failure cell specifically (e.g., via a row-level pattern) rather than any digit anywhere in the table output.

Useful? React with 👍 / 👎.