Repository navigation
Tests: one conftest, shared fakes, and 51 files become 15 - #14
Merged
Merged
Conversation
…d setup
- tests/conftest.py sets the env placeholders before any palmer import, so
the eight module-level load_dotenv() calls go, and with them the last way
real keys could reach a test. PALMER_NO_SCHEDULER is set there too, which
retires the three `with patch("...BackgroundScheduler.start"): import
main` blocks and the collection-order dependency they hid.
- fresh_db fixture replaces eight _fresh/_fresh_db helpers and thirteen
inline copies (ten of them hand-numbered t.db..t12.db in one file). Files
whose tests reach the setup through a local helper keep that helper.
- Two autouse guards. The socket guard makes any fetch that escapes its
patch fail by name instead of succeeding silently on a machine with live
keys. The Anthropic guard came out of turning the first one on: a dozen
tests were calling the real API with a placeholder key on every run,
spending 1.4s each in the SDK's retry backoff and swallowing the error.
The suite is 5s now against 7s before, for that reason.
- tests/helpers.py holds llm_reply (was _haiku_reply x4, _haiku, _resp,
_reply), Block/Resp (x4, byte-identical), tool_by_name (x2) and
drive_tool, the get_reply-through-one-tool-call scaffold that four files
each carried in full. Each of those keeps a short wrapper naming its
patches; the wrappers carry distinct names so the files can merge.
1,430 tests before and after.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018QpUhnYC3U7KxF8QCeenfQ
Each new file is the concatenation of the old ones, section by section,
with every original docstring kept as a comment block at the top of its
section — those carry the incident that led to the test and are the part
worth keeping. Imports are hoisted and merged; the two identical PHONE
constants collapse to one. Names that would have collided were given
distinct names in the previous commit.
test_db test_db_claims, test_phantom_history, test_local_day_guards
test_userprofile test_profile_schema, test_profile_freshness, test_user_covered, test_onboard
test_time test_timeutil, test_reminder_time, test_reminder_recurrence, test_clock_correctness
test_weather test_weather, test_weather_source, test_weather_locations, test_forecast_audit
test_shopping test_price_watches, test_price_daily_cap, test_price_alert, test_amazon_watches
test_watches test_watches_story, test_watches_genre, test_corroboration, test_subject_dedup, test_rubrics, test_followup
test_news test_sources, test_trends, test_tickers
test_morning test_morning_link, test_morning_schedule, test_briefing_shape, test_weather_city, test_repetition
test_page test_page, test_cards, test_my_page_tool, test_arrange_page_tool
test_opening test_opening, test_shows
test_sms test_sms_chunking, test_link_integrity, test_tapback, test_reaction_interpret
test_agent test_calibration, test_reply_resilience, test_commute, test_add_price_watch_tool
test_guards test_guards_and_flights, test_scheduler_config
test_home, test_sports unchanged
Three placements are by what the file tests rather than by its name:
test_weather_city tests the morning drafter and the agent write path (its
own docstring says weather.py was innocent), test_add_price_watch_tool
tests dispatch, and test_shows shares Opening's seeded-cache machinery.
1,430 tests before and after.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018QpUhnYC3U7KxF8QCeenfQ
3 of 4 tasks
fresh_db's monkeypatch used raising=False, which silently no-ops when DATABASE_URL is set instead of failing loudly like the old per-file fixtures did — a dev with DATABASE_URL exported would have tests run against a real Postgres database instead of a fresh SQLite file. Restore the loud failure with an explicit assertion. Also finish the consolidation: six files still reimplemented fresh_db's two-line setup locally instead of using the shared fixture, and three more rebuilt helpers.llm_reply's mock shape by hand in files that already import it. Fix four leftover "Run: pytest test_X.py" comments pointing at files this PR deleted, and correct CLAUDE.md's claim that the network guard "fails by name" — netutil._http_get_json's six callers catch every exception by contract and return None, so an escape there still reads as an empty result rather than a named failure. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Re-applies the morning-format, preview-card and local-day test changes into the consolidated test_morning.py, test_page.py and test_userprofile.py. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Stage 3 of 4 (hygiene → package layout → test consolidation → docs split). Stacked on #13; merge that first and this retargets to
main. 1,430 tests before and after, and the suite is faster (5s vs 7s) for a reason worth reading.Commit 1 —
tests/conftest.pyandtests/helpers.pypalmerimport; the eight module-levelload_dotenv()calls go, and with them the last way real keys could reach a test.PALMER_NO_SCHEDULERis set there too, retiring the threewith patch("…BackgroundScheduler.start"): import mainblocks and the collection-order dependency they hid.fresh_dbfixture replaces eight_fresh/_fresh_dbhelpers and thirteen inline copies (ten hand-numberedt.db…t12.dbin one file).helpers.py:llm_reply(was_haiku_reply×4,_haiku,_resp,_reply),Block/Resp(×4 byte-identical),tool_by_name(×2), anddrive_tool, the get_reply-through-one-tool-call scaffold four files carried in full.Commit 2 — 51 files → 15, grouped by module under test. Concatenated section by section; every original docstring is kept as a comment block heading its section, because those carry the incident behind the test. Imports hoisted and merged. Three placements go by what the file tests rather than its name:
test_weather_city→ morning (its docstring says weather.py was innocent),test_add_price_watch_tool→ agent (it tests dispatch),test_shows→ opening (shared seeded-cache machinery).Test plan
ruff check .cleanpytest -q— 1,430 passed after each commit, same asmain.envand no exported keys--durationsshows nothing over 0.5s; the retry-backoff tests are gone🤖 Generated with Claude Code
https://claude.ai/code/session_018QpUhnYC3U7KxF8QCeenfQ