# Current test inventory β€” what exists and what it asserts Status: complete inventory as of 2026-07-07. Source of truth: the six files in `/srv/apps/work/tests/*.test.ts` (read directly), cross-checked against `docs/tests-catalog.md` and the extraction in `scratchpad/extract/docs-features-d.md`. Where the catalog disagrees with the test files, **the test files win** β€” every disagreement is listed in "Catalog staleness" below. This document is the input to the (not-yet-written) `06-test-plan.md`. The R1 success criterion in `00-vision-and-scope.md` is "all ported tests green (the current 116 + ports of the RPC/permission semantics)" β€” the count **116** is verified here (see totals table). ## How the current suite runs - Runner: **vitest** (`"test": "vitest run"`, `"test:watch": "vitest"` in `package.json`; vitest `^4.1.7`, coverage via `@vitest/coverage-v8`). - All 116 tests are **pure unit tests of `lib/*` helpers** β€” no database, no network, no Supabase, no React rendering. Imports use the `@/lib/...` path alias. - Everything DB/auth-dependent is deliberately untested (the 🟑 catalog-only list at the bottom). - One suite (`filters.test.ts`) pins the clock with `vi.useFakeTimers()` + `vi.setSystemTime(new Date('2026-05-15T12:00:00Z'))` (reset in `afterEach`); FLT-004 re-pins to `2026-01-15T12:00:00Z` inside the test. Any port must reproduce clock-pinning. ## Totals and ports-to summary | File | Suite prefix(es) | # tests | Module under test | Ports to | |---|---|---:|---|---| | `tests/clockify.test.ts` | CLK | 43 | `lib/clockify.ts` | Python (pytest) β€” import pipeline; **except** CLK-019/020 β†’ dropped | | `tests/cc-expenses.test.ts` | CCE | 29 | `lib/cc-expenses.ts` | Python (pytest) β€” CC-expense paste pipeline + exports | | `tests/toggl.test.ts` | TGL | 14 | `lib/toggl.ts` | dropped-with-feature β€” **except** TGL-009/011/012 (blocked predicate) β†’ Python | | `tests/invoices.test.ts` | INV | 11 | `lib/invoices.ts` | Python (paste block); Vitest (date default) | | `tests/filters.test.ts` | FLT | 10 | `lib/filters.ts` | Vitest β€” filter serialization (with a pytest twin at the API boundary) | | `tests/format.test.ts` | FMT | 9 | `lib/format.ts` | Vitest β€” display formatting | | **Total** | | **116** | | | Ports-to legend (from `04-architecture.md` "Testing"): - **Python port** β€” pytest under `tests/` mirroring `services/` 1:1; the architecture doc explicitly says "import pipeline ports with the CLK-* fixtures". - **Vitest port** β€” SPA-side logic ("filter serialization, table interaction hooks") plus pure display helpers. - **dropped-with-feature** β€” the Toggl export is a confirmed phase-out (`00-vision-and-scope.md`: rian 2026-07-06, "I no longer have any need for toggl"); `rate_proportion` / converted-hours machinery is never implemented in the rebuild (ADR #034). --- ## File-by-file inventory Format per test: **ID** β€” Given / When / Then in one line β†’ *ports to*. ### 1. `tests/clockify.test.ts` (43 tests) β€” `lib/clockify.ts` Functions under test: `parseDateOnly(s, fmt)`, `parseTimeOnly(s, fmt?)`, `detectDateFormat(samples)`, `formatDateRangeLabel(range|null)`, `parseDurationSeconds(s)`, `parseClockifyCsv(csv, dateFmt, timeFmt)`, `splitClientProject(s)`, `applyConversion(entry, TeamLookup)`, `computeBilloutAmount(seconds|null, rate|null)`. #### `parseDateOnly` (6) - **CLK-001** β€” Given `'2026-05-20'` and `'2026-1-5'` with fmt `'YYYY-MM-DD'` / when parsed / then returns zero-padded `'2026-05-20'`, `'2026-01-05'`. β†’ *Python* - **CLK-002** β€” Given `'5/20/2026'` with fmt `'MM/DD/YYYY'` / then `'2026-05-20'` (US order). β†’ *Python* - **CLK-003** β€” Given `'2/12/2025'` / then `'DD/MM/YYYY'` yields `'2025-12-02'` while `'MM/DD/YYYY'` yields `'2025-02-12'` β€” same input, format-dependent result. β†’ *Python* - **CLK-004** β€” Given garbage (`'not a date'`, `''`) / then `null`. β†’ *Python* - **CLK-005** β€” Given out-of-range month/day (`'2026-13-01'`, `'2026-02-32'`, `'13/01/2026'` as MM/DD) / then `null`. β†’ *Python* - **CLK-006** β€” Given 2-digit years (`'5/20/26'`, `'20/5/26'`) / then interpreted as 20xx (`'2026-05-20'`). β†’ *Python* #### `parseTimeOnly` (5) β€” NOTE: the `fmt` arg is now **ignored**; AM/PM presence auto-detects 12h vs 24h per string - **CLK-007** β€” Given 24-hour strings (`'13:45:30'`, `'0:00:00'`, `'9:05'`) / then normalized `'HH:MM:SS'` (missing seconds β†’ `:00`). β†’ *Python* - **CLK-008** β€” Given 12-hour strings with AM/PM (case-insensitive) / then converted to 24h (`'1:45:30 PM'` β†’ `'13:45:30'`). β†’ *Python* - **CLK-009** β€” Given `12:xx AM` / then hour `00`; given `12:xx PM` / then hour `12`. β†’ *Python* - **CLK-010** β€” Given `'25:00:00'`, `''`, or `'13:45:30 PM'` (hour 13 with PM) / then `null`. β†’ *Python* - **CLK-026** β€” Given a 24h string / then result identical for fmt `'12h'`, `'24h'`, or omitted (fmt arg ignored). β†’ *Python* - **CLK-027** β€” Given a 12h string with PM / then result identical regardless of fmt arg. β†’ *Python* - **CLK-028** β€” Given `'1:45:30'` with no AM/PM suffix / then treated as 24h β†’ `'01:45:30'`, not `null` (regression: pre-auto-detect this failed under the 12h parser). β†’ *Python* (That block is 7 `it`s; CLK-026/027/028 were added 2026-06-10 with the auto-detect change.) #### `detectDateFormat` (9) - **CLK-029** β€” Given samples with a 4-digit-year prefix / then `'YYYY-MM-DD'`. β†’ *Python* - **CLK-030** β€” Given slash dates where ANY first slot > 12 (`'25/05/2026'`) / then `'DD/MM/YYYY'`. β†’ *Python* - **CLK-031** β€” Given slash dates where ANY second slot > 12 and no DD signal (`'5/20/2026'`) / then `'MM/DD/YYYY'`. β†’ *Python* - **CLK-032** β€” Given `[]` or all-gibberish samples / then `null`. β†’ *Python* - **CLK-032b** β€” Given an ambiguous week `01/06..08/06/2026` (all slots ≀ 12) / then **tightest-span heuristic** picks `'DD/MM/YYYY'` (one month vs eight). Regression for the 2026-06-09 incident where the fallback scattered a week across 8 months. β†’ *Python* - **CLK-032c** β€” Given the inverse US export `06/01..06/08/2026` / then `'MM/DD/YYYY'` (June 1–8, not Jan–Aug). β†’ *Python* - **CLK-032d** β€” Given `['1/2/2026','2/1/2026']` where both interpretations span equal months / then `null` (don't guess). β†’ *Python* - **CLK-033** β€” Given blanks/whitespace/gibberish mixed with one real `'25/05/2026'` / then unparseable entries ignored β†’ `'DD/MM/YYYY'`. β†’ *Python* - **CLK-034** β€” Given mixed ISO + numeric rows / then lands on one of the two valid answers without crashing (first non-null signal wins; test accepts either). β†’ *Python* #### `formatDateRangeLabel` (5) - **CLK-040** β€” Given `{min:'2026-06-02', max:'2026-06-08'}` / then `'Jun 2 – Jun 8, 2026'` (same-month). β†’ *Python* (import-preview label; move to Vitest if the rebuilt SPA renders the label client-side) - **CLK-041** β€” Given cross-month same-year / then `'May 28 – Jun 3, 2026'`. β†’ *Python* (same note) - **CLK-042** β€” Given cross-year / then both years: `'Dec 28, 2025 – Jan 3, 2026'`. β†’ *Python* (same note) - **CLK-043** β€” Given min == max / then single date `'Jun 2, 2026'`. β†’ *Python* (same note) - **CLK-044** β€” Given `null` / then `null`. β†’ *Python* (same note) #### `parseDurationSeconds` (4) - **CLK-011** β€” Given `'1:00:00'`, `'0:01:30'`, `'5:20:36'` (HH:MM:SS) / then exact seconds. β†’ *Python* - **CLK-012** β€” Given `'1:30'`, `'0:30'` (HH:MM) / then minutesΓ—60. β†’ *Python* - **CLK-013** β€” Given decimal hours `'1.5'`, `'0.25'`, `'5.34'` / then `Math.round(h Γ— 3600)`. β†’ *Python* - **(no ID)** β€” Given `''` or `'forever'` / then `null`. **Un-ID'd test** β€” assign an ID (suggest CLK-013b) when porting. β†’ *Python* #### `parseClockifyCsv` (6 in file; fixture = the 17-column Clockify header `Project,Client,Description,Task,User,Group,Email,Tags,Billable,Start Date,Start Time,End Date,End Time,Duration (h),Duration (decimal),Billable Rate (USD),Billable Amount (USD)`) - **CLK-014** β€” Given a header + 2 rows / then 2 entries, 0 errors; asserts `start_at`/`end_at` as naive `'YYYY-MM-DD HH:MM:SS'`, `duration_seconds`, `billable:true` from `'Yes'`; **CSV "Client" β†’ `operator`; CSV "Project" is split** (`project` set, `client` null when no separator). β†’ *Python* - **CLK-021** β€” Given `splitClientProject` inputs / then splits on the FIRST `' : '` (space-colon-space, strict): `'Foo : Bar : Baz'` β†’ client `'Foo'`, project `'Bar : Baz'`; `'NoColon'` and `'Foo:Bar'` (no spaces) β†’ client `null`, project unchanged. β†’ *Python* - **CLK-022** β€” Given rows with `Project: 'DaxTech : Website'` and `'PromptVictoria'` / then csv-level split into `client`/`project`; colon-less β†’ client `null`. β†’ *Python* - **CLK-015** β€” Given a CSV missing required headers / then 0 entries and `errors[0].message` matches `/missing required columns/i`. β†’ *Python* - **CLK-016** β€” Given rows with duplicate emails and scattered dates / then `uniqueEmails` deduped (`['a@x.com','b@x.com']`) and `dateRange` = `{min, max}` of parsed dates. β†’ *Python* - **CLK-017** β€” Given `'Mixed.Case@EXAMPLE.com'` / then `source_user_email` lowercased. β†’ *Python* #### `applyConversion` (4) β€” TeamLookup = `Map` - **CLK-018** β€” Given an email not in the lookup / then ALL null: `{team_member_id, converted_user, converted_duration_seconds, cost_rate_usd, billout_rate_usd}` all `null`. β†’ *Python, rewritten*: the surviving semantic is "unknown worker email leaves the entry unresolved" (engagement resolution in the new model); the converted_* fields themselves are dropped. - **CLK-019** β€” Given a member with `consolidate_as:'Tingang'` vs one without / then `converted_user` = `consolidate_as` if set, else `display_name`. β†’ **dropped-with-feature** (consolidate-as is Toggl-export plumbing). - **CLK-020** β€” Given 7200s with `rate_proportion` 0.5 vs 1.0 / then `converted_duration_seconds` = duration Γ— proportion (3600 / 7200). β†’ **dropped-with-feature** (ADR #034: no proportions, ever β€” judgment call #12). - **CLK-021 (DUPLICATE ID** β€” second use, distinct test) β€” Given a lookup member with `billout_rate_usd: 25` / then `applyConversion` surfaces `billout_rate_usd` in its result. β†’ *Python, rewritten*: billout rate comes from engagement compensation terms; the "rate is stamped from the member's current rate at import" semantic survives. #### `computeBilloutAmount` (2) - **CLK-022 (DUPLICATE ID** β€” second use) β€” Given `null` seconds or `null` rate / then `null`. β†’ *Python* (money-stamping math; invariant Β§5 of `02-domain-model.md`) - **CLK-023** β€” Given (3600, 25) β†’ 25; (1800, 25) β†’ 12.5; (60, 25) β‰ˆ 0.42 / then amount = `converted_seconds / 3600 Γ— rate`, rounded to 2dp. β†’ *Python* (same; in the rebuild the input is source seconds, not converted seconds) ### 2. `tests/cc-expenses.test.ts` (29 tests) β€” `lib/cc-expenses.ts` Functions: `parseExpenseDate(s)`, `parseCurrency(s)`, `parseCcLine(line)`, `parseCcPaste(text)`, `findHighlightMatch(desc, keywords)`, `findAutoRuleMatch(desc, rules)`, `formatCcPasteBlock(lines)`. All parsing/matching runs in the server action (`appendCcLines` in `app/(app)/tools/cc-expenses/actions.ts`) β†’ all port to **Python**. #### `parseExpenseDate` (5) - **CCE-D-001** β€” `'May 20, 2026'` β†’ `'2026-05-20'` (long format). β†’ *Python* - **CCE-D-002** β€” ISO `'2026-05-20'` passes through unchanged. β†’ *Python* - **CCE-D-003** β€” `'may 20 2026'` / `'MAY 20 2026'` (mixed case, no comma) still parse. β†’ *Python* - **CCE-D-004** β€” `'nonsense'`, `''`, `'Foobar 20, 2026'` β†’ `null`. β†’ *Python* - **CCE-D-005** β€” `'Jan 5, 2026'` β†’ `'2026-01-05'` (day zero-padded). β†’ *Python* #### `parseCurrency` (3) - **CCE-C-001** β€” Strips `$` and commas: `'$7.15'`β†’7.15, `'$3,856.56'`β†’3856.56, `'1234'`β†’1234. β†’ *Python* - **CCE-C-002** β€” `''`, whitespace, `'abc'` β†’ `null`. β†’ *Python* - **CCE-C-003** β€” Rounds to 2dp: `'$0.999'`β†’1, `'$10.234'`β†’10.23. β†’ *Python* #### `parseCcLine` (5) β€” input is one tab-separated bank-paste line: dateβ‡₯descriptionβ‡₯debitβ‡₯creditβ‡₯balance - **CCE-L-001** β€” Standard 5-column line (trailing tab OK) / then `{expense_date, description, amount_usd (from debit), balance_usd}` all parsed. β†’ *Python* - **CCE-L-002** β€” Debit blank, credit `'$50.00'` / then `amount_usd` = 50 (credit column used). β†’ *Python* - **CCE-L-003** β€” Empty/whitespace-only line β†’ `null`. β†’ *Python* - **CCE-L-004** β€” The verbatim input is preserved as `raw_line` (audit). β†’ *Python* - **CCE-L-005** β€” Partial parse: unparseable date β†’ `expense_date:null` but description + amount still extracted. β†’ *Python* #### `parseCcPaste` (2) - **CCE-P-001** β€” Splits on newlines, skips blank/whitespace lines, preserves order. β†’ *Python* - **CCE-P-002** β€” Handles CRLF (`\r\n`) line endings. β†’ *Python* #### `findHighlightMatch` (5) β€” keywords = `[{keyword: string}]` - **CCE-H-001** β€” Substring contains β†’ returns the matched keyword string (`'UBER *EATS'` β†’ `'UBER'`). β†’ *Python* - **CCE-H-002** β€” Case-insensitive. β†’ *Python* - **CCE-H-003** β€” No match / `null` desc / `''` desc β†’ `null`. β†’ *Python* - **CCE-H-004** β€” Multi-match β†’ first-in-list wins (list order semantics; NOTE flag #6 in the extraction β€” differs from auto-rules' created_at-DESC ordering, normalize in rebuild). β†’ *Python* - **CCE-H-005** β€” An entry with `keyword: ''` is skipped (doesn't match everything). β†’ *Python* #### `findAutoRuleMatch` (5) β€” rules = `[{id, keyword, project_id, assignment_description, category}]` - **CCE-R-001** β€” Match returns the whole rule object (all fields, for pre-fill). β†’ *Python* - **CCE-R-002** β€” Case-insensitive substring contains. β†’ *Python* - **CCE-R-003** β€” No matching rule β†’ `null`. β†’ *Python* - **CCE-R-004** β€” First-match wins; **caller controls order** (call sites pass created_at DESC β€” newest rule wins). β†’ *Python* - **CCE-R-005** β€” `null` / `''` description β†’ `null`. β†’ *Python* #### `formatCcPasteBlock` (4) β€” 7-column tab-separated export, one line per assigned expense - **CCE-X-001** β€” `[]` β†’ `''`. β†’ *Python* - **CCE-X-002** β€” Column shape: `['{client} : {project}', assignment_description, amount 2dp, expense_date, '', 'Bowden Works', category]`. β†’ *Python* - **CCE-X-003** β€” Amount always 2dp, never a `$`. β†’ *Python* - **CCE-X-004** β€” Multiple lines joined with `\n`. β†’ *Python* ### 3. `tests/toggl.test.ts` (14 tests) β€” `lib/toggl.ts` The Toggl export (`/transfer`, `buildTogglCsv`, `formatDurationSeconds`, `splitNaiveTimestamp`, `TOGGL_COLUMNS`, `ADI_EMAIL`) is a **confirmed phase-out** β€” every test on those is dropped-with-feature. **Exception:** `blockerReason` is not Toggl-only β€” it is imported by `entries/entries-table.tsx` (blocked-status display) and mirrored as a SQL predicate in `entries/actions.ts` (source comment: "Update all three together"), and it gates invoice eligibility. Its 3 tests port to Python as the blocked-entry predicate. - **TGL-001** β€” `formatDurationSeconds(0)` β†’ `'00:00:00'`. β†’ **dropped-with-feature** - **TGL-002** β€” 3661β†’`'01:01:01'`, 7200β†’`'02:00:00'`, 90β†’`'00:01:30'`. β†’ **dropped-with-feature** - **TGL-003** β€” `splitNaiveTimestamp` accepts both `'T'` and space separators β†’ `{date, time}`. β†’ **dropped-with-feature** (only used inside `buildTogglCsv`) - **TGL-004** β€” `buildTogglCsv([entry])` emits NO header row by default (raw data for the invoicing sheet; first cell is the member value, not `'Member'`). β†’ **dropped-with-feature** - **TGL-005** β€” Source employee email goes in the **Tags** column (last), converted user in the Member column (first). β†’ **dropped-with-feature** - **TGL-006** β€” Email column (2nd) is always `ADI_EMAIL` regardless of source. β†’ **dropped-with-feature** - **TGL-015** β€” `buildTogglCsv(entries, {header:true})` re-enables the header row and it equals `TOGGL_COLUMNS` (14 columns: Member, Email, Client, Project, Project end, Description, Teams, Start date, Stop date, Project start, Start time, Stop time, Duration, Tags). β†’ **dropped-with-feature** - **TGL-016** β€” `{teams:'2026-04'}` option populates column G (index 6, "Teams"). β†’ **dropped-with-feature** - **TGL-017** β€” Teams defaults to `'-'` placeholder when omitted or whitespace-only. β†’ **dropped-with-feature** - **TGL-009** β€” `blockerReason({...eligible, team_member_id:null})` matches `/team member/i` (unknown email). β†’ *Python* (blocked predicate) - **TGL-011** β€” Nulling any required field flags it: `operator`, `client`, `project`, `description`, `end_at`, `source_user_email` (fixture also requires `start_at`; converted_* fields intentionally NOT required post-invoice-migration). β†’ *Python* (blocked predicate) - **TGL-012** β€” A fully-resolved entry β†’ `null` (not blocked). β†’ *Python* (blocked predicate) - **TGL-013** β€” Toggl CSV recombination: Project column = `'{client} : {project}'`, Client column = `operator`. β†’ **dropped-with-feature** - **TGL-014** β€” Client `null` β†’ Project column is just the project name (no separator). β†’ **dropped-with-feature** ### 4. `tests/invoices.test.ts` (11 tests) β€” `lib/invoices.ts` - **INV-D-001** β€” `defaultInvoiceDate(new Date(2026,4,15))` β†’ `'2026-04-30'` (last day of previous month). β†’ *Vitest* (form default pre-fill; add a pytest twin if the create-invoice API computes the default server-side) - **INV-D-002** β€” On the 1st (Jun 1) β†’ `'2026-05-31'` (end of the month that just finished). β†’ *Vitest* (same) - **INV-D-003** β€” January rolls back to previous December (`Jan 10, 2026` β†’ `'2025-12-31'`). β†’ *Vitest* (same) - **INV-D-004** β€” Leap-year February: `Mar 5, 2024` β†’ `'2024-02-29'`; `Mar 5, 2026` β†’ `'2026-02-28'`. β†’ *Vitest* (same) - **INV-P-001** β€” `formatPasteBlock([], date)` β†’ `''`. β†’ *Python* (paste-block export carries over unchanged per judgment call #15; server-generated text) - **INV-P-002** β€” One tab-separated line per breakdown row, fixed 7 columns: `['{client} : {project}', 'Bowden Works Team', amount 2dp, invoice_date, '', 'Bowden Works', 'Labour']`. β†’ *Python* - **INV-P-003** β€” Amount 2dp, no `$`. β†’ *Python* - **INV-P-004** β€” Multiple rows joined with `\n`; row order is the caller's (block does not sort). β†’ *Python* - **INV-P-005** β€” `client:null` / `project:null` β†’ `'(no client) : (no project)'`. β†’ *Python* - **INV-P-006** β€” Separator is `' : '` colon-with-spaces (same convention as the Clockify/Toggl split). β†’ *Python* - **INV-P-007** β€” 5th column (between date and `'Bowden Works'`) is empty string. β†’ *Python* Breakdown row shape used by fixtures: `{client, project, row_count, source_seconds, cost_usd, billout_amount_usd}` (only client, project, billout_amount_usd affect the block). ### 5. `tests/filters.test.ts` (10 tests) β€” `lib/filters.ts` Function: `readFiltersFromSearchParams(searchParams, defaults?)` β†’ `{startBound, endBound, status, batch, q, imported_by, ...}`. Bounds are naive `'YYYY-MM-DD HH:MM:SS'` strings; ranges are half-open `[start, end)`. Clock pinned to 2026-05-15T12:00Z. - **FLT-001** β€” No `date` param β†’ this-month default: `'2026-05-01 00:00:00'` / `'2026-06-01 00:00:00'`. β†’ *Vitest* - **FLT-002** β€” `date=all` β†’ both bounds `null`. β†’ *Vitest* - **FLT-003** β€” `date=this-month` β†’ [first-of-this-month, first-of-next-month). β†’ *Vitest* - **FLT-004** β€” `date=last-month` in January β†’ `'2025-12-01'`..`'2026-01-01'` (year rollover). β†’ *Vitest* - **FLT-005** β€” `date=custom&start=2026-04-10&end=2026-04-15` β†’ endBound `'2026-04-16 00:00:00'` (**inclusive-end-day**: end + 1 day). β†’ *Vitest* - **FLT-006** β€” `status` passes through (`transferred`, `blocked`), defaults to `'all'` (server applies the actual blocker filter). β†’ *Vitest* (note: `transferred` status value dies with the Toggl phase-out; `blocked`/`pending`/`all` survive) - **FLT-007** β€” `batch=all` β†’ `null`; any other string preserved. β†’ *Vitest* - **FLT-008** β€” Empty/whitespace `q` β†’ `null`; non-empty preserved. β†’ *Vitest* - **FLT-009** β€” `imported_by` passes through as string; empty/absent β†’ `null`. β†’ *Vitest* - **FLT-010** β€” A `defaults` argument (`{date:'all', status:'pending'}`) overrides built-in defaults when the URL has no params (this is how `/entries` gets its `all`+`pending` default). β†’ *Vitest* Rebuild note: in the SPA+FastAPI split, URLβ†’filter-state parsing stays Vitest, but the **date-window computation and half-open semantics must also be pinned by pytest** at the API layer, since the API now owns the query (04-architecture: filter serialization is named Vitest turf; the SQL predicate equivalence of `status=blocked` is ENT-016's territory). ### 6. `tests/format.test.ts` (9 tests) β€” `lib/format.ts` All null/undefined inputs render the em-dash `'β€”'`. - **FMT-001** β€” `fmtHoursDecimal`: 3600β†’`'1.00'`, 5400β†’`'1.50'`, nullβ†’`'β€”'`. β†’ *Vitest* - **FMT-002** β€” `fmtHoursMinutes`: 5400β†’`'1:30'`, 3600β†’`'1:00'`, 0β†’`'0:00'`. β†’ *Vitest* - **FMT-003** β€” `fmtHoursMinutes` rounds to nearest minute: 5430β†’`'1:31'`, 5429β†’`'1:30'`. β†’ *Vitest* - **FMT-004** β€” `fmtHoursMinutes(null|undefined)` β†’ `'β€”'`. β†’ *Vitest* - **FMT-005** β€” `fmtUsd`: 1593.89β†’`'$1,593.89'` (thousands separator), 0β†’`'$0.00'`, nullβ†’`'β€”'`. β†’ *Vitest* - **FMT-006** β€” `fmtMoneyHours(5400, 14)` β†’ `'$21.00 (1:30)'` (dollars = seconds/3600 Γ— rate, plus hh:mm). β†’ *Vitest* - **FMT-007** β€” `fmtMoneyHours` β†’ `'β€”'` for 0 / null / undefined seconds. β†’ *Vitest* - **FMT-009** β€” `fmtHoursMinutesSeconds` exact, no rounding: 5400β†’`'1:30:00'`, 3661β†’`'1:01:01'`, 0β†’`'0:00:00'`, null/undefinedβ†’`'β€”'`. β†’ *Vitest* - **FMT-008** β€” `fmtTimestampShort`: truncates seconds and normalizes `'T'`β†’space: `'2026-05-20T09:30:00'` and `'2026-05-20 09:30:00'` both β†’ `'2026-05-20 09:30'`; nullβ†’`'β€”'`. β†’ *Vitest* --- ## `docs/tests-catalog.md` reconciliation The catalog is the declared "stack-agnostic checklist the new test suite must cover" with stable IDs (βœ… = vitest test exists, 🟑 = catalog only). It matches the FMT and FLT sections exactly. It is **stale** everywhere else: ### Catalog staleness (catalog vs actual test files β€” test files win) 1. **CLK section stops at CLK-022.** Missing from the catalog but present in `clockify.test.ts`: CLK-026/027/028 (time auto-detect), CLK-029–034 + 032b/032c/032d (`detectDateFormat` incl. the tightest-span regression), CLK-040–044 (`formatDateRangeLabel`), CLK-023 + the duplicate-ID CLK-021/CLK-022 (`billout_rate_usd` surfacing, `computeBilloutAmount`) β€” ~20 tests uncataloged. 2. **CLK ID collisions.** `CLK-021` is used twice (splitClientProject AND billout_rate_usd surfacing); `CLK-022` is used twice (parseClockifyCsv split AND computeBilloutAmount nulls). Renumber during the port. 3. **One un-ID'd test** (`parseDurationSeconds` garbage β†’ null). 4. **TGL-004 meaning drifted.** Catalog: "emits the 13 expected columns in the documented order." Actual test: "emits no header row by default"; the column-order assertion moved to TGL-015 (`{header:true}` equals `TOGGL_COLUMNS`) β€” and `TOGGL_COLUMNS` now has **14** columns (Teams was added), not 13. 5. **TGL-007, TGL-008, TGL-010 are ghosts.** Catalog marks them βœ… (Billable Yes/No; Task empty; blockerReason flags missing converted duration) but no such tests exist in `toggl.test.ts`. TGL-010's removal is deliberate β€” post-invoice-migration, converted_* fields are no longer eligibility requirements (comment in both the test file and `lib/toggl.ts`). TGL-007/008 assertions were folded away when the header/format tests were reworked. 6. **TGL-011 description stale.** Catalog says it flags missing "client / project / description / start / end / source email / converted user"; the actual test checks `operator`, client, project, description, `end_at`, `source_user_email` β€” operator added, converted-user and start checks gone. 7. **TGL-015/016/017 uncataloged** (header option, Teams column, `'-'` placeholder). 8. **The INV-* suite (11 tests) is entirely absent from the catalog.** `invoices.test.ts` exists and passes but has no catalog section. 9. **The CCE-* suite (29 tests) is entirely absent from the catalog.** The CC-expenses feature doc lists ~29 βœ… CCE IDs + 5 🟑 CCE-A IDs, none merged into the catalog (extraction flag #3 β€” if the catalog is the rebuild checklist, these would silently drop). ### Catalog 🟑 list β€” behaviors with NO test (integration/manual only) These 44 IDs are cataloged but untested because they need the DB or Supabase Auth. Verbatim from `docs/tests-catalog.md`, with the rebuild owner added. Legend: **py** = pytest+httpx API/domain test, **pw** = Playwright smoke, **drop** = dies with a phased-out feature, **rewrite** = semantic survives in a different mechanism. | ID | Behavior (one line) | Rebuild owner | |---|---|---| | AUTH-001 | First-login redirects to /change-password | **drop** (must_change_password dies with id-auth Pattern B) | | AUTH-002 | /change-password flips `must_change_password=false` then redirects to / | **drop** (same) | | RBAC-001 | Adi (manager) sees /transfer but no "Mark as transferred" button | **drop** β€” also STALE: contradicts RBAC-005/006 and the permissions matrix; extraction flag #2 says retire it | | RBAC-002 | Rian as super admin sees the "View as β–Ύ" switcher | **py + pw** (view-as survives as app-layer impersonation) | | RBAC-003 | View-as cookie surfaces banner and disables owner-only buttons | **pw** | | IMP-001 | Uploading a Clockify CSV creates a batch + N entries | **py** (+pw smoke: importβ†’resolveβ†’commit) | | IMP-002 | Unknown emails leave `team_member_id` NULL on those rows | **py** (rewrite to engagement resolution) | | IMP-003 | Cannot delete an import batch if any entry is transferred | **rewrite β†’ py**: "transferred" is dead; the surviving lock is *invoiced* entries block batch delete | | TRN-001 | Mark-selected assigns one shared `transfer_batch_id` | **drop** (Toggl transfer) | | TRN-002 | Undo batch clears transferred_at/by/transfer_batch_id | **drop** | | TRN-003 | Download CSV does not write to the DB | **drop** | | SUM-001 | `summary_by_project` RPC: one row per project, this/prev/all-time seconds | **py** (RPC semantics port to a service query) | | SUM-002 | Summary dollars = (converted_seconds/3600) Γ— adi_rate | **rewrite β†’ py**: converted seconds die; dollars come from stamped billout/cost | | ENT-005 | Bulk Mark explicit IDs: one shared batch id; transferred+blocked skipped | **drop** | | ENT-006 | Bulk Mark "all matching filter" respects active filters | **drop** (but see ENT-020 β€” the scope-derivation invariant survives) | | ENT-007 | Bulk Unmark clears transfer fields only on transferred rows | **drop** | | ENT-008 | "All dates" applies unbounded filter (strip-on-`all` regression) | **Vitest + py** | | ENT-009 | Cost column = (converted_duration_seconds/3600) Γ— org.default_hourly_rate_usd | **rewrite β†’ py**: cost is a locked-at-write stamp, not a live computation (LOCK-001 semantics won) | | ENT-010 | Sortable column headers update ?sort=/?dir= and reorder | **Vitest** (table interaction hook) + **py** (order-by) | | TEAM-003 | Base-rate update writes org.default_hourly_rate_usd, reflected in Effective rate / Cost / Summary | **rewrite β†’ py**: rates become engagement compensation terms | | TEAM-004 | Effective rate displays as base Γ— proportion, read-only | **drop** (proportions die, ADR #034) | | LOCK-001 | `billout_cost_usd` captured at import, not changed by later rate updates | **py** β€” load-bearing invariant (02-domain-model Β§5) | | LOCK-002 | Editing billout duration recomputes `billout_cost_usd` at CURRENT rate | **py** β€” explicit re-stamp invariant | | LOCK-003 | Re-resolve recomputes from current proportion Γ— current base rate | **py** (rewrite minus proportion) | | RBAC-004 | Manager cannot edit/delete/re-resolve a transferred entry; UI shows "Locked" | **rewrite β†’ py**: lock trigger becomes *invoiced*, not transferred | | RBAC-005 | No `canTransfer` β†’ no Transfer nav link | **drop** (page dies) β€” capability truth-table grid replaces it | | RBAC-006 | Direct /transfer as non-transfer user redirects to / | **drop** (same) | | SUM-003 | Summary hides projects with zero hours in both prev+this (still in all-time) | **py** | | ENT-011 | Totals card (`entries_filter_totals` RPC) covers FULL filter match across pages | **py** | | ENT-012 | Project/User/Source-email filter narrows list AND totals; bulk respects | **py** | | ENT-013 | Client filter narrows list AND totals; bulk respects | **py** | | ENT-014 | `bulkUpdateFields`: empty input no-ops that field; values update client/project/converted_user | **py** (converted_user part drops) | | ENT-015 | `bulkUpdateFields` as manager touches only `transferred_at IS NULL` rows | **rewrite β†’ py** (invoiced-lock instead) | | ENT-016 | `status=blocked` is a SQL-level predicate packed onto page 1, not JS post-filter | **py** β€” pairs with the TGL-009/011/012 predicate port | | ENT-017 | Operator column + filter sort/filter correctly | **rewrite β†’ py** (operator becomes an FK-resolved party, not a text column) | | ENT-018 | Filter pivots from `entries_filter_options` RPC reflect the filtered set | **py** | | COM-101 | Nav-bell comment visible to every org member | **py** | | COM-102 | Resolve sets resolved_at/resolved_by; badge decrements | **py + pw** | | COM-103 | Delete allowed for author or super_admin; others RLS-denied | **py** (authz grid β€” RLS denial becomes service-layer denial) | | ENT-020 | Bulk with all_matching=1 + status + operator only updates rows matching BOTH (missing-applyBulkFilter-dimension regression) | **py** β€” the "3,274-rows-from-a-353-row-selection" incident; judgment call #11 makes scope-derivation architectural | | ENT-021 | Bulk aborts when pre-count β‰  user-confirmed `expected_count`; nothing written | **py** β€” expected-count precondition is an R1 requirement (judgment call #11) | | ENT-022 | Per-row Edit Save persists via `updateEntry` (nested-form silent-fail regression) | **pw** (form wiring) + **py** (updateEntry) | | ENT-023 | Fresh /entries applies date=all + status=pending to query AND dropdown UI | **Vitest + pw** | | ENT-024 | "Clear filters" strips every filter param and resets to defaults | **Vitest** | | SUM-004 | `summary_by_project` returns one row per (operator, client, project) TRIPLE; three sortable columns | **py** | Also cataloged only in the CC-expenses feature doc (not in tests-catalog.md), all 🟑: **CCE-A-001** (re-paste preserves order via `line_index`), **CCE-A-002** (per-row Save updates only that row), **CCE-A-003** (delete batch cascades to lines), **CCE-A-004** (auto-rule fires at paste, pre-fills assignment), **CCE-A-005** (new rule doesn't touch existing lines β€” rules apply only at paste time). All β†’ **py**. --- ## Rebuild deltas (vs 00 / 02 / 04 / 07) - **Where tests live in the rebuild** (04-architecture): pytest+httpx for API/domain β€” authz truth-table grids (actor Γ— action Γ— resource, per 02-domain-model Β§4: capability functions `canTransfer`, `canManageTeam`, `canManageImports`, `canEditEntry`, `canSeeIncome`, `canEnterViewAs` port with **identical truth tables** β€” meaning the *effective enforcement*; for canManageTeam that is the RLS rule, not the dead function (02 Β§4) β€” before any new semantics), money stamping, and the import pipeline "with the CLK-* fixtures". Vitest for frontend logic (filter serialization, table interaction hooks). Playwright smoke for importβ†’resolveβ†’commit, entry edit, invoice create+lock β€” behavior assertions, not render assertions. CI runs pytest + vitest + tsc from R0. - **`06-test-plan.md` does not exist yet** (04-architecture references it); this inventory is its primary input. - **The diff harness is a separate oracle** (R0, built FIRST): it compares data exports against the live app (invoice totals, monthly summaries, per-project aggregates). It is not part of this unit-test inventory but covers much of what the 🟑 list leaves untested. - **Counts**: 116 vitest tests verified by direct count (29 CCE + 43 CLK + 10 FLT + 9 FMT + 11 INV + 14 TGL), matching 00-vision-and-scope's "the current 116". Judgment call #14 says the import pipeline ports "with its 44-test suite" β€” `clockify.test.ts` has **43** `it()` blocks; treat 43 as correct (44 is off by one, or counts the planned re-ID of the un-ID'd duration test). - **Net port math**: of 116 existing tests, ~14 drop outright (11 Toggl-CSV/format tests + CLK-019 + CLK-020, plus TGL-003's helper), ~3 port with rewritten semantics (CLK-018, CLK-021-dup, eligibility fixtures), and the remaining ~99 port mechanically (Python: CLK, CCE, INV-P, TGL-009/011/012; Vitest: FMT, FLT, INV-D). - **Fixture-language rule** (04-architecture): rebuilt fixtures use the real business graph (Adi, Brentwood, PlusROI), not user1/ project1 β€” the current tests' anonymous fixtures (`dev@example.com`, 'Test Client') should be re-expressed on the seed slices. - **Explicitly new-in-rebuild test surface** (no current equivalent): the authz capability grids, audit-log writes on every mutation, Pattern B auth middleware (with `AUTH_DEV_USER` dev-mode stub, judgment call #13), structured `{error_code, summary, detail}` errors, and zero-cost-worker rates ("Gary: cost $0, billout $25" β€” judgment call #12, "tested explicitly"). ## Quirks a rebuilder must not "fix" silently - `' : '` (space-colon-space) is the universal client/project separator, split on FIRST occurrence, strict (no-space `'Foo:Bar'` does NOT split). Appears in CLK-021, TGL-013, INV-P-006, CCE-X-002. - Naive local timestamps everywhere (`'YYYY-MM-DD HH:MM:SS'`, no timezone) β€” CLK-014, TGL-003, FLT bounds. 02-domain-model carries this over deliberately. - Half-open date ranges `[start, end)` with inclusive-end-day for custom ranges (FLT-005). - Em-dash `'β€”'` (not `'-'`, not empty) for null display values (FMT). - 2dp rounding at format time for money; `computeBilloutAmount` rounds at computation time (CLK-023). - `detectDateFormat`'s tightest-span heuristic and its "don't guess" null (CLK-032b/c/d) encode a real 2026-06-09 data-corruption incident β€” port the fixtures verbatim.