Core and API: multi-pattern presence model, migration and routes (schema 9) #231

2026-09-24T19:36:23-04:00

dev_agent commented

2026-09-24 19:36:23 -04:00

Task e63a8e64 (Feature: Multiple presence patterns per child), task-branch PR into the feature branch.

Done-condition: a child holds zero or more patterns (name, kind home|away, cycle 1..4, anchor, per-week weekday set); overrides win; away beats home; a child with no patterns is present every day; a child with home patterns but none covering the date is absent (old single-schedule behaviour); migration 9 converts every presence_schedules row into a home pattern named "Home schedule", retires the old table, and proves presence results unchanged for a pre-migration fixture; admin CRUD routes for patterns with 422 naming the field and 404 for unknown child/pattern; old one-per-child route removed; overrides routes unchanged; TZ=UTC uv run pytest -q passes.

Developer evidence at ef6940d937: TZ=UTC uv run pytest -q = 1976 passed, 0 failed; mutation checks confirm the suite catches a wrong default rule (15 fail), ISO-week arithmetic (10 fail) and home-before-away (2 fail); new tests/test_presence_patterns.py covers the owner scenario week by week, the 2026->2027 boundary, the 2028 leap day, no-patterns, overrides-beat-patterns and away-beats-home.

Spec amendment flagged by the developer and accepted by the orchestrator: rule 4 is "a child with at least one home pattern, none covering the date, is absent", not "otherwise present" — otherwise a home pattern could never make a child absent and the unchanged-migration test could not pass.

Task e63a8e64 (Feature: Multiple presence patterns per child), task-branch PR into the feature branch. Done-condition: a child holds zero or more patterns (name, kind home|away, cycle 1..4, anchor, per-week weekday set); overrides win; away beats home; a child with no patterns is present every day; a child with home patterns but none covering the date is absent (old single-schedule behaviour); migration 9 converts every presence_schedules row into a home pattern named "Home schedule", retires the old table, and proves presence results unchanged for a pre-migration fixture; admin CRUD routes for patterns with 422 naming the field and 404 for unknown child/pattern; old one-per-child route removed; overrides routes unchanged; TZ=UTC uv run pytest -q passes. Developer evidence at ef6940d937502c20feed5f25059097d282621e14: TZ=UTC uv run pytest -q = 1976 passed, 0 failed; mutation checks confirm the suite catches a wrong default rule (15 fail), ISO-week arithmetic (10 fail) and home-before-away (2 fail); new tests/test_presence_patterns.py covers the owner scenario week by week, the 2026->2027 boundary, the 2028 leap day, no-patterns, overrides-beat-patterns and away-beats-home. Spec amendment flagged by the developer and accepted by the orchestrator: rule 4 is "a child with at least one home pattern, none covering the date, is absent", not "otherwise present" — otherwise a home pattern could never make a child absent and the unchanged-migration test could not pass.

dev_agent added 1 commit 2026-09-24 19:36:24 -04:00

Replace presence schedules with multiple presence patterns per child (schema 9) ef6940d937

A child now carries zero or more home/away presence patterns instead of
at most one schedule. Migration 9 creates presence_patterns, copies every
presence_schedules row across as a 'home' pattern named "Home schedule"
(cycle, anchor and pattern verbatim), and drops the old table in one
transaction. Migration 1 keeps creating the retired table from a frozen
copy of its DDL; schema.py no longer declares it.

PresenceEngine resolves override -> any covering away pattern (absent)
-> any covering home pattern (present) -> otherwise absent when the child
has a home pattern, else present. The home-pattern default keeps every
migrated child's presence unchanged.

The admin plane's one-per-child presence-schedule routes are replaced by
pattern CRUD: GET/POST /children/{child_id}/presence-patterns and
PATCH/DELETE /presence-patterns/{pattern_id}.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

overseer referenced this issue from a commit

2026-09-24 19:45:55 -04:00

Move the Admin PWA Schedule tab to multi-pattern presence

overseer added 3 commits 2026-09-24 19:45:55 -04:00

Feature 19 follow-up: calendar week starts on Sunday (#230 ) d3c52974f5

Co-authored-by: dev_agent <17+dev_agent@code.cubecraftcreations.com>
Co-committed-by: dev_agent <17+dev_agent@code.cubecraftcreations.com>

Merge remote-tracking branch 'origin/dev' into task/multi-pattern-presence-model_e63a8e64-2f1a-4818-baf2-0d0431bcd014 1c2a7222b3

Move the Admin PWA Schedule tab to multi-pattern presence 39417615a2

The admin client still called the removed one-per-child
/presence-schedule route, so the Schedule tab could not load or save
(review finding F1 on PR #231). The client now uses the presence-pattern
routes: list and create under /children/{id}/presence-patterns, and
PATCH / DELETE on /presence-patterns/{id}.

The Schedule screen lists every pattern of the selected child (name,
Home/Away, cycle length, weekday set) with Add, Edit and Delete, and a
pushed pattern editor sets name, kind, cycle length, anchor date and the
per-week weekday set. Edits send only the changed fields. The month grid
shows the combined presence with the core's first-match rule: override,
then any covering away pattern, then any covering home pattern, then
away when the child has a home pattern, else home. The grid keeps the
Sunday-first week; stored weekdays stay Monday=0.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

overseer added 1 commit 2026-09-24 19:55:47 -04:00

Reject non-integer weekday values in presence pattern requests b91ecbd5f6

The create and edit pattern models typed weekday values as lax int, so
JSON true, "1" and 1.0 were coerced to weekday 1. Type them StrictInt
so each is a 422 naming the pattern field, and add create and edit
regression cases for bool, string, float and out-of-range weekdays.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

overseer added 1 commit 2026-09-24 20:03:06 -04:00

Allow saving presence patterns that cover no weekday 0d763519a0

The pattern editor disabled Save unless a weekday was selected, but an
all-empty weekday set is valid in the core model and the API, and
migration 9 produces one for a legacy all-absent schedule. Save now only
requires a name (plus a valid anchor date and a change to save).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

review_agent approved these changes

2026-09-24 20:08:37 -04:00

review_agent left a comment

Independent Codex gpt-5.6-terra review completed for this exact commit.

dev_agent merged commit 88b4733a2d into feature/c1f79a07-da1e-4b29-bdb9-946bb33499ce-multiple-presence-patterns

2026-09-24 20:09:57 -04:00

dev_agent referenced this pull request

2026-09-24 20:11:33 -04:00

Feature: multiple presence patterns per child (schema 9) #232

Sign in to join this conversation.