4.0 regression map: which test guards which check of design 12
Design § 12 ran three scripts and a dump build against the design's DDL in SQLite. This page maps every check of the three scripts
to the test in the suite that guards it now, against the real brickwork/store4/*.sql, so a later reader can see each failure the
design's review demonstrated (§ 13) guarded. Written for M.9 (issue #111). The scripts and their logs are in the design's scratchpad
(not in the repo); the labels below are the logs' own.
tests/test_store4_schema.py::test_halves_equal_the_design_ddl holds the two halves of the shipped DDL equal to schema-4.0.sql, so a
check that passed on the design's copy passes on what ships.
| Script | Checks | Covered before M.9 | Ported in M.9 | Not portable |
|---|---|---|---|---|
test_constraints.py (constraints.log) |
97 | 97 (M.1) | 0 | 0 |
test_procedures.py (procedures.log) |
16 | 15 (M.2, M.7) | 1 (A3, second half) | 0 |
test_journey.py (journey.log) |
1 replay, 8 steps | step checks in module tests; no replay | 1 test, through the app's writers | 0 (3 rows differ from the journey text, see below; setup's system flags and event were added to match it) |
1. Constraint checks (97)
C1 to C9 run against catalogue.sql, L0 to L15 against brickwork.sql; each test file function is one database, in the log's order.
The review case each guards is in section 4.
| # | Check | Test |
|---|---|---|
| 1 | C1 colour 1 = Rebrickable 9999 | tests/test_store4_schema.py::test_catalogue_constraints |
| 2 | C1 colour 2 = Rebrickable -1 | tests/test_store4_schema.py::test_catalogue_constraints |
| 3 | C1 fixed id 1 cannot hold another colour | tests/test_store4_schema.py::test_catalogue_constraints |
| 4 | C1 Rebrickable 9999 cannot take another id | tests/test_store4_schema.py::test_catalogue_constraints |
| 5 | C2 a part cannot be classified by theme | tests/test_store4_schema.py::test_catalogue_constraints |
| 6 | C2 namespace kind must match the namespace | tests/test_store4_schema.py::test_catalogue_constraints |
| 7 | C2 born key unique (provider, namespace, key) | tests/test_store4_schema.py::test_catalogue_constraints |
| 8 | C2 same key in another namespace is another item | tests/test_store4_schema.py::test_catalogue_constraints |
| 9 | C3 kind change keeps the id (theme guess corrected by BrickLink) | tests/test_store4_schema.py::test_catalogue_constraints |
| 10 | C3 id 15 still the same born key | tests/test_store4_schema.py::test_catalogue_constraints |
| 11 | C3 a part cannot change kind | tests/test_store4_schema.py::test_catalogue_constraints |
| 12 | C4 catalogue cannot mint into the local range | tests/test_store4_schema.py::test_catalogue_constraints |
| 13 | C4 catalogue adopts a local id (set fetched via API before the dump) | tests/test_store4_schema.py::test_catalogue_constraints |
| 14 | C5 one inventory per (item, version) | tests/test_store4_schema.py::test_catalogue_constraints |
| 15 | C5 one latest inventory per item | tests/test_store4_schema.py::test_catalogue_constraints |
| 16 | C6 inventory line key is the primary key (no surrogate id) | tests/test_store4_schema.py::test_catalogue_constraints |
| 17 | C6 inventory_part is WITHOUT ROWID | tests/test_store4_schema.py::test_catalogue_constraints |
| 18 | C7 leg rule: 970c03 in Black -> BrickLink 970c00 (colour-scoped) | tests/test_store4_schema.py::test_catalogue_constraints |
| 19 | C7 leg rule: 970c03 in other colours -> 970c11 (unscoped) | tests/test_store4_schema.py::test_catalogue_constraints |
| 20 | C7 duplicate unscoped link refused (NULL colour coalesced) | tests/test_store4_schema.py::test_catalogue_constraints |
| 21 | C7 two preferred links for one (item, provider, colour) | tests/test_store4_schema.py::test_catalogue_constraints |
| 22 | C7 a catalogue link needs an item | tests/test_store4_schema.py::test_catalogue_constraints |
| 23 | C7 status 'unmatched' no longer exists on links | tests/test_store4_schema.py::test_catalogue_constraints |
| 24 | C8 BrickLink 72 <- Rebrickable 1051 (inverse preferred) | tests/test_store4_schema.py::test_catalogue_constraints |
| 25 | C8 BrickLink 72 <- Rebrickable 313 too | tests/test_store4_schema.py::test_catalogue_constraints |
| 26 | C8 only one inverse per BrickLink colour | tests/test_store4_schema.py::test_catalogue_constraints |
| 27 | C9 catalogue foreign_key_check clean | tests/test_store4_schema.py::test_catalogue_constraints |
| 28 | L0 setup seeds colours 1 and 2 | tests/test_store4_schema.py::test_collection_constraints |
| 29 | L0 a second Rebrickable 9999 colour | tests/test_store4_schema.py::test_collection_constraints |
| 30 | L0 a BrickLink-only colour is local, above the base | tests/test_store4_schema.py::test_collection_constraints |
| 31 | L1 a local item below the base | tests/test_store4_schema.py::test_collection_constraints |
| 32 | L1 an unidentified item: local, no born key | tests/test_store4_schema.py::test_collection_constraints |
| 33 | L1 a catalogue item needs a Rebrickable born key | tests/test_store4_schema.py::test_collection_constraints |
| 34 | L2 job source 'import' (pipeline) is a seeded source | tests/test_store4_schema.py::test_collection_constraints |
| 35 | L2 copy on Rebrickable v1 with BrickLink list pending | tests/test_store4_schema.py::test_collection_constraints |
| 36 | L2 pending without its job | tests/test_store4_schema.py::test_collection_constraints |
| 37 | L2 source 'bricklink' needs a fetched list | tests/test_store4_schema.py::test_collection_constraints |
| 38 | L2 set_preference 'bricklink' before any list was fetched | tests/test_store4_schema.py::test_collection_constraints |
| 39 | L2 a copy cannot use another set's fetched list (composite FK) | tests/test_store4_schema.py::test_collection_constraints |
| 40 | L2 two latest fetched lists per (item, source) | tests/test_store4_schema.py::test_collection_constraints |
| 41 | L2 a second fetch is seq 2, latest; seq 1 stays for copies built from it | tests/test_store4_schema.py::test_collection_constraints |
| 42 | L2 fetched line key is the primary key | tests/test_store4_schema.py::test_collection_constraints |
| 43 | L3 copy_part cannot name another copy's figure (composite FK) | tests/test_store4_schema.py::test_collection_constraints |
| 44 | L3 copy_part of its own figure | tests/test_store4_schema.py::test_collection_constraints |
| 45 | L3 one row per line key in a copy | tests/test_store4_schema.py::test_collection_constraints |
| 46 | L4 missing=2, damaged=1 of 2 (today's per-field bounds, checker.py:454-468) | tests/test_store4_schema.py::test_collection_constraints |
| 47 | L4 missing above quantity | tests/test_store4_schema.py::test_collection_constraints |
| 48 | L5 the initial design's 'replace the rows' fails once a Take/allocation exists | tests/test_store4_schema.py::test_collection_constraints |
| 49 | L5 switch by in-place update + retire commits with Take, allocation and check session in place | tests/test_store4_schema.py::test_collection_constraints |
| 50 | L5 switching back re-activates the same row id | tests/test_store4_schema.py::test_collection_constraints |
| 51 | L5 stock_transfer still points at row 104 | tests/test_store4_schema.py::test_collection_constraints |
| 52 | L6 a live catalogue item cannot be merged away | tests/test_store4_schema.py::test_collection_constraints |
| 53 | L6 a local item cannot be merged while referenced | tests/test_store4_schema.py::test_collection_constraints |
| 54 | L6 local BrickLink-born 970c00 merged into catalogue 970c03 after moving references | tests/test_store4_schema.py::test_collection_constraints |
| 55 | L6 a redirect must match its merge | tests/test_store4_schema.py::test_collection_constraints |
| 56 | L6 INSERT naming a merged item (copy_part) | tests/test_store4_schema.py::test_collection_constraints |
| 57 | L6 UPDATE to a merged item (copy_part) | tests/test_store4_schema.py::test_collection_constraints |
| 58 | L6 INSERT naming a merged item (stock_part, not guarded before) | tests/test_store4_schema.py::test_collection_constraints |
| 59 | L6 INSERT naming a merged item (price_observation, not guarded before) | tests/test_store4_schema.py::test_collection_constraints |
| 60 | L6 materialising a catalogue line still works (no catalogue row names a merged id) | tests/test_store4_schema.py::test_collection_constraints |
| 61 | L6 a retired catalogue item can be merged only after its references move | tests/test_store4_schema.py::test_collection_constraints |
| 62 | L6 a retired, unreferenced catalogue item (a Rebrickable rename) can be merged | tests/test_store4_schema.py::test_collection_constraints |
| 63 | L6 merge into an item of another kind | tests/test_store4_schema.py::test_collection_constraints |
| 64 | L7 3023 White used, no lot, no storage | tests/test_store4_schema.py::test_collection_constraints |
| 65 | L7 the identical line again (NULL lot and storage) | tests/test_store4_schema.py::test_collection_constraints |
| 66 | L7 3023 White with condition NULL | tests/test_store4_schema.py::test_collection_constraints |
| 67 | L7 the same with condition NULL again | tests/test_store4_schema.py::test_collection_constraints |
| 68 | L7 a new live line after the old one was archived | tests/test_store4_schema.py::test_collection_constraints |
| 69 | L7 two identical loose figures stay two rows | tests/test_store4_schema.py::test_collection_constraints |
| 70 | L7 a loose figure needs a part-list source (G6) | tests/test_store4_schema.py::test_collection_constraints |
| 71 | L7 a loose part has no part-list source | tests/test_store4_schema.py::test_collection_constraints |
| 72 | L8 item 2339 has no BrickLink id: item_provider_status | tests/test_store4_schema.py::test_collection_constraints |
| 73 | L8 a migrated figure with no candidate: to_check without a key | tests/test_store4_schema.py::test_collection_constraints |
| 74 | L8 item_link without an item | tests/test_store4_schema.py::test_collection_constraints |
| 75 | L8 item_link status 'unmatched' | tests/test_store4_schema.py::test_collection_constraints |
| 76 | L8 same_number candidate as to_check link | tests/test_store4_schema.py::test_collection_constraints |
| 77 | L8 the same link twice (NULL colour scope coalesced) | tests/test_store4_schema.py::test_collection_constraints |
| 78 | L8 'manual' only from the user | tests/test_store4_schema.py::test_collection_constraints |
| 79 | L9 an unreferenced theme-guessed item changes kind, same id | tests/test_store4_schema.py::test_collection_constraints |
| 80 | L9 a set referenced by a copy cannot become gear | tests/test_store4_schema.py::test_collection_constraints |
| 81 | L9 gear held as stock cannot silently become a set | tests/test_store4_schema.py::test_collection_constraints |
| 82 | L9 ... reconcile records it instead (reason 'kind_differs') | tests/test_store4_schema.py::test_collection_constraints |
| 83 | L9 UPDATE copy.item_id to a part | tests/test_store4_schema.py::test_collection_constraints |
| 84 | L9 INSERT copy of a gear item | tests/test_store4_schema.py::test_collection_constraints |
| 85 | L9 UPDATE stock.kind away from the item kind | tests/test_store4_schema.py::test_collection_constraints |
| 86 | L9 INSERT wish with a kind that is not the item kind | tests/test_store4_schema.py::test_collection_constraints |
| 87 | L10 preferred 970c11 for 970c03 in Red (colour-scoped row) | tests/test_store4_schema.py::test_collection_constraints |
| 88 | L10 preferred 970c11 for 970c03 in any other colour | tests/test_store4_schema.py::test_collection_constraints |
| 89 | L10 a second preferred link for (970c03, bricklink, Black) | tests/test_store4_schema.py::test_collection_constraints |
| 90 | L11 wish for a set (colour NULL) | tests/test_store4_schema.py::test_collection_constraints |
| 91 | L11 the same live wish again (NULL colour coalesced) | tests/test_store4_schema.py::test_collection_constraints |
| 92 | L12 events are append-only | tests/test_store4_schema.py::test_collection_constraints |
| 93 | L13 step error 'unidentified' (G9) | tests/test_store4_schema.py::test_collection_constraints |
| 94 | L13 unknown step error code | tests/test_store4_schema.py::test_collection_constraints |
| 95 | L14 a system flag cannot be archived | tests/test_store4_schema.py::test_collection_constraints |
| 96 | L15 collection foreign_key_check clean | tests/test_store4_schema.py::test_collection_constraints |
| 97 | L15 collection integrity_check ok | tests/test_store4_schema.py::test_collection_constraints |
L2 "job source 'import' is a seeded source" is an INSERT of a job with source='import' under foreign_keys=ON (the test's own SELECT 1 is the
design's, the insert before it is the check).
2. Procedure checks (16)
| Check | Test |
|---|---|
| A1 re-download (rows reordered, one vanished, one new): every old key keeps its id | tests/test_store4_builder.py::test_a1_redownload_keeps_ids_retires_and_revivestests/test_store4_reconcile.py::test_done_when_1_a_redownload_keeps_every_id_the_collection_uses |
| A1 the vanished key is carried, retired, same id | tests/test_store4_builder.py::test_a1_redownload_keeps_ids_retires_and_revives |
| A1 the new key is numbered above the high-water mark | tests/test_store4_builder.py::test_a1_redownload_keeps_ids_retires_and_revives |
| A1 catalogue_id kept | tests/test_store4_builder.py::test_a1_redownload_keeps_ids_retires_and_revives |
| A2 rebuild without the old catalogue: every id the collection references is kept | tests/test_store4_builder.py::test_a2_catalogue_lost_rebuild_from_the_collectiontests/test_store4_reconcile.py::test_done_when_2_a_rebuild_from_scratch_seeded_by_the_collection_keeps_the_ids |
| A2 unreferenced keys get numbers above the collection's high-water mark | tests/test_store4_builder.py::test_a2_catalogue_lost_rebuild_from_the_collection |
| A2 and the pair binds (same catalogue_id) | tests/test_store4_builder.py::test_a2_catalogue_lost_rebuild_from_the_collection |
| A3 another install numbers differently (decided) and has another catalogue_id | tests/test_store4_regression.py::test_a3_an_install_with_another_history_numbers_the_same_dump_differentlytests/test_store4_builder.py::test_a3_another_install_has_a_new_catalogue_id_and_the_same_ids |
| A4 a set first fetched through the API is adopted by the next dump with its local id | tests/test_store4_builder.py::test_a4_a_local_item_is_adopted_by_the_next_buildtests/test_store4_reconcile.py::test_a4_an_adopted_local_item_is_flipped_by_reconcile_and_a_merged_one_is_listed |
| B1 colour-scoped inverse (970c00/Black to 970c03, 970c11/Red to 970c03, 970c00/Red to 970c05) | tests/test_store4_mint.py::test_leg_rule_candidates_are_scoped_to_the_minted_colourtests/test_store4_identity.py::test_inverse_finds_item_and_colour_with_the_leg_rule |
| B1 BrickLink colour 72 resolves to the inverse-preferred Rebrickable colour | tests/test_store4_mint.py::test_colour_links_prefer_the_inverse_preferred_onetests/test_store4_identity.py::test_inverse_bricklink_colour_72_takes_the_preferred_rebrickable_colour |
| B2 sticker sheets: the first keeps the catalogue item, the second becomes a local item | tests/test_store4_mint.py::test_b3_the_sticker_sheets_are_kept_aparttests/test_store4_migrate.py::test_sticker_sheets_stay_aparttests/test_store4_identity.py::test_inverse_keeps_sticker_sheets_apart_and_is_injectivetests/test_store4_regression.py::test_b5_b6_eight_lines_of_one_copy_insert_under_the_line_key_and_read_back_in_bricklink_space |
| B3 an unmapped BrickLink part becomes a local BrickLink-born item | tests/test_store4_mint.py::test_rule_5_no_candidate_is_a_local_bricklink_born_itemtests/test_store4_regression.py::test_b5_b6_eight_lines_of_one_copy_insert_under_the_line_key_and_read_back_in_bricklink_space |
| B4 the mint is injective on (item, colour) | tests/test_store4_mint.py::test_b5_the_mint_is_injectivetests/test_store4_mint.py::test_injectivity_runs_across_holderstests/test_store4_identity.py::test_inverse_keeps_sticker_sheets_apart_and_is_injective |
| B5 all eight 3.3 lines insert without a unique-key collision | tests/test_store4_regression.py::test_b5_b6_eight_lines_of_one_copy_insert_under_the_line_key_and_read_back_in_bricklink_spacetests/test_store4_migrate.py::test_every_regression_fixture_migrates |
| B6 proof in BrickLink space: the 4.0 lines read back through the mint table equal the 3.3 lines | tests/test_store4_mint.py::test_b6_read_back_in_bricklink_space_equals_the_3_3_keystests/test_store4_migrate_run.py::test_the_proof_passes_on_the_synthetic_storetests/test_store4_migrate_run.py::test_the_proof_passes_on_every_regression_fixturetests/test_store4_regression.py::test_b5_b6_eight_lines_of_one_copy_insert_under_the_line_key_and_read_back_in_bricklink_space |
The numbering of the mint tests differs from the design's (the file's docstring says so): its test_b3_... is the design's B2, its test_b5_...
the design's B4. Rows above use the design's numbers.
3. The journey replay
user-journey.md is replayed by one test, tests/test_store4_regression.py::test_the_user_journey_replays_through_the_apps_writers, on a fixture catalogue (the fixture
dump plus the set 6081-1, its figures and the arch 2339, built at test time by the real builder). Every row comes from a writer of the
app: setup and activation (the pair fixture), core.keys (save, check with a fake transport), the Add routes (review, confirm), the BrickLink
part-list job (store4.jobs), the pipeline's six steps with fake transports, core.identify.store. It ends with foreign_key_check and
integrity_check clean.
Journey step (journey.log) |
Where in the test | Writer |
|---|---|---|
| § 0 setup | "0 and 1.1" block | connect, store4.activate |
| § 1.1 activate: colours, colour links, one event | "0 and 1.1" block | store4.activate |
| § 1.2 key saved and tested | "1.2" block | core.keys.save, core.keys.check |
§ 2 copy of 6081-1: items projected, figure lines, copy_part rows with spares |
"2" block | Add review and confirm |
| § 2 jobs: picture row, price observation, set link upgraded by the reply | "the job" block | pipeline steps pictures, market_price |
| § 2.1 pending copy switched in place to BrickLink's list, lines retired, none deleted | "2.1" block | Add with the preference, jobs.drain |
§ 3 three stock rows, merge on repeat, 2339 no-match status, prices with provider_key, step error unidentified |
"3" block | Add (loose), pipeline matching, market_price |
| § 4 two loose figures, the waiter with no match and a candidate | "4" block | Add (loose figure), matching, identify.store |
| end: foreign_key_check clean, integrity ok | last block | PRAGMA |
Where the app does not do what the journey text says, the test asserts what the app does:
core.identify.storewrites theidentificationrow only; the journey'sitem_linkto Brickognize with statusto_checkis not written by any writer, so it is not asserted.- Without a Brickset key the
metadatastep endsfailedwithsource_not_configured(the journey says the step errors without a key and does not say which state). - The BrickLink list in § 2.1 leaves the Rebrickable arch
2339unmatched, so BrickLink's own2339becomes a local item and the Rebrickable line is retired (two retired lines in the test, one in the journey, whose BrickLink list differed by one line only).
4. The review's demonstrated failures (design § 12 and § 13)
| Review case | Guarded by |
|---|---|
| FK failure on re-materialisation (L5) | tests/test_store4_schema.py::test_collection_constraints (L5)tests/test_store4_switch.py::test_references_survive_a_switch_away_and_back_and_the_rows_come_back_under_the_same_idtests/test_store4_switch.py::test_a_switch_updates_inserts_and_retires_by_key_and_deletes_nothing |
| Merged id in a live inventory aborts the next Add (L6) | tests/test_store4_schema.py::test_collection_constraints (L6)tests/test_store4_schema.py::test_every_item_reference_has_both_guardstests/test_store4_merge.py |
| NULLs in the stock merge key (L7) | tests/test_store4_schema.py::test_collection_constraints (L7)tests/test_bw_add.py::test_loose_parts_merge_and_undo_restores_the_quantitytests/test_store4_regression.py::test_the_user_journey_replays_through_the_apps_writers |
| Unmatched with an item (L8, C7) | tests/test_store4_schema.py::test_collection_constraints (L8)tests/test_store4_schema.py::test_catalogue_constraints (C7)tests/test_store4_regression.py::test_the_user_journey_replays_through_the_apps_writers |
| Kind change (C3, L9) | tests/test_store4_schema.py::test_catalogue_constraints (C3)tests/test_store4_schema.py::test_collection_constraints (L9)tests/test_store4_builder.py::test_bricklink_corrected_kind_is_kepttests/test_store4_reconcile.py::test_a_referenced_kind_is_kept_and_recorded_as_kind_differs_then_cleared |
| Pending BrickLink state (L2) | tests/test_store4_schema.py::test_collection_constraints (L2)tests/test_store4_pending.pytests/test_bw_add.py::test_bricklink_pending_settles_end_to_end_through_addtests/test_store4_regression.py::test_the_user_journey_replays_through_the_apps_writers |
| Mint collisions, findings 1 (B1 to B6) | section 2 |
| Review 2, fragile line references (L2, L5, C6) | tests/test_store4_schema.py::test_collection_constraints (L5)tests/test_store4_schema.py::test_catalogue_constraints (C6)tests/test_store4_switch.py |
| Review 8, constraints not enforced (L3, L6, L7, L11) | tests/test_store4_schema.py::test_collection_constraintstests/test_store4_schema.py::test_every_item_reference_has_both_guardstests/test_store4_schema.py::test_every_collection_foreign_key_has_an_index_but_three_recorded |
The build_dump.py run (a real 22 October dump: 110,746 items, 0 duplicate line keys, 0 FK violations) needs the real dump and is not in the
suite. scripts/store4_build_dump.py builds it by hand; the builder's own tests (tests/test_store4_builder.py) cover the same invariants on
invented dumps. Not testable here either: anything that needs prod data (§ 8.11), timing on the Pi, the BrickStore file format.