fix(dv2): rename record_source example x5__* -> mp__* (B2)

docs/domain.md

@@ -197,13 +197,13 @@ renames. The two exceptions are listed in §5.4.
 | `wb__` | Wildberries seller API | FBS orders, commissions, returns |
 | `excel__` | Logistics spreadsheets | Container manifests, cross-dock — the inbound-container storyline |
 | `pg_ops__` | Postgres OLTP (hot tier) via CDC | Operational order/customer rows promoted into the vault |
-| `mp__` (currently `x5__`) | Consolidated marketplace order feed | High-volume retail order history (see §5.4) |
+| `mp__` | Consolidated marketplace order feed | High-volume retail order history |
 
 ### 5.4 Planned renames / repins (the only code changes the legend requires)
 
 | Change | Scope | Status |
 | ------ | ----- | ------ |
-| `x5__*` → `mp__*` record_source (+ governance SQL, officer probes, admission tests) | The prefix currently carries the name of the Kaggle seed dataset (X5 Retail Hero) that the demo loader replays as transaction history. Under the legend it is the **consolidated marketplace feed**, and the prefix should say so. Dataset attribution stays in the loader README | Planned (data phase) |
+| `x5__*` → `mp__*` record_source (+ governance SQL, officer probes, admission tests) | The prefix carried the name of the Kaggle seed dataset (X5 Retail Hero) that the demo loader replays as transaction history. Under the legend it is the **consolidated marketplace feed**, and the prefix says so. Dataset attribution stays in the loader README | **Done** (B2) |
 | Demo value repin: currencies to `RUB` (primary), `AED`/`KZT` in branch stories; demo revenue/counts consistent with §1–2 | `contracts/entities/order.yaml` currency examples, NL demo answers, seeded `ORD-*` rows | Planned (data phase, after the generator spec) |
 
 Vocabulary guardrails for all public docs: the company is an **own-brand /

tests/unit/test_dv2_business_vault_ddl.py

@@ -4,8 +4,8 @@
 this file. It pins that every view parses under sqlglot's ClickHouse dialect and
 that the customer MDM views admit hub rows by branch via
 ``splitByString('__', record_source)[2]`` — NOT by a hard-coded
-``record_source = '1c__<branch>'`` filter that silently dropped OLTP/X5-promoted
-customers (``record_source`` ``pg_ops__`` / ``x5__``).
+``record_source = '1c__<branch>'`` filter that silently dropped OLTP/marketplace-
+promoted customers (``record_source`` ``pg_ops__`` / ``mp__``).
 
 This is the ClickHouse half of audit_28_06_26 #12. The PostgreSQL port was fixed
 in ``test_dv2_postgres_ddl.py::test_customer_mdm_views_admit_all_source_conventions``;
@@ -59,15 +59,17 @@ def test_customer_mdm_views_admit_all_source_conventions():
     """audit_28_06_26 #12 (ClickHouse half): the customer MDM views must select
     hub rows by branch via ``splitByString('__', record_source)[2]``, NOT by a
     hard-coded ``record_source = '1c__<branch>'`` filter that silently drops
-    OLTP/X5-promoted customers (record_source ``pg_ops__`` / ``x5__``). Proven
-    live on PostgreSQL in #99: the buggy filter returns 1 of 2 seeded customers,
-    the source-agnostic filter returns both. This keeps the ClickHouse views in
-    lock-step with the PostgreSQL port so the engines cannot diverge again."""
+    OLTP/marketplace-promoted customers (record_source ``pg_ops__`` / ``mp__``).
+    Proven live on PostgreSQL in #99: the buggy filter returns 1 of 2 seeded
+    customers, the source-agnostic filter returns both. This keeps the
+    ClickHouse views in lock-step with the PostgreSQL port so the engines
+    cannot diverge again."""
     for branch in MDM_BRANCHES:
         path = BV_DIR / f"bv_customer_mdm__{branch}.sql"
         # Strip block/line comments: the headers deliberately quote the old,
         # buggy ``record_source = '1c__<branch>'`` filter to explain the fix.
-        body = _strip_comments(path.read_text(encoding="utf-8"))
+        raw = path.read_text(encoding="utf-8")
+        body = _strip_comments(raw)
         assert f"CREATE OR REPLACE VIEW rv.bv_customer_mdm__{branch}" in body, (
             f"bv_customer_mdm__{branch}.sql must define rv.bv_customer_mdm__{branch}"
         )
@@ -80,3 +82,13 @@ def test_customer_mdm_views_admit_all_source_conventions():
             f"hard-coded record_source = '1c__<branch>' filter in "
             f"bv_customer_mdm__{branch} reintroduces audit #12"
         )
+        # B2 (domain.md §5.4): the legend's marketplace-feed vocabulary replaces
+        # the Kaggle dataset name in the third-source-convention example (checked
+        # on the RAW text — the example lives in the header comment, which body
+        # strips).
+        assert "x5__" not in raw, (
+            f"stale Kaggle-dataset record_source prefix x5__ leaked back into bv_customer_mdm__{branch}.sql"
+        )
+        assert "mp__" in raw, (
+            f"bv_customer_mdm__{branch}.sql should document the mp__ marketplace-feed convention (domain.md §5.3)"
+        )

tests/unit/test_dv2_postgres_ddl.py

@@ -147,10 +147,12 @@ def test_business_vault_uses_postgres_collapse():
 def test_customer_mdm_views_admit_all_source_conventions():
     """audit_28_06_26 #12: the customer MDM views must select hub rows by
     branch via split_part(record_source, '__', 2), NOT by a hard-coded
-    record_source = '1c__<branch>' filter that silently drops OLTP/X5-promoted
-    customers (record_source pg_ops__/x5__). Proven live on PG: the buggy filter
-    returns 1 of 2 seeded customers, the split_part filter returns both."""
-    body = _strip_comments((PG_DIR / "03_business_vault.sql").read_text(encoding="utf-8")).lower()
+    record_source = '1c__<branch>' filter that silently drops OLTP/marketplace-
+    promoted customers (record_source pg_ops__/mp__). Proven live on PG: the
+    buggy filter returns 1 of 2 seeded customers, the split_part filter returns
+    both."""
+    raw = (PG_DIR / "03_business_vault.sql").read_text(encoding="utf-8").lower()
+    body = _strip_comments(raw)
     branches = ("msk", "spb", "ekb", "dxb", "ala")
     for branch in branches:
         assert f"view rv.bv_customer_mdm__{branch}" in body, f"missing PG view for {branch}"
@@ -161,6 +163,15 @@ def test_customer_mdm_views_admit_all_source_conventions():
     assert "record_source = '1c__" not in body, (
         "hard-coded record_source = '1c__<branch>' filter reintroduces audit #12"
     )
+    # B2 (domain.md §5.4): the legend's marketplace-feed vocabulary replaces the
+    # Kaggle dataset name in the third-source-convention example (checked on the
+    # RAW text — the example lives in the header comment, which body strips).
+    assert "x5__" not in raw, (
+        "stale Kaggle-dataset record_source prefix x5__ leaked back into 03_business_vault.sql"
+    )
+    assert "mp__" in raw, (
+        "03_business_vault.sql should document the mp__ marketplace-feed convention (domain.md §5.3)"
+    )
 
 
 def test_hubs_and_links_have_bytea_primary_keys():

warehouse/agentflow/dv2/business_vault/bv_customer_mdm__ala.sql

@@ -3,7 +3,7 @@ Purpose: Canonical customer record for the ALA branch.
 Layer:   Business Vault.
 Branch:  ala (KZ jurisdiction; Bitrix loyalty not wired in ALA).
 Hub admission: splitByString('__', record_source)[2] = 'ala' (source-agnostic:
-         1c__/pg_ops__/x5__ all integrated, not only 1C; audit_28_06_26 #12;
+         1c__/pg_ops__/mp__ all integrated, not only 1C; audit_28_06_26 #12;
          mirrors the PostgreSQL port's split_part(record_source,'__',2)).
 Security: SQL SECURITY DEFINER (ADR 0006 Phase 2) — readers query this view
          under the definer's rights, so the column-limited grants in

warehouse/agentflow/dv2/business_vault/bv_customer_mdm__dxb.sql

@@ -8,7 +8,7 @@ Conflict policy:
     keeps the loyalty columns for schema parity with bv_customer_mdm__msk so
     downstream marts can UNION ALL the two branches without renaming.
 Hub admission: splitByString('__', record_source)[2] = 'dxb' (source-agnostic:
-         1c__/pg_ops__/x5__ all integrated, not only 1C; audit_28_06_26 #12;
+         1c__/pg_ops__/mp__ all integrated, not only 1C; audit_28_06_26 #12;
          mirrors the PostgreSQL port's split_part(record_source,'__',2)).
 Security: SQL SECURITY DEFINER (ADR 0006 Phase 2) — readers query this view
          under the definer's rights, so the column-limited grants in

warehouse/agentflow/dv2/business_vault/bv_customer_mdm__ekb.sql

@@ -3,7 +3,7 @@ Purpose: Canonical customer record for the EKB branch.
 Layer:   Business Vault.
 Branch:  ekb (RU jurisdiction; same conflict policy as MSK).
 Hub admission: splitByString('__', record_source)[2] = 'ekb' (source-agnostic:
-         1c__/pg_ops__/x5__ all integrated, not only 1C; audit_28_06_26 #12;
+         1c__/pg_ops__/mp__ all integrated, not only 1C; audit_28_06_26 #12;
          mirrors the PostgreSQL port's split_part(record_source,'__',2)).
 Security: SQL SECURITY DEFINER (ADR 0006 Phase 2) — readers query this view
          under the definer's rights, so the column-limited grants in

warehouse/agentflow/dv2/business_vault/bv_customer_mdm__msk.sql

@@ -9,9 +9,9 @@ Conflict policy:
   - If a customer exists only in Bitrix, PII columns are NULL but the row
     is still returned so loyalty-only customers stay visible.
 Hub admission: splitByString('__', record_source)[2] = 'msk', so a customer
-         promoted under ANY source convention (1c__msk, pg_ops__msk, x5__msk,
+         promoted under ANY source convention (1c__msk, pg_ops__msk, mp__msk,
          ...) is integrated, not only 1C. The old record_source = '1c__msk'
-         filter silently dropped OLTP/X5-promoted customers (audit_28_06_26 #12);
+         filter silently dropped OLTP/marketplace-promoted customers (audit_28_06_26 #12);
          this mirrors the PostgreSQL port's split_part(record_source,'__',2).
 Security: SQL SECURITY DEFINER (ADR 0006 Phase 2) — readers query this view
          under the definer's rights, so the column-limited grants in

warehouse/agentflow/dv2/business_vault/bv_customer_mdm__spb.sql

@@ -3,7 +3,7 @@ Purpose: Canonical customer record for the SPB branch.
 Layer:   Business Vault.
 Branch:  spb (RU jurisdiction; same conflict policy as MSK).
 Hub admission: splitByString('__', record_source)[2] = 'spb' (source-agnostic:
-         1c__/pg_ops__/x5__ all integrated, not only 1C; audit_28_06_26 #12;
+         1c__/pg_ops__/mp__ all integrated, not only 1C; audit_28_06_26 #12;
          mirrors the PostgreSQL port's split_part(record_source,'__',2)).
 Security: SQL SECURITY DEFINER (ADR 0006 Phase 2) — readers query this view
          under the definer's rights, so the column-limited grants in

warehouse/agentflow/dv2/postgres/03_business_vault.sql

@@ -144,10 +144,10 @@ Dialect: postgresql.
 Branch:  one view per jurisdiction by design — PII stays in branch.
 Hub admission: split_part(record_source,'__',2) = '<branch>', so a customer
          promoted under ANY source convention (1c__<branch>, pg_ops__<branch>,
-         x5__<branch>, ...) is integrated. This mirrors bv_order_canonical's
+         mp__<branch>, ...) is integrated. This mirrors bv_order_canonical's
          order_branch derivation above. The ClickHouse views hard-code
          record_source = '1c__<branch>', which silently dropped every OLTP- and
-         X5-promoted customer (record_source pg_ops__/x5__) from the MDM result
+         marketplace-promoted customer (record_source pg_ops__/mp__) from the MDM result
          (audit_28_06_26 #12). The hash keys were never incompatible —
          customer_hk = md5(business_key) is identical across loaders; only the
          hub record_source filter excluded them. split_part is the source-

warehouse/agentflow/dv2/postgres/governance/verify_live.sh

@@ -49,7 +49,7 @@ INSERT INTO rv.hub_customer (customer_hk, customer_bk, record_source) VALUES
  (decode(md5('CUST-MSK-5'),'hex'),'CUST-MSK-5','1c__msk'),
  (decode(md5('CUST-MSK-6'),'hex'),'CUST-MSK-6','1c__msk'),
  (decode(md5('CUST-MSK-7'),'hex'),'CUST-MSK-7','pg_ops__msk'),
- (decode(md5('CUST-MSK-8'),'hex'),'CUST-MSK-8','x5__msk'),
+ (decode(md5('CUST-MSK-8'),'hex'),'CUST-MSK-8','mp__msk'),
  (decode(md5('CUST-DXB-1'),'hex'),'CUST-DXB-1','1c__dxb'),
  (decode(md5('CUST-DXB-2'),'hex'),'CUST-DXB-2','1c__dxb')
 ON CONFLICT (customer_hk) DO NOTHING;