recon + sqlglot validator + drill_down package; guard ReAct dimension picks against candidate list

tests/test_recon.py

@@ -0,0 +1,106 @@
+"""Regression tests for the recon package — type round-trips, join_path,
+and column_to_tables index."""
+
+from api.recon.types import Recon, Relationship, Table, Column, Metric
+
+
+def _sample_recon() -> Recon:
+    cols_account = [
+        Column("account_id", "BIGINT", False),
+        Column("district_id", "BIGINT", True),
+    ]
+    cols_district = [
+        Column("district_id", "BIGINT", False),
+        Column("A2", "TEXT", True),
+        Column("A11", "BIGINT", True),
+    ]
+    cols_loan = [
+        Column("loan_id", "BIGINT", False),
+        Column("account_id", "BIGINT", True),
+        Column("amount", "BIGINT", True),
+        Column("status", "TEXT", True),
+    ]
+    rels = [
+        Relationship("account", "district_id", "district", "district_id"),
+        Relationship("loan", "account_id", "account", "account_id"),
+    ]
+    c2t = {
+        "account_id": sorted(["account", "loan"]),
+        "district_id": sorted(["account", "district"]),
+        "A2": ["district"],
+        "A11": ["district"],
+        "loan_id": ["loan"],
+        "amount": ["loan"],
+        "status": ["loan"],
+    }
+    return Recon(
+        schema="financial",
+        tables={
+            "account": Table("account", "Accounts", cols_account),
+            "district": Table("district", "Districts", cols_district),
+            "loan": Table("loan", "Loans", cols_loan),
+        },
+        metrics={
+            "loan_volume": Metric("loan_volume", "Total CZK lent",
+                                  "SUM(amount)", "loan", None, "CZK"),
+        },
+        column_to_tables=c2t,
+        relationships=rels,
+    )
+
+
+def test_owning_tables():
+    r = _sample_recon()
+    assert r.owning_tables("A2") == ["district"]
+    assert r.owning_tables("amount") == ["loan"]
+    assert r.owning_tables("account_id") == ["account", "loan"]
+    assert r.owning_tables("nonsense") == []
+
+
+def test_join_path_direct():
+    r = _sample_recon()
+    assert r.join_path("loan", "account") == ["loan", "account"]
+    assert r.join_path("account", "district") == ["account", "district"]
+
+
+def test_join_path_multi_hop():
+    r = _sample_recon()
+    assert r.join_path("loan", "district") == ["loan", "account", "district"]
+
+
+def test_join_path_same():
+    r = _sample_recon()
+    assert r.join_path("loan", "loan") == ["loan"]
+
+
+def test_join_path_unknown_returns_none():
+    r = _sample_recon()
+    assert r.join_path("loan", "does_not_exist") is None
+
+
+def test_recon_to_dict_from_dict_roundtrip():
+    r = _sample_recon()
+    d = r.to_dict()
+    r2 = Recon.from_dict(d)
+    assert r2.schema == r.schema
+    assert sorted(r2.tables) == sorted(r.tables)
+    assert sorted(r2.metrics) == sorted(r.metrics)
+    assert r2.column_to_tables == r.column_to_tables
+    assert len(r2.relationships) == len(r.relationships)
+    assert r2.join_path("loan", "district") == ["loan", "account", "district"]
+
+
+def test_relationship_parse_dotted():
+    r = Relationship.parse({"from": "loan.account_id", "to": "account.account_id"})
+    assert r.from_table == "loan"
+    assert r.from_column == "account_id"
+    assert r.to_table == "account"
+    assert r.to_column == "account_id"
+
+
+def test_relationship_parse_explicit():
+    r = Relationship.parse({
+        "from_table": "loan", "from_column": "account_id",
+        "to_table": "account", "to_column": "account_id",
+    })
+    assert r.from_table == "loan"