feat: DM-4438 expose the document path on a schema discovery result

HISTORY.rst

@@ -5,6 +5,9 @@ History
 1.6.0 (unreleased)
 ------------------
 
+* Added an optional ``path`` field to ``SchemaDiscoveryColumn``: where a document-database field sits within its document,
+  as a list of keys with ``*`` for an array position, such as ``["contact", "*", "email"]``.
+  It is ``None`` for a relational column.
 * Added a read-only ``project`` field to ``ConnectionConfig``, ``Ruleset``, ``RulesetLibrary``, ``DiscoveryConfig``,
   ``DiscoveryConfigLibrary``, ``RGConfig`` and ``TableReference``:
   the ``<id>-<slug>`` reference of the project the resource belongs to, such as ``1-default``.

datamasque/client/models/discovery.py

@@ -8,6 +8,7 @@
 from datamasque.client.models.connection import ConnectionConfig, ConnectionId, unwrap_connection_id
 from datamasque.client.models.data_selection import (
     HashColumnsTableConfig,
+    JsonPath,
     Locator,
     UserSelection,
     deserialize_locator,
@@ -375,7 +376,14 @@ class LengthUnit(Enum):
 
 
 class SchemaDiscoveryColumn(BaseModel):
-    """Column-level data in a schema discovery result."""
+    """
+    Column-level data in a schema discovery result.
+
+    `path` is where a document-database field sits within its document, as a list of keys,
+    with `*` for an array position, such as `["contact", "*", "email"]`.
+    The result's `column` holds the same path joined with dots.
+    It is absent for a relational column, and for a result stored by a DataMasque version that did not record paths.
+    """
 
     model_config = ConfigDict(extra="allow")
 
@@ -394,6 +402,7 @@ class SchemaDiscoveryColumn(BaseModel):
     referencing_foreign_keys: list[ReferencingForeignKey]
     constraint: str  # Primary or Unique, or empty string if column does not participate in a PK/UK
     safe_data_preview: Optional[SafeDataPreview] = None
+    path: Optional[JsonPath] = None
 
 
 class SchemaDiscoveryResult(BaseModel):

tests/test_discovery.py

@@ -812,6 +812,42 @@ def test_schema_discovery_result_without_max_length_unit_is_none() -> None:
     assert result.data.max_length_unit is None
 
 
+def test_schema_discovery_result_parses_document_path():
+    row = _schema_discovery_row(1, "contact.*.email")
+    row["data"]["path"] = ["contact", "*", "email"]
+
+    result = SchemaDiscoveryResult.model_validate(row)
+
+    assert result.column == "contact.*.email"
+    assert result.data.path == ["contact", "*", "email"]
+
+
+def test_schema_discovery_result_parses_an_integer_path_segment():
+    """The server's path type allows an array index, so one must not fail to parse."""
+    row = _schema_discovery_row(1, "contact.0.email")
+    row["data"]["path"] = ["contact", 0, "email"]
+
+    result = SchemaDiscoveryResult.model_validate(row)
+
+    assert result.data.path == ["contact", 0, "email"]
+
+
+def test_schema_discovery_result_refuses_a_path_that_is_not_a_list():
+    """`path` is typed, so a value of the wrong shape is an error rather than passed through."""
+    row = _schema_discovery_row(1, "contact.email")
+    row["data"]["path"] = "contact.email"
+
+    with pytest.raises(ValidationError, match="path"):
+        SchemaDiscoveryResult.model_validate(row)
+
+
+def test_schema_discovery_result_without_path_is_none():
+    """A relational column, or a result stored before DataMasque recorded paths, carries no path."""
+    result = SchemaDiscoveryResult.model_validate(_schema_discovery_row(1, "email"))
+
+    assert result.data.path is None
+
+
 def test_list_schema_discovery_results_follows_pagination(client):
     run_id = RunId(42)
     page1 = {