acryl-datahub 1.1.0.5rc7__py3-none-any.whl → 1.1.0.5rc9__py3-none-any.whl

datahub/ingestion/source/sql/vertica.py

@@ -45,7 +45,6 @@ from datahub.metadata.com.linkedin.pegasus2avro.dataset import UpstreamLineage
 from datahub.metadata.com.linkedin.pegasus2avro.metadata.snapshot import DatasetSnapshot
 from datahub.metadata.com.linkedin.pegasus2avro.mxe import MetadataChangeEvent
 from datahub.metadata.schema_classes import (
-    ChangeTypeClass,
     DatasetLineageTypeClass,
     DatasetPropertiesClass,
     SubTypesClass,
@@ -501,10 +500,7 @@ class VerticaSource(SQLAlchemySource):
         if dpi_aspect:
             yield dpi_aspect
         yield MetadataChangeProposalWrapper(
-            entityType="dataset",
-            changeType=ChangeTypeClass.UPSERT,
             entityUrn=dataset_urn,
-            aspectName="subTypes",
             aspect=SubTypesClass(typeNames=[DatasetSubTypes.PROJECTIONS]),
         ).as_workunit()
 

datahub/ingestion/source/sql_queries.py

@@ -66,7 +66,7 @@ class SqlQueriesSourceConfig(PlatformInstanceConfigMixin, EnvConfigMixin):
         description="The default schema to use for unqualified table names",
         default=None,
     )
-    default_dialect: Optional[str] = Field(
+    override_dialect: Optional[str] = Field(
         description="The SQL dialect to use when parsing queries. Overrides automatic dialect detection.",
         default=None,
     )
@@ -181,7 +181,7 @@ class SqlQueriesSource(Source):
             schema_resolver=self.schema_resolver,
             default_db=self.config.default_db,
             default_schema=self.config.default_schema,
-            default_dialect=self.config.default_dialect,
+            override_dialect=self.config.override_dialect,
         )
         if result.debug_info.table_error:
             logger.info(f"Error parsing table lineage, {result.debug_info.table_error}")

datahub/ingestion/source/superset.py

@@ -658,6 +658,7 @@ class SupersetSource(StatefulIngestionSourceBase):
         if datasource_id:
             dataset_info = self.get_dataset_info(datasource_id).get("result", {})
             dataset_column_info = dataset_info.get("columns", [])
+            dataset_metric_info = dataset_info.get("metrics", [])
 
             for column in dataset_column_info:
                 col_name = column.get("column_name", "")
@@ -671,6 +672,17 @@ class SupersetSource(StatefulIngestionSourceBase):
                     continue
 
                 dataset_columns.append((col_name, col_type, col_description))
+
+            for metric in dataset_metric_info:
+                metric_name = metric.get("metric_name", "")
+                metric_type = metric.get("metric_type", "")
+                metric_description = metric.get("description", "")
+
+                if metric_name == "" or metric_type == "":
+                    logger.info(f"could not construct metric lineage for {metric}")
+                    continue
+
+                dataset_columns.append((metric_name, metric_type, metric_description))
         else:
             # if no datasource id, cannot build cll, just return
             logger.warning(
@@ -972,19 +984,44 @@ class SupersetSource(StatefulIngestionSourceBase):
             schema_fields.append(field)
         return schema_fields
 
+    def gen_metric_schema_fields(
+        self, metric_data: List[Dict[str, Any]]
+    ) -> List[SchemaField]:
+        schema_fields: List[SchemaField] = []
+        for metric in metric_data:
+            metric_type = metric.get("metric_type", "")
+            data_type = resolve_sql_type(metric_type)
+            if data_type is None:
+                data_type = NullType()
+
+            field = SchemaField(
+                fieldPath=metric.get("metric_name", ""),
+                type=SchemaFieldDataType(data_type),
+                nativeDataType=metric_type or "",
+                description=metric.get("description", ""),
+                nullable=True,
+            )
+            schema_fields.append(field)
+        return schema_fields
+
     def gen_schema_metadata(
         self,
         dataset_response: dict,
     ) -> SchemaMetadata:
         dataset_response = dataset_response.get("result", {})
         column_data = dataset_response.get("columns", [])
+        metric_data = dataset_response.get("metrics", [])
+
+        column_fields = self.gen_schema_fields(column_data)
+        metric_fields = self.gen_metric_schema_fields(metric_data)
+
         schema_metadata = SchemaMetadata(
             schemaName=dataset_response.get("table_name", ""),
             platform=make_data_platform_urn(self.platform),
             version=0,
             hash="",
             platformSchema=MySqlDDL(tableSchema=""),
-            fields=self.gen_schema_fields(column_data),
+            fields=column_fields + metric_fields,
         )
         return schema_metadata
 
@@ -1049,6 +1086,8 @@ class SupersetSource(StatefulIngestionSourceBase):
         # To generate column level lineage, we can manually decode the metadata
         # to produce the ColumnLineageInfo
         columns = dataset_response.get("result", {}).get("columns", [])
+        metrics = dataset_response.get("result", {}).get("metrics", [])
+
         fine_grained_lineages: List[FineGrainedLineageClass] = []
 
         for column in columns:
@@ -1067,6 +1106,22 @@ class SupersetSource(StatefulIngestionSourceBase):
                 )
             )
 
+        for metric in metrics:
+            metric_name = metric.get("metric_name", "")
+            if not metric_name:
+                continue
+
+            downstream = [make_schema_field_urn(datasource_urn, metric_name)]
+            upstreams = [make_schema_field_urn(upstream_dataset, metric_name)]
+            fine_grained_lineages.append(
+                FineGrainedLineageClass(
+                    downstreamType=FineGrainedLineageDownstreamTypeClass.FIELD,
+                    downstreams=downstream,
+                    upstreamType=FineGrainedLineageUpstreamTypeClass.FIELD_SET,
+                    upstreams=upstreams,
+                )
+            )
+
         upstream_lineage = UpstreamLineageClass(
             upstreams=[
                 UpstreamClass(

datahub/ingestion/source/tableau/tableau.py

@@ -149,7 +149,6 @@ from datahub.metadata.com.linkedin.pegasus2avro.schema import (
 )
 from datahub.metadata.schema_classes import (
     BrowsePathsClass,
-    ChangeTypeClass,
     ChartInfoClass,
     ChartUsageStatisticsClass,
     DashboardInfoClass,
@@ -529,6 +528,14 @@ class TableauConfig(
         default=False,
         description="Ingest details for tables external to (not embedded in) tableau as entities.",
     )
+    emit_all_published_datasources: bool = Field(
+        default=False,
+        description="Ingest all published data sources. When False (default), only ingest published data sources that belong to an ingested workbook.",
+    )
+    emit_all_embedded_datasources: bool = Field(
+        default=False,
+        description="Ingest all embedded data sources. When False (default), only ingest embedded data sources that belong to an ingested workbook.",
+    )
 
     env: str = Field(
         default=builder.DEFAULT_ENV,
@@ -2180,32 +2187,32 @@ class TableauSiteSource:
                     else []
                 )
 
-                # The Tableau SQL parser much worse than our sqlglot based parser,
-                # so relying on metadata parsed by Tableau from SQL queries can be
-                # less accurate. This option allows us to ignore Tableau's parser and
-                # only use our own.
-                if self.config.force_extraction_of_lineage_from_custom_sql_queries:
-                    logger.debug("Extracting TLL & CLL from custom sql (forced)")
+                tableau_table_list = csql.get(c.TABLES, [])
+                if self.config.force_extraction_of_lineage_from_custom_sql_queries or (
+                    not tableau_table_list
+                    and self.config.extract_lineage_from_unsupported_custom_sql_queries
+                ):
+                    if not tableau_table_list:
+                        # custom sql tables may contain unsupported sql, causing incomplete lineage
+                        # we extract the lineage from the raw queries
+                        logger.debug(
+                            "Parsing TLL & CLL from custom sql (tableau metadata incomplete)"
+                        )
+                    else:
+                        # The Tableau SQL parser is much worse than our sqlglot based parser,
+                        # so relying on metadata parsed by Tableau from SQL queries can be
+                        # less accurate. This option allows us to ignore Tableau's parser and
+                        # only use our own.
+                        logger.debug("Parsing TLL & CLL from custom sql (forced)")
+
                     yield from self._create_lineage_from_unsupported_csql(
                         csql_urn, csql, columns
                     )
                 else:
-                    tables = csql.get(c.TABLES, [])
-
-                    if tables:
-                        # lineage from custom sql -> datasets/tables #
-                        yield from self._create_lineage_to_upstream_tables(
-                            csql_urn, tables, datasource
-                        )
-                    elif (
-                        self.config.extract_lineage_from_unsupported_custom_sql_queries
-                    ):
-                        logger.debug("Extracting TLL & CLL from custom sql")
-                        # custom sql tables may contain unsupported sql, causing incomplete lineage
-                        # we extract the lineage from the raw queries
-                        yield from self._create_lineage_from_unsupported_csql(
-                            csql_urn, csql, columns
-                        )
+                    # lineage from custom sql -> datasets/tables #
+                    yield from self._create_lineage_to_upstream_tables(
+                        csql_urn, tableau_table_list, datasource
+                    )
 
             #  Schema Metadata
             schema_metadata = self.get_schema_metadata_for_custom_sql(columns)
@@ -2243,7 +2250,6 @@ class TableauSiteSource:
             yield self.get_metadata_change_event(dataset_snapshot)
             yield self.get_metadata_change_proposal(
                 dataset_snapshot.urn,
-                aspect_name=c.SUB_TYPES,
                 aspect=SubTypesClass(typeNames=[DatasetSubTypes.VIEW, c.CUSTOM_SQL]),
             )
 
@@ -2408,7 +2414,6 @@ class TableauSiteSource:
             upstream_lineage = UpstreamLineage(upstreams=upstream_tables)
             yield self.get_metadata_change_proposal(
                 csql_urn,
-                aspect_name=c.UPSTREAM_LINEAGE,
                 aspect=upstream_lineage,
             )
             self.report.num_tables_with_upstream_lineage += 1
@@ -2594,7 +2599,6 @@ class TableauSiteSource:
         )
         yield self.get_metadata_change_proposal(
             csql_urn,
-            aspect_name=c.UPSTREAM_LINEAGE,
             aspect=upstream_lineage,
         )
         self.report.num_tables_with_upstream_lineage += 1
@@ -2640,14 +2644,10 @@ class TableauSiteSource:
     def get_metadata_change_proposal(
         self,
         urn: str,
-        aspect_name: str,
         aspect: Union["UpstreamLineage", "SubTypesClass"],
     ) -> MetadataWorkUnit:
         return MetadataChangeProposalWrapper(
-            entityType=c.DATASET,
-            changeType=ChangeTypeClass.UPSERT,
             entityUrn=urn,
-            aspectName=aspect_name,
             aspect=aspect,
         ).as_workunit()
 
@@ -2755,7 +2755,6 @@ class TableauSiteSource:
                 )
                 yield self.get_metadata_change_proposal(
                     datasource_urn,
-                    aspect_name=c.UPSTREAM_LINEAGE,
                     aspect=upstream_lineage,
                 )
                 self.report.num_tables_with_upstream_lineage += 1
@@ -2774,7 +2773,6 @@ class TableauSiteSource:
         yield self.get_metadata_change_event(dataset_snapshot)
         yield self.get_metadata_change_proposal(
             dataset_snapshot.urn,
-            aspect_name=c.SUB_TYPES,
             aspect=SubTypesClass(
                 typeNames=(
                     ["Embedded Data Source"]
@@ -2860,7 +2858,11 @@ class TableauSiteSource:
         return datasource
 
     def emit_published_datasources(self) -> Iterable[MetadataWorkUnit]:
-        datasource_filter = {c.ID_WITH_IN: self.datasource_ids_being_used}
+        datasource_filter = (
+            {}
+            if self.config.emit_all_published_datasources
+            else {c.ID_WITH_IN: self.datasource_ids_being_used}
+        )
 
         for datasource in self.get_connection_objects(
             query=published_datasource_graphql_query,
@@ -3553,7 +3555,11 @@ class TableauSiteSource:
         return browse_paths
 
     def emit_embedded_datasources(self) -> Iterable[MetadataWorkUnit]:
-        datasource_filter = {c.ID_WITH_IN: self.embedded_datasource_ids_being_used}
+        datasource_filter = (
+            {}
+            if self.config.emit_all_embedded_datasources
+            else {c.ID_WITH_IN: self.embedded_datasource_ids_being_used}
+        )
 
         for datasource in self.get_connection_objects(
             query=embedded_datasource_graphql_query,

datahub/ingestion/source/tableau/tableau_constant.py

@@ -50,7 +50,6 @@ TABLES = "tables"
 DESCRIPTION = "description"
 SQL = "SQL"
 QUERY = "query"
-SUB_TYPES = "subTypes"
 VIEW = "view"
 CUSTOM_SQL = "Custom SQL"
 REMOTE_TYPE = "remoteType"
@@ -58,7 +57,6 @@ UNKNOWN = "UNKNOWN"
 PUBLISHED_DATA_SOURCE = "PublishedDatasource"
 LUID = "luid"
 EMBEDDED_DATA_SOURCE = "EmbeddedDatasource"
-UPSTREAM_LINEAGE = "upstreamLineage"
 OWNER = "owner"
 USERNAME = "username"
 HAS_EXTRACTS = "hasExtracts"

datahub/ingestion/source/unity/source.py

@@ -56,6 +56,7 @@ from datahub.ingestion.source.aws.s3_util import (
 from datahub.ingestion.source.common.subtypes import (
     DatasetContainerSubTypes,
     DatasetSubTypes,
+    SourceCapabilityModifier,
 )
 from datahub.ingestion.source.state.stale_entity_removal_handler import (
     StaleEntityRemovalHandler,
@@ -152,7 +153,14 @@ logger: logging.Logger = logging.getLogger(__name__)
 @capability(SourceCapability.USAGE_STATS, "Enabled by default")
 @capability(SourceCapability.PLATFORM_INSTANCE, "Enabled by default")
 @capability(SourceCapability.DOMAINS, "Supported via the `domain` config field")
-@capability(SourceCapability.CONTAINERS, "Enabled by default")
+@capability(
+    SourceCapability.CONTAINERS,
+    "Enabled by default",
+    subtype_modifier=[
+        SourceCapabilityModifier.CATALOG,
+        SourceCapabilityModifier.SCHEMA,
+    ],
+)
 @capability(SourceCapability.OWNERSHIP, "Supported via the `include_ownership` config")
 @capability(
     SourceCapability.DATA_PROFILING, "Supported via the `profiling.enabled` config"

datahub/sdk/lineage_client.py

@@ -478,7 +478,7 @@ class LineageClient:
         env: str = "PROD",
         default_db: Optional[str] = None,
         default_schema: Optional[str] = None,
-        default_dialect: Optional[str] = None,
+        override_dialect: Optional[str] = None,
     ) -> None:
         """Add lineage by parsing a SQL query."""
         from datahub.sql_parsing.sqlglot_lineage import (
@@ -494,7 +494,7 @@ class LineageClient:
             platform_instance=platform_instance,
             env=env,
             graph=self._client._graph,
-            default_dialect=default_dialect,
+            override_dialect=override_dialect,
         )
 
         if parsed_result.debug_info.table_error:

datahub/sql_parsing/sql_parsing_aggregator.py

@@ -1577,27 +1577,33 @@ class SqlParsingAggregator(Closeable):
 
         @dataclasses.dataclass
         class QueryLineageInfo:
-            upstreams: List[UrnStr]  # this is direct upstreams, with *no temp tables*
-            column_lineage: List[ColumnLineageInfo]
+            upstreams: OrderedSet[
+                UrnStr
+            ]  # this is direct upstreams, with *no temp tables*
+            column_lineage: OrderedSet[ColumnLineageInfo]
             confidence_score: float
 
             def _merge_lineage_from(self, other_query: "QueryLineageInfo") -> None:
-                self.upstreams += other_query.upstreams
-                self.column_lineage += other_query.column_lineage
+                self.upstreams.update(other_query.upstreams)
+                self.column_lineage.update(other_query.column_lineage)
                 self.confidence_score = min(
                     self.confidence_score, other_query.confidence_score
                 )
 
+        cache: Dict[str, QueryLineageInfo] = {}
+
         def _recurse_into_query(
             query: QueryMetadata, recursion_path: List[QueryId]
         ) -> QueryLineageInfo:
             if query.query_id in recursion_path:
                 # This is a cycle, so we just return the query as-is.
                 return QueryLineageInfo(
-                    upstreams=query.upstreams,
-                    column_lineage=query.column_lineage,
+                    upstreams=OrderedSet(query.upstreams),
+                    column_lineage=OrderedSet(query.column_lineage),
                     confidence_score=query.confidence_score,
                 )
+            if query.query_id in cache:
+                return cache[query.query_id]
             recursion_path = [*recursion_path, query.query_id]
             composed_of_queries.add(query.query_id)
 
@@ -1612,7 +1618,7 @@ class SqlParsingAggregator(Closeable):
                         upstream_query = self._query_map.get(upstream_query_id)
                         if (
                             upstream_query
-                            and upstream_query.query_id not in composed_of_queries
+                            and upstream_query.query_id not in recursion_path
                         ):
                             temp_query_lineage_info = _recurse_into_query(
                                 upstream_query, recursion_path
@@ -1672,11 +1678,14 @@ class SqlParsingAggregator(Closeable):
                 ]
             )
 
-            return QueryLineageInfo(
-                upstreams=list(new_upstreams),
-                column_lineage=new_cll,
+            ret = QueryLineageInfo(
+                upstreams=new_upstreams,
+                column_lineage=OrderedSet(new_cll),
                 confidence_score=new_confidence_score,
             )
+            cache[query.query_id] = ret
+
+            return ret
 
         resolved_lineage_info = _recurse_into_query(base_query, [])
 
@@ -1716,8 +1725,8 @@ class SqlParsingAggregator(Closeable):
             base_query,
             query_id=composite_query_id,
             formatted_query_string=merged_query_text,
-            upstreams=resolved_lineage_info.upstreams,
-            column_lineage=resolved_lineage_info.column_lineage,
+            upstreams=list(resolved_lineage_info.upstreams),
+            column_lineage=list(resolved_lineage_info.column_lineage),
             confidence_score=resolved_lineage_info.confidence_score,
         )
 

datahub/sql_parsing/sqlglot_lineage.py

@@ -56,6 +56,7 @@ from datahub.sql_parsing.sql_parsing_common import (
     QueryTypeProps,
 )
 from datahub.sql_parsing.sqlglot_utils import (
+    DialectOrStr,
     get_dialect,
     get_query_fingerprint_debug,
     is_dialect_instance,
@@ -124,6 +125,17 @@ class _DownstreamColumnRef(_ParserBaseModel):
 
 
 class DownstreamColumnRef(_ParserBaseModel):
+    """
+    TODO: Instead of implementing custom __hash__ function this class should simply inherit from _FrozenModel.
+          What stops us is that `column_type` field of type `SchemaFieldDataTypeClass` is not hashable - it's an
+          auto-generated class from .pdl model files. We need generic solution allowing us to either:
+          1. Implement hashing for .pdl model objects
+          2. Reliably provide pydantic (both v1 and v2) with information to skip particular fields from default
+             hash function - with a twist here that _FrozenModel implements its own `__lt__` function - it needs
+             to understand that instruction as well.
+          Instances of this class needs to be hashable as we store them in a set when processing lineage from queries.
+    """
+
     table: Optional[Urn] = None
     column: str
     column_type: Optional[SchemaFieldDataTypeClass] = None
@@ -139,8 +151,11 @@ class DownstreamColumnRef(_ParserBaseModel):
             return v
         return SchemaFieldDataTypeClass.from_obj(v)
 
+    def __hash__(self) -> int:
+        return hash((self.table, self.column, self.native_column_type))
 
-class ColumnTransformation(_ParserBaseModel):
+
+class ColumnTransformation(_FrozenModel):
     is_direct_copy: bool
     column_logic: str
 
@@ -153,11 +168,21 @@ class _ColumnLineageInfo(_ParserBaseModel):
 
 
 class ColumnLineageInfo(_ParserBaseModel):
+    """
+    TODO: Instead of implementing custom __hash__ function this class should simply inherit from _FrozenModel.
+          To achieve this, we need to change `upstreams` to `Tuple[ColumnRef, ...]` - along with many code lines
+          depending on it.
+          Instances of this class needs to be hashable as we store them in a set when processing lineage from queries.
+    """
+
     downstream: DownstreamColumnRef
     upstreams: List[ColumnRef]
 
     logic: Optional[ColumnTransformation] = pydantic.Field(default=None)
 
+    def __hash__(self) -> int:
+        return hash((self.downstream, tuple(self.upstreams), self.logic))
+
 
 class _JoinInfo(_ParserBaseModel):
     join_type: str
@@ -1231,12 +1256,12 @@ def _sqlglot_lineage_inner(
     schema_resolver: SchemaResolverInterface,
     default_db: Optional[str] = None,
     default_schema: Optional[str] = None,
-    default_dialect: Optional[str] = None,
+    override_dialect: Optional[DialectOrStr] = None,
 ) -> SqlParsingResult:
-    if not default_dialect:
-        dialect = get_dialect(schema_resolver.platform)
+    if override_dialect:
+        dialect = get_dialect(override_dialect)
     else:
-        dialect = get_dialect(default_dialect)
+        dialect = get_dialect(schema_resolver.platform)
 
     default_db = _normalize_db_or_schema(default_db, dialect)
     default_schema = _normalize_db_or_schema(default_schema, dialect)
@@ -1423,7 +1448,7 @@ def _sqlglot_lineage_nocache(
     schema_resolver: SchemaResolverInterface,
     default_db: Optional[str] = None,
     default_schema: Optional[str] = None,
-    default_dialect: Optional[str] = None,
+    override_dialect: Optional[DialectOrStr] = None,
 ) -> SqlParsingResult:
     """Parse a SQL statement and generate lineage information.
 
@@ -1441,8 +1466,8 @@ def _sqlglot_lineage_nocache(
     can be brittle with respect to missing schema information and complex
     SQL logic like UNNESTs.
 
-    The SQL dialect can be given as an argument called default_dialect or it can
-    be inferred from the schema_resolver's platform.
+    The SQL dialect will be inferred from the schema_resolver's platform.
+    That inference can be overridden by passing an override_dialect argument.
     The set of supported dialects is the same as sqlglot's. See their
     `documentation <https://sqlglot.com/sqlglot/dialects/dialect.html#Dialects>`_
     for the full list.
@@ -1457,7 +1482,7 @@ def _sqlglot_lineage_nocache(
         schema_resolver: The schema resolver to use for resolving table schemas.
         default_db: The default database to use for unqualified table names.
         default_schema: The default schema to use for unqualified table names.
-        default_dialect: A default dialect to override the dialect provided by 'schema_resolver'.
+        override_dialect: Override the dialect provided by 'schema_resolver'.
 
     Returns:
         A SqlParsingResult object containing the parsed lineage information.
@@ -1482,7 +1507,7 @@ def _sqlglot_lineage_nocache(
             schema_resolver=schema_resolver,
             default_db=default_db,
             default_schema=default_schema,
-            default_dialect=default_dialect,
+            override_dialect=override_dialect,
         )
     except Exception as e:
         return SqlParsingResult.make_from_error(e)
@@ -1520,15 +1545,15 @@ def sqlglot_lineage(
     schema_resolver: SchemaResolverInterface,
     default_db: Optional[str] = None,
     default_schema: Optional[str] = None,
-    default_dialect: Optional[str] = None,
+    override_dialect: Optional[DialectOrStr] = None,
 ) -> SqlParsingResult:
     if schema_resolver.includes_temp_tables():
         return _sqlglot_lineage_nocache(
-            sql, schema_resolver, default_db, default_schema, default_dialect
+            sql, schema_resolver, default_db, default_schema, override_dialect
         )
     else:
         return _sqlglot_lineage_cached(
-            sql, schema_resolver, default_db, default_schema, default_dialect
+            sql, schema_resolver, default_db, default_schema, override_dialect
         )
 
 
@@ -1580,7 +1605,7 @@ def create_lineage_sql_parsed_result(
     default_schema: Optional[str] = None,
     graph: Optional[DataHubGraph] = None,
     schema_aware: bool = True,
-    default_dialect: Optional[str] = None,
+    override_dialect: Optional[DialectOrStr] = None,
 ) -> SqlParsingResult:
     schema_resolver = create_schema_resolver(
         platform=platform,
@@ -1600,7 +1625,7 @@ def create_lineage_sql_parsed_result(
             schema_resolver=schema_resolver,
             default_db=default_db,
             default_schema=default_schema,
-            default_dialect=default_dialect,
+            override_dialect=override_dialect,
         )
     except Exception as e:
         return SqlParsingResult.make_from_error(e)