diff --git a/metadata-ingestion/docs/dev_guides/classification.md b/metadata-ingestion/docs/dev_guides/classification.md index f20638a2ab5bd..39eac229a6601 100644 --- a/metadata-ingestion/docs/dev_guides/classification.md +++ b/metadata-ingestion/docs/dev_guides/classification.md @@ -10,7 +10,7 @@ Note that a `.` is used to denote nested fields in the YAML recipe. | ------------------------- | -------- | --------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------- | | enabled | | boolean | Whether classification should be used to auto-detect glossary terms | False | | sample_size | | int | Number of sample values used for classification. | 100 | -| max_workers | | int | Number of worker threads to use for classification. Set to 1 to disable. | Number of cpu cores or 4 | +| max_workers | | int | Number of worker processes to use for classification. Set to 1 to disable. | Number of cpu cores or 4 | | info_type_to_term | | Dict[str,string] | Optional mapping to provide glossary term identifier for info type. | By default, info type is used as glossary term identifier. | | classifiers | | Array of object | Classifiers to use to auto-detect glossary terms. If more than one classifier, infotype predictions from the classifier defined later in sequence take precedance. | [{'type': 'datahub', 'config': None}] | | table_pattern | | AllowDenyPattern (see below for fields) | Regex patterns to filter tables for classification. This is used in combination with other patterns in parent config. Specify regex to match the entire table name in `database.schema.table` format. e.g. to match all tables starting with customer in Customer database and public schema, use the regex 'Customer.public.customer.*' | {'allow': ['.*'], 'deny': [], 'ignoreCase': True} | diff --git a/metadata-ingestion/src/datahub/ingestion/glossary/classifier.py b/metadata-ingestion/src/datahub/ingestion/glossary/classifier.py index 99789a49c0b43..ddcb74e354613 100644 --- a/metadata-ingestion/src/datahub/ingestion/glossary/classifier.py +++ b/metadata-ingestion/src/datahub/ingestion/glossary/classifier.py @@ -39,7 +39,7 @@ class ClassificationConfig(ConfigModel): max_workers: int = Field( default=(os.cpu_count() or 4), - description="Number of worker threads to use for classification. Set to 1 to disable.", + description="Number of worker processes to use for classification. Set to 1 to disable.", ) table_pattern: AllowDenyPattern = Field( diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery.py index 5046f52cdce26..7a96b2f0643ab 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery.py @@ -2,24 +2,9 @@ import functools import logging import os -import re -import traceback -from collections import defaultdict -from datetime import datetime, timedelta -from typing import Dict, Iterable, List, Optional, Set, Type, Union, cast +from typing import Iterable, List, Optional -from google.cloud import bigquery -from google.cloud.bigquery.table import TableListItem - -from datahub.configuration.pattern_utils import is_schema_allowed, is_tag_allowed -from datahub.emitter.mce_builder import ( - make_data_platform_urn, - make_dataplatform_instance_urn, - make_dataset_urn, - make_tag_urn, -) -from datahub.emitter.mcp import MetadataChangeProposalWrapper -from datahub.emitter.mcp_builder import BigQueryDatasetKey, ContainerKey, ProjectIdKey +from datahub.emitter.mce_builder import make_dataset_urn from datahub.ingestion.api.common import PipelineContext from datahub.ingestion.api.decorators import ( SupportStatus, @@ -30,54 +15,31 @@ ) from datahub.ingestion.api.incremental_lineage_helper import auto_incremental_lineage from datahub.ingestion.api.source import ( - CapabilityReport, MetadataWorkUnitProcessor, SourceCapability, TestableSource, TestConnectionReport, ) from datahub.ingestion.api.workunit import MetadataWorkUnit -from datahub.ingestion.glossary.classification_mixin import ( - SAMPLE_SIZE_MULTIPLIER, - ClassificationHandler, - classification_workunit_processor, -) from datahub.ingestion.source.bigquery_v2.bigquery_audit import ( BigqueryTableIdentifier, BigQueryTableRef, ) from datahub.ingestion.source.bigquery_v2.bigquery_config import BigQueryV2Config -from datahub.ingestion.source.bigquery_v2.bigquery_data_reader import BigQueryDataReader -from datahub.ingestion.source.bigquery_v2.bigquery_helper import ( - unquote_and_decode_unicode_escape_seq, -) from datahub.ingestion.source.bigquery_v2.bigquery_report import BigQueryV2Report from datahub.ingestion.source.bigquery_v2.bigquery_schema import ( - BigqueryColumn, - BigqueryDataset, BigqueryProject, BigQuerySchemaApi, - BigqueryTable, - BigqueryTableSnapshot, - BigqueryView, ) -from datahub.ingestion.source.bigquery_v2.common import ( - BQ_EXTERNAL_DATASET_URL_TEMPLATE, - BQ_EXTERNAL_TABLE_URL_TEMPLATE, +from datahub.ingestion.source.bigquery_v2.bigquery_schema_gen import ( + BigQuerySchemaGenerator, +) +from datahub.ingestion.source.bigquery_v2.bigquery_test_connection import ( + BigQueryTestConnection, ) from datahub.ingestion.source.bigquery_v2.lineage import BigqueryLineageExtractor from datahub.ingestion.source.bigquery_v2.profiler import BigqueryProfiler from datahub.ingestion.source.bigquery_v2.usage import BigQueryUsageExtractor -from datahub.ingestion.source.common.subtypes import ( - DatasetContainerSubTypes, - DatasetSubTypes, -) -from datahub.ingestion.source.sql.sql_utils import ( - add_table_to_schema_container, - gen_database_container, - gen_schema_container, - get_domain_wu, -) from datahub.ingestion.source.state.profiling_state_handler import ProfilingHandler from datahub.ingestion.source.state.redundant_run_skip_handler import ( RedundantLineageRunSkipHandler, @@ -89,57 +51,11 @@ from datahub.ingestion.source.state.stateful_ingestion_base import ( StatefulIngestionSourceBase, ) -from datahub.ingestion.source_report.ingestion_stage import ( - METADATA_EXTRACTION, - PROFILING, -) -from datahub.metadata.com.linkedin.pegasus2avro.common import ( - Status, - SubTypes, - TimeStamp, -) -from datahub.metadata.com.linkedin.pegasus2avro.dataset import ( - DatasetProperties, - ViewProperties, -) -from datahub.metadata.com.linkedin.pegasus2avro.schema import ( - ArrayType, - BooleanType, - BytesType, - DateType, - MySqlDDL, - NullType, - NumberType, - RecordType, - SchemaField, - SchemaFieldDataType, - SchemaMetadata, - StringType, - TimeType, -) -from datahub.metadata.schema_classes import ( - DataPlatformInstanceClass, - GlobalTagsClass, - TagAssociationClass, -) from datahub.sql_parsing.schema_resolver import SchemaResolver -from datahub.utilities.file_backed_collections import FileBackedDict -from datahub.utilities.hive_schema_to_avro import ( - HiveColumnToAvroConverter, - get_schema_fields_for_hive_column, -) -from datahub.utilities.mapping import Constants -from datahub.utilities.perf_timer import PerfTimer -from datahub.utilities.ratelimiter import RateLimiter from datahub.utilities.registries.domain_registry import DomainRegistry logger: logging.Logger = logging.getLogger(__name__) -# Handle table snapshots -# See https://cloud.google.com/bigquery/docs/table-snapshots-intro. -SNAPSHOT_TABLE_REGEX = re.compile(r"^(.+)@(\d{13})$") -CLUSTERING_COLUMN_TAG = "CLUSTERING_COLUMN" - # We can't use close as it is not called if the ingestion is not successful def cleanup(config: BigQueryV2Config) -> None: @@ -178,58 +94,18 @@ def cleanup(config: BigQueryV2Config) -> None: supported=True, ) class BigqueryV2Source(StatefulIngestionSourceBase, TestableSource): - # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types - # Note: We use the hive schema parser to parse nested BigQuery types. We also have - # some extra type mappings in that file. - BIGQUERY_FIELD_TYPE_MAPPINGS: Dict[ - str, - Type[ - Union[ - ArrayType, - BytesType, - BooleanType, - NumberType, - RecordType, - StringType, - TimeType, - DateType, - NullType, - ] - ], - ] = { - "BYTES": BytesType, - "BOOL": BooleanType, - "DECIMAL": NumberType, - "NUMERIC": NumberType, - "BIGNUMERIC": NumberType, - "BIGDECIMAL": NumberType, - "FLOAT64": NumberType, - "INT": NumberType, - "INT64": NumberType, - "SMALLINT": NumberType, - "INTEGER": NumberType, - "BIGINT": NumberType, - "TINYINT": NumberType, - "BYTEINT": NumberType, - "STRING": StringType, - "TIME": TimeType, - "TIMESTAMP": TimeType, - "DATE": DateType, - "DATETIME": TimeType, - "GEOGRAPHY": NullType, - "JSON": RecordType, - "INTERVAL": NullType, - "ARRAY": ArrayType, - "STRUCT": RecordType, - } - def __init__(self, ctx: PipelineContext, config: BigQueryV2Config): super().__init__(config, ctx) self.config: BigQueryV2Config = config self.report: BigQueryV2Report = BigQueryV2Report() - self.classification_handler = ClassificationHandler(self.config, self.report) self.platform: str = "bigquery" + self.domain_registry: Optional[DomainRegistry] = None + if self.config.domain: + self.domain_registry = DomainRegistry( + cached_domains=[k for k in self.config.domain], graph=self.ctx.graph + ) + BigqueryTableIdentifier._BIGQUERY_DEFAULT_SHARDED_TABLE_REGEX = ( self.config.sharded_table_pattern ) @@ -247,12 +123,6 @@ def __init__(self, ctx: PipelineContext, config: BigQueryV2Config): self.sql_parser_schema_resolver = self._init_schema_resolver() - self.data_reader: Optional[BigQueryDataReader] = None - if self.classification_handler.is_classification_enabled(): - self.data_reader = BigQueryDataReader.create( - self.config.get_bigquery_client() - ) - redundant_lineage_run_skip_handler: Optional[ RedundantLineageRunSkipHandler ] = None @@ -289,12 +159,6 @@ def __init__(self, ctx: PipelineContext, config: BigQueryV2Config): redundant_run_skip_handler=redundant_usage_run_skip_handler, ) - self.domain_registry: Optional[DomainRegistry] = None - if self.config.domain: - self.domain_registry = DomainRegistry( - cached_domains=[k for k in self.config.domain], graph=self.ctx.graph - ) - self.profiling_state_handler: Optional[ProfilingHandler] = None if self.config.enable_stateful_profiling: self.profiling_state_handler = ProfilingHandler( @@ -307,17 +171,15 @@ def __init__(self, ctx: PipelineContext, config: BigQueryV2Config): config, self.report, self.profiling_state_handler ) - # Global store of table identifiers for lineage filtering - self.table_refs: Set[str] = set() - - # Maps project -> view_ref, so we can find all views in a project - self.view_refs_by_project: Dict[str, Set[str]] = defaultdict(set) - # Maps project -> snapshot_ref, so we can find all snapshots in a project - self.snapshot_refs_by_project: Dict[str, Set[str]] = defaultdict(set) - # Maps view ref -> actual sql - self.view_definitions: FileBackedDict[str] = FileBackedDict() - # Maps snapshot ref -> Snapshot - self.snapshots_by_ref: FileBackedDict[BigqueryTableSnapshot] = FileBackedDict() + self.bq_schema_extractor = BigQuerySchemaGenerator( + self.config, + self.report, + self.bigquery_data_dictionary, + self.domain_registry, + self.sql_parser_schema_resolver, + self.profiler, + self.gen_dataset_urn, + ) self.add_config_to_report() atexit.register(cleanup, config) @@ -327,161 +189,9 @@ def create(cls, config_dict: dict, ctx: PipelineContext) -> "BigqueryV2Source": config = BigQueryV2Config.parse_obj(config_dict) return cls(ctx, config) - @staticmethod - def connectivity_test(client: bigquery.Client) -> CapabilityReport: - ret = client.query("select 1") - if ret.error_result: - return CapabilityReport( - capable=False, failure_reason=f"{ret.error_result['message']}" - ) - else: - return CapabilityReport(capable=True) - - @property - def store_table_refs(self): - return self.config.include_table_lineage or self.config.include_usage_statistics - - @staticmethod - def metadata_read_capability_test( - project_ids: List[str], config: BigQueryV2Config - ) -> CapabilityReport: - for project_id in project_ids: - try: - logger.info(f"Metadata read capability test for project {project_id}") - client: bigquery.Client = config.get_bigquery_client() - assert client - bigquery_data_dictionary = BigQuerySchemaApi( - BigQueryV2Report().schema_api_perf, client - ) - result = bigquery_data_dictionary.get_datasets_for_project_id( - project_id, 10 - ) - if len(result) == 0: - return CapabilityReport( - capable=False, - failure_reason=f"Dataset query returned empty dataset. It is either empty or no dataset in project {project_id}", - ) - tables = bigquery_data_dictionary.get_tables_for_dataset( - project_id=project_id, - dataset_name=result[0].name, - tables={}, - with_data_read_permission=config.have_table_data_read_permission, - ) - if len(list(tables)) == 0: - return CapabilityReport( - capable=False, - failure_reason=f"Tables query did not return any table. It is either empty or no tables in project {project_id}.{result[0].name}", - ) - - except Exception as e: - return CapabilityReport( - capable=False, - failure_reason=f"Dataset query failed with error: {e}", - ) - - return CapabilityReport(capable=True) - - @staticmethod - def lineage_capability_test( - connection_conf: BigQueryV2Config, - project_ids: List[str], - report: BigQueryV2Report, - ) -> CapabilityReport: - lineage_extractor = BigqueryLineageExtractor( - connection_conf, report, lambda ref: "" - ) - for project_id in project_ids: - try: - logger.info(f"Lineage capability test for project {project_id}") - lineage_extractor.test_capability(project_id) - except Exception as e: - return CapabilityReport( - capable=False, - failure_reason=f"Lineage capability test failed with: {e}", - ) - - return CapabilityReport(capable=True) - - @staticmethod - def usage_capability_test( - connection_conf: BigQueryV2Config, - project_ids: List[str], - report: BigQueryV2Report, - ) -> CapabilityReport: - usage_extractor = BigQueryUsageExtractor( - connection_conf, - report, - schema_resolver=SchemaResolver(platform="bigquery"), - dataset_urn_builder=lambda ref: "", - ) - for project_id in project_ids: - try: - logger.info(f"Usage capability test for project {project_id}") - failures_before_test = len(report.failures) - usage_extractor.test_capability(project_id) - if failures_before_test != len(report.failures): - return CapabilityReport( - capable=False, - failure_reason="Usage capability test failed. Check the logs for further info", - ) - except Exception as e: - return CapabilityReport( - capable=False, - failure_reason=f"Usage capability test failed with: {e} for project {project_id}", - ) - return CapabilityReport(capable=True) - @staticmethod def test_connection(config_dict: dict) -> TestConnectionReport: - test_report = TestConnectionReport() - _report: Dict[Union[SourceCapability, str], CapabilityReport] = dict() - - try: - connection_conf = BigQueryV2Config.parse_obj_allow_extras(config_dict) - client: bigquery.Client = connection_conf.get_bigquery_client() - assert client - - test_report.basic_connectivity = BigqueryV2Source.connectivity_test(client) - - connection_conf.start_time = datetime.now() - connection_conf.end_time = datetime.now() + timedelta(minutes=1) - - report: BigQueryV2Report = BigQueryV2Report() - project_ids: List[str] = [] - projects = client.list_projects() - - for project in projects: - if connection_conf.project_id_pattern.allowed(project.project_id): - project_ids.append(project.project_id) - - metadata_read_capability = BigqueryV2Source.metadata_read_capability_test( - project_ids, connection_conf - ) - if SourceCapability.SCHEMA_METADATA not in _report: - _report[SourceCapability.SCHEMA_METADATA] = metadata_read_capability - - if connection_conf.include_table_lineage: - lineage_capability = BigqueryV2Source.lineage_capability_test( - connection_conf, project_ids, report - ) - if SourceCapability.LINEAGE_COARSE not in _report: - _report[SourceCapability.LINEAGE_COARSE] = lineage_capability - - if connection_conf.include_usage_statistics: - usage_capability = BigqueryV2Source.usage_capability_test( - connection_conf, project_ids, report - ) - if SourceCapability.USAGE_STATS not in _report: - _report[SourceCapability.USAGE_STATS] = usage_capability - - test_report.capability_report = _report - return test_report - - except Exception as e: - test_report.basic_connectivity = CapabilityReport( - capable=False, failure_reason=f"{e}" - ) - return test_report + return BigQueryTestConnection.test_connection(config_dict) def _init_schema_resolver(self) -> SchemaResolver: schema_resolution_required = ( @@ -509,83 +219,6 @@ def _init_schema_resolver(self) -> SchemaResolver: ) return SchemaResolver(platform=self.platform, env=self.config.env) - def get_dataplatform_instance_aspect( - self, dataset_urn: str, project_id: str - ) -> MetadataWorkUnit: - aspect = DataPlatformInstanceClass( - platform=make_data_platform_urn(self.platform), - instance=( - make_dataplatform_instance_urn(self.platform, project_id) - if self.config.include_data_platform_instance - else None - ), - ) - return MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=aspect - ).as_workunit() - - def gen_dataset_key(self, db_name: str, schema: str) -> ContainerKey: - return BigQueryDatasetKey( - project_id=db_name, - dataset_id=schema, - platform=self.platform, - env=self.config.env, - backcompat_env_as_instance=True, - ) - - def gen_project_id_key(self, database: str) -> ContainerKey: - return ProjectIdKey( - project_id=database, - platform=self.platform, - env=self.config.env, - backcompat_env_as_instance=True, - ) - - def gen_project_id_containers(self, database: str) -> Iterable[MetadataWorkUnit]: - database_container_key = self.gen_project_id_key(database) - - yield from gen_database_container( - database=database, - name=database, - sub_types=[DatasetContainerSubTypes.BIGQUERY_PROJECT], - domain_registry=self.domain_registry, - domain_config=self.config.domain, - database_container_key=database_container_key, - ) - - def gen_dataset_containers( - self, dataset: str, project_id: str, tags: Optional[Dict[str, str]] = None - ) -> Iterable[MetadataWorkUnit]: - schema_container_key = self.gen_dataset_key(project_id, dataset) - - tags_joined: Optional[List[str]] = None - if tags and self.config.capture_dataset_label_as_tag: - tags_joined = [ - f"{k}:{v}" - for k, v in tags.items() - if is_tag_allowed(self.config.capture_dataset_label_as_tag, k) - ] - - database_container_key = self.gen_project_id_key(database=project_id) - - yield from gen_schema_container( - database=project_id, - schema=dataset, - sub_types=[DatasetContainerSubTypes.BIGQUERY_DATASET], - domain_registry=self.domain_registry, - domain_config=self.config.domain, - schema_container_key=schema_container_key, - database_container_key=database_container_key, - external_url=( - BQ_EXTERNAL_DATASET_URL_TEMPLATE.format( - project=project_id, dataset=dataset - ) - if self.config.include_external_url - else None - ), - tags=tags_joined, - ) - def get_workunit_processors(self) -> List[Optional[MetadataWorkUnitProcessor]]: return [ *super().get_workunit_processors(), @@ -603,25 +236,23 @@ def get_workunits_internal(self) -> Iterable[MetadataWorkUnit]: return if self.config.include_schema_metadata: - for project_id in projects: - self.report.set_ingestion_stage(project_id.id, METADATA_EXTRACTION) - logger.info(f"Processing project: {project_id.id}") - yield from self._process_project(project_id) + for project in projects: + yield from self.bq_schema_extractor.get_project_workunits(project) if self.config.include_usage_statistics: yield from self.usage_extractor.get_usage_workunits( - [p.id for p in projects], self.table_refs + [p.id for p in projects], self.bq_schema_extractor.table_refs ) if self.config.include_table_lineage: yield from self.lineage_extractor.get_lineage_workunits( [p.id for p in projects], self.sql_parser_schema_resolver, - self.view_refs_by_project, - self.view_definitions, - self.snapshot_refs_by_project, - self.snapshots_by_ref, - self.table_refs, + self.bq_schema_extractor.view_refs_by_project, + self.bq_schema_extractor.view_definitions, + self.bq_schema_extractor.snapshot_refs_by_project, + self.bq_schema_extractor.snapshots_by_ref, + self.bq_schema_extractor.table_refs, ) def _get_projects(self) -> List[BigqueryProject]: @@ -636,15 +267,25 @@ def _get_projects(self) -> List[BigqueryProject]: return list(self._query_project_list()) def _query_project_list(self) -> Iterable[BigqueryProject]: - projects = self.bigquery_data_dictionary.get_projects() - if not projects: # Report failure on exception and if empty list is returned - self.report.report_failure( - "metadata-extraction", - "Get projects didn't return any project. " - "Maybe resourcemanager.projects.get permission is missing for the service account. " + try: + projects = self.bigquery_data_dictionary.get_projects() + + if ( + not projects + ): # Report failure on exception and if empty list is returned + self.report.failure( + title="Get projects didn't return any project. ", + message="Maybe resourcemanager.projects.get permission is missing for the service account. " + "You can assign predefined roles/bigquery.metadataViewer role to your service account.", + ) + except Exception as e: + self.report.failure( + title="Failed to get BigQuery Projects", + message="Maybe resourcemanager.projects.get permission is missing for the service account. " "You can assign predefined roles/bigquery.metadataViewer role to your service account.", + exc=e, ) - return + projects = [] for project in projects: if self.config.project_id_pattern.allowed(project.id): @@ -652,567 +293,6 @@ def _query_project_list(self) -> Iterable[BigqueryProject]: else: self.report.report_dropped(project.id) - def _process_project( - self, bigquery_project: BigqueryProject - ) -> Iterable[MetadataWorkUnit]: - db_tables: Dict[str, List[BigqueryTable]] = {} - db_views: Dict[str, List[BigqueryView]] = {} - db_snapshots: Dict[str, List[BigqueryTableSnapshot]] = {} - - project_id = bigquery_project.id - try: - bigquery_project.datasets = ( - self.bigquery_data_dictionary.get_datasets_for_project_id(project_id) - ) - except Exception as e: - error_message = f"Unable to get datasets for project {project_id}, skipping. The error was: {e}" - if self.config.is_profiling_enabled(): - error_message = f"Unable to get datasets for project {project_id}, skipping. Does your service account has bigquery.datasets.get permission? The error was: {e}" - logger.error(error_message) - self.report.report_failure( - "metadata-extraction", - f"{project_id} - {error_message}", - ) - return None - - if len(bigquery_project.datasets) == 0: - more_info = ( - "Either there are no datasets in this project or missing bigquery.datasets.get permission. " - "You can assign predefined roles/bigquery.metadataViewer role to your service account." - ) - if self.config.exclude_empty_projects: - self.report.report_dropped(project_id) - warning_message = f"Excluded project '{project_id}' since no were datasets found. {more_info}" - else: - yield from self.gen_project_id_containers(project_id) - warning_message = ( - f"No datasets found in project '{project_id}'. {more_info}" - ) - logger.warning(warning_message) - return - - yield from self.gen_project_id_containers(project_id) - - self.report.num_project_datasets_to_scan[project_id] = len( - bigquery_project.datasets - ) - for bigquery_dataset in bigquery_project.datasets: - if not is_schema_allowed( - self.config.dataset_pattern, - bigquery_dataset.name, - project_id, - self.config.match_fully_qualified_names, - ): - self.report.report_dropped(f"{bigquery_dataset.name}.*") - continue - try: - # db_tables, db_views, and db_snapshots are populated in the this method - yield from self._process_schema( - project_id, bigquery_dataset, db_tables, db_views, db_snapshots - ) - - except Exception as e: - error_message = f"Unable to get tables for dataset {bigquery_dataset.name} in project {project_id}, skipping. Does your service account has bigquery.tables.list, bigquery.routines.get, bigquery.routines.list permission? The error was: {e}" - if self.config.is_profiling_enabled(): - error_message = f"Unable to get tables for dataset {bigquery_dataset.name} in project {project_id}, skipping. Does your service account has bigquery.tables.list, bigquery.routines.get, bigquery.routines.list permission, bigquery.tables.getData permission? The error was: {e}" - - trace = traceback.format_exc() - logger.error(trace) - logger.error(error_message) - self.report.report_failure( - "metadata-extraction", - f"{project_id}.{bigquery_dataset.name} - {error_message} - {trace}", - ) - continue - - if self.config.is_profiling_enabled(): - logger.info(f"Starting profiling project {project_id}") - self.report.set_ingestion_stage(project_id, PROFILING) - yield from self.profiler.get_workunits( - project_id=project_id, - tables=db_tables, - ) - - def _process_schema( - self, - project_id: str, - bigquery_dataset: BigqueryDataset, - db_tables: Dict[str, List[BigqueryTable]], - db_views: Dict[str, List[BigqueryView]], - db_snapshots: Dict[str, List[BigqueryTableSnapshot]], - ) -> Iterable[MetadataWorkUnit]: - dataset_name = bigquery_dataset.name - - yield from self.gen_dataset_containers( - dataset_name, project_id, bigquery_dataset.labels - ) - - columns = None - - rate_limiter: Optional[RateLimiter] = None - if self.config.rate_limit: - rate_limiter = RateLimiter( - max_calls=self.config.requests_per_min, period=60 - ) - - if ( - self.config.include_tables - or self.config.include_views - or self.config.include_table_snapshots - ): - columns = self.bigquery_data_dictionary.get_columns_for_dataset( - project_id=project_id, - dataset_name=dataset_name, - column_limit=self.config.column_limit, - run_optimized_column_query=self.config.run_optimized_column_query, - extract_policy_tags_from_catalog=self.config.extract_policy_tags_from_catalog, - report=self.report, - rate_limiter=rate_limiter, - ) - - if self.config.include_tables: - db_tables[dataset_name] = list( - self.get_tables_for_dataset(project_id, dataset_name) - ) - - for table in db_tables[dataset_name]: - table_columns = columns.get(table.name, []) if columns else [] - table_wu_generator = self._process_table( - table=table, - columns=table_columns, - project_id=project_id, - dataset_name=dataset_name, - ) - yield from classification_workunit_processor( - table_wu_generator, - self.classification_handler, - self.data_reader, - [project_id, dataset_name, table.name], - data_reader_kwargs=dict( - sample_size_percent=( - self.config.classification.sample_size - * SAMPLE_SIZE_MULTIPLIER - / table.rows_count - if table.rows_count - else None - ) - ), - ) - elif self.store_table_refs: - # Need table_refs to calculate lineage and usage - for table_item in self.bigquery_data_dictionary.list_tables( - dataset_name, project_id - ): - identifier = BigqueryTableIdentifier( - project_id=project_id, - dataset=dataset_name, - table=table_item.table_id, - ) - if not self.config.table_pattern.allowed(identifier.raw_table_name()): - self.report.report_dropped(identifier.raw_table_name()) - continue - try: - self.table_refs.add( - str(BigQueryTableRef(identifier).get_sanitized_table_ref()) - ) - except Exception as e: - logger.warning( - f"Could not create table ref for {table_item.path}: {e}" - ) - - if self.config.include_views: - db_views[dataset_name] = list( - self.bigquery_data_dictionary.get_views_for_dataset( - project_id, - dataset_name, - self.config.is_profiling_enabled(), - self.report, - ) - ) - - for view in db_views[dataset_name]: - view_columns = columns.get(view.name, []) if columns else [] - yield from self._process_view( - view=view, - columns=view_columns, - project_id=project_id, - dataset_name=dataset_name, - ) - - if self.config.include_table_snapshots: - db_snapshots[dataset_name] = list( - self.bigquery_data_dictionary.get_snapshots_for_dataset( - project_id, - dataset_name, - self.config.is_profiling_enabled(), - self.report, - ) - ) - - for snapshot in db_snapshots[dataset_name]: - snapshot_columns = columns.get(snapshot.name, []) if columns else [] - yield from self._process_snapshot( - snapshot=snapshot, - columns=snapshot_columns, - project_id=project_id, - dataset_name=dataset_name, - ) - - # This method is used to generate the ignore list for datatypes the profiler doesn't support we have to do it here - # because the profiler doesn't have access to columns - def generate_profile_ignore_list(self, columns: List[BigqueryColumn]) -> List[str]: - ignore_list: List[str] = [] - for column in columns: - if not column.data_type or any( - word in column.data_type.lower() - for word in ["array", "struct", "geography", "json"] - ): - ignore_list.append(column.field_path) - return ignore_list - - def _process_table( - self, - table: BigqueryTable, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - table_identifier = BigqueryTableIdentifier(project_id, dataset_name, table.name) - - self.report.report_entity_scanned(table_identifier.raw_table_name()) - - if not self.config.table_pattern.allowed(table_identifier.raw_table_name()): - self.report.report_dropped(table_identifier.raw_table_name()) - return - - if self.store_table_refs: - self.table_refs.add( - str(BigQueryTableRef(table_identifier).get_sanitized_table_ref()) - ) - table.column_count = len(columns) - - # We only collect profile ignore list if profiling is enabled and profile_table_level_only is false - if ( - self.config.is_profiling_enabled() - and not self.config.profiling.profile_table_level_only - ): - table.columns_ignore_from_profiling = self.generate_profile_ignore_list( - columns - ) - - if not table.column_count: - logger.warning( - f"Table doesn't have any column or unable to get columns for table: {table_identifier}" - ) - - # If table has time partitioning, set the data type of the partitioning field - if table.partition_info: - table.partition_info.column = next( - ( - column - for column in columns - if column.name == table.partition_info.field - ), - None, - ) - yield from self.gen_table_dataset_workunits( - table, columns, project_id, dataset_name - ) - - def _process_view( - self, - view: BigqueryView, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - table_identifier = BigqueryTableIdentifier(project_id, dataset_name, view.name) - - self.report.report_entity_scanned(table_identifier.raw_table_name(), "view") - - if not self.config.view_pattern.allowed(table_identifier.raw_table_name()): - self.report.report_dropped(table_identifier.raw_table_name()) - return - - if self.store_table_refs: - table_ref = str( - BigQueryTableRef(table_identifier).get_sanitized_table_ref() - ) - self.table_refs.add(table_ref) - if self.config.lineage_parse_view_ddl and view.view_definition: - self.view_refs_by_project[project_id].add(table_ref) - self.view_definitions[table_ref] = view.view_definition - - view.column_count = len(columns) - if not view.column_count: - logger.warning( - f"View doesn't have any column or unable to get columns for table: {table_identifier}" - ) - - yield from self.gen_view_dataset_workunits( - table=view, - columns=columns, - project_id=project_id, - dataset_name=dataset_name, - ) - - def _process_snapshot( - self, - snapshot: BigqueryTableSnapshot, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - table_identifier = BigqueryTableIdentifier( - project_id, dataset_name, snapshot.name - ) - - self.report.snapshots_scanned += 1 - - if not self.config.table_snapshot_pattern.allowed( - table_identifier.raw_table_name() - ): - self.report.report_dropped(table_identifier.raw_table_name()) - return - - snapshot.columns = columns - snapshot.column_count = len(columns) - if not snapshot.column_count: - logger.warning( - f"Snapshot doesn't have any column or unable to get columns for table: {table_identifier}" - ) - - if self.store_table_refs: - table_ref = str( - BigQueryTableRef(table_identifier).get_sanitized_table_ref() - ) - self.table_refs.add(table_ref) - if snapshot.base_table_identifier: - self.snapshot_refs_by_project[project_id].add(table_ref) - self.snapshots_by_ref[table_ref] = snapshot - - yield from self.gen_snapshot_dataset_workunits( - table=snapshot, - columns=columns, - project_id=project_id, - dataset_name=dataset_name, - ) - - def gen_table_dataset_workunits( - self, - table: BigqueryTable, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - custom_properties: Dict[str, str] = {} - if table.expires: - custom_properties["expiration_date"] = str(table.expires) - - if table.partition_info: - custom_properties["partition_info"] = str(table.partition_info) - - if table.size_in_bytes: - custom_properties["size_in_bytes"] = str(table.size_in_bytes) - - if table.active_billable_bytes: - custom_properties["billable_bytes_active"] = str( - table.active_billable_bytes - ) - - if table.long_term_billable_bytes: - custom_properties["billable_bytes_long_term"] = str( - table.long_term_billable_bytes - ) - - if table.max_partition_id: - custom_properties["number_of_partitions"] = str(table.num_partitions) - custom_properties["max_partition_id"] = str(table.max_partition_id) - custom_properties["is_partitioned"] = str(True) - - sub_types: List[str] = [DatasetSubTypes.TABLE] - if table.max_shard_id: - custom_properties["max_shard_id"] = str(table.max_shard_id) - custom_properties["is_sharded"] = str(True) - sub_types = ["sharded table"] + sub_types - - tags_to_add = None - if table.labels and self.config.capture_table_label_as_tag: - tags_to_add = [] - tags_to_add.extend( - [ - make_tag_urn(f"""{k}:{v}""") - for k, v in table.labels.items() - if is_tag_allowed(self.config.capture_table_label_as_tag, k) - ] - ) - - yield from self.gen_dataset_workunits( - table=table, - columns=columns, - project_id=project_id, - dataset_name=dataset_name, - sub_types=sub_types, - tags_to_add=tags_to_add, - custom_properties=custom_properties, - ) - - def gen_view_dataset_workunits( - self, - table: BigqueryView, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - tags_to_add = None - if table.labels and self.config.capture_view_label_as_tag: - tags_to_add = [ - make_tag_urn(f"{k}:{v}") - for k, v in table.labels.items() - if is_tag_allowed(self.config.capture_view_label_as_tag, k) - ] - yield from self.gen_dataset_workunits( - table=table, - columns=columns, - project_id=project_id, - dataset_name=dataset_name, - tags_to_add=tags_to_add, - sub_types=[DatasetSubTypes.VIEW], - ) - - view = cast(BigqueryView, table) - view_definition_string = view.view_definition - view_properties_aspect = ViewProperties( - materialized=view.materialized, - viewLanguage="SQL", - viewLogic=view_definition_string or "", - ) - yield MetadataChangeProposalWrapper( - entityUrn=self.gen_dataset_urn( - project_id=project_id, dataset_name=dataset_name, table=table.name - ), - aspect=view_properties_aspect, - ).as_workunit() - - def gen_snapshot_dataset_workunits( - self, - table: BigqueryTableSnapshot, - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - ) -> Iterable[MetadataWorkUnit]: - custom_properties: Dict[str, str] = {} - if table.ddl: - custom_properties["snapshot_ddl"] = table.ddl - if table.snapshot_time: - custom_properties["snapshot_time"] = str(table.snapshot_time) - if table.size_in_bytes: - custom_properties["size_in_bytes"] = str(table.size_in_bytes) - if table.rows_count: - custom_properties["rows_count"] = str(table.rows_count) - yield from self.gen_dataset_workunits( - table=table, - columns=columns, - project_id=project_id, - dataset_name=dataset_name, - sub_types=[DatasetSubTypes.BIGQUERY_TABLE_SNAPSHOT], - custom_properties=custom_properties, - ) - - def gen_dataset_workunits( - self, - table: Union[BigqueryTable, BigqueryView, BigqueryTableSnapshot], - columns: List[BigqueryColumn], - project_id: str, - dataset_name: str, - sub_types: List[str], - tags_to_add: Optional[List[str]] = None, - custom_properties: Optional[Dict[str, str]] = None, - ) -> Iterable[MetadataWorkUnit]: - dataset_urn = self.gen_dataset_urn( - project_id=project_id, dataset_name=dataset_name, table=table.name - ) - - status = Status(removed=False) - yield MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=status - ).as_workunit() - - datahub_dataset_name = BigqueryTableIdentifier( - project_id, dataset_name, table.name - ) - - yield self.gen_schema_metadata( - dataset_urn, table, columns, datahub_dataset_name - ) - - dataset_properties = DatasetProperties( - name=datahub_dataset_name.get_table_display_name(), - description=( - unquote_and_decode_unicode_escape_seq(table.comment) - if table.comment - else "" - ), - qualifiedName=str(datahub_dataset_name), - created=( - TimeStamp(time=int(table.created.timestamp() * 1000)) - if table.created is not None - else None - ), - lastModified=( - TimeStamp(time=int(table.last_altered.timestamp() * 1000)) - if table.last_altered is not None - else None - ), - externalUrl=( - BQ_EXTERNAL_TABLE_URL_TEMPLATE.format( - project=project_id, dataset=dataset_name, table=table.name - ) - if self.config.include_external_url - else None - ), - ) - if custom_properties: - dataset_properties.customProperties.update(custom_properties) - - yield MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=dataset_properties - ).as_workunit() - - if tags_to_add: - yield self.gen_tags_aspect_workunit(dataset_urn, tags_to_add) - - yield from add_table_to_schema_container( - dataset_urn=dataset_urn, - parent_container_key=self.gen_dataset_key(project_id, dataset_name), - ) - yield self.get_dataplatform_instance_aspect( - dataset_urn=dataset_urn, project_id=project_id - ) - - subTypes = SubTypes(typeNames=sub_types) - yield MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=subTypes - ).as_workunit() - - if self.domain_registry: - yield from get_domain_wu( - dataset_name=str(datahub_dataset_name), - entity_urn=dataset_urn, - domain_registry=self.domain_registry, - domain_config=self.config.domain, - ) - - def gen_tags_aspect_workunit( - self, dataset_urn: str, tags_to_add: List[str] - ) -> MetadataWorkUnit: - tags = GlobalTagsClass( - tags=[TagAssociationClass(tag_to_add) for tag_to_add in tags_to_add] - ) - return MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=tags - ).as_workunit() - def gen_dataset_urn( self, project_id: str, dataset_name: str, table: str, use_raw_name: bool = False ) -> str: @@ -1235,241 +315,9 @@ def gen_dataset_urn_from_raw_ref(self, ref: BigQueryTableRef) -> str: use_raw_name=True, ) - def gen_dataset_urn_from_ref(self, ref: BigQueryTableRef) -> str: - return self.gen_dataset_urn( - ref.table_identifier.project_id, - ref.table_identifier.dataset, - ref.table_identifier.table, - ) - - def gen_schema_fields(self, columns: List[BigqueryColumn]) -> List[SchemaField]: - schema_fields: List[SchemaField] = [] - - # Below line affects HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR in global scope - # TODO: Refractor this such that - # converter = HiveColumnToAvroConverter(struct_type_separator=" "); - # converter.get_schema_fields_for_hive_column(...) - original_struct_type_separator = ( - HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR - ) - HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR = " " - _COMPLEX_TYPE = re.compile("^(struct|array)") - last_id = -1 - for col in columns: - # if col.data_type is empty that means this column is part of a complex type - if col.data_type is None or _COMPLEX_TYPE.match(col.data_type.lower()): - # If the we have seen the ordinal position that most probably means we already processed this complex type - if last_id != col.ordinal_position: - schema_fields.extend( - get_schema_fields_for_hive_column( - col.name, col.data_type.lower(), description=col.comment - ) - ) - - # We have to add complex type comments to the correct level - if col.comment: - for idx, field in enumerate(schema_fields): - # Remove all the [version=2.0].[type=struct]. tags to get the field path - if ( - re.sub( - r"\[.*?\]\.", - "", - field.fieldPath.lower(), - 0, - re.MULTILINE, - ) - == col.field_path.lower() - ): - field.description = col.comment - schema_fields[idx] = field - break - else: - tags = [] - if col.is_partition_column: - tags.append( - TagAssociationClass(make_tag_urn(Constants.TAG_PARTITION_KEY)) - ) - - if col.cluster_column_position is not None: - tags.append( - TagAssociationClass( - make_tag_urn( - f"{CLUSTERING_COLUMN_TAG}_{col.cluster_column_position}" - ) - ) - ) - - if col.policy_tags: - for policy_tag in col.policy_tags: - tags.append(TagAssociationClass(make_tag_urn(policy_tag))) - field = SchemaField( - fieldPath=col.name, - type=SchemaFieldDataType( - self.BIGQUERY_FIELD_TYPE_MAPPINGS.get(col.data_type, NullType)() - ), - nativeDataType=col.data_type, - description=col.comment, - nullable=col.is_nullable, - globalTags=GlobalTagsClass(tags=tags), - ) - schema_fields.append(field) - last_id = col.ordinal_position - HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR = ( - original_struct_type_separator - ) - return schema_fields - - def gen_schema_metadata( - self, - dataset_urn: str, - table: Union[BigqueryTable, BigqueryView, BigqueryTableSnapshot], - columns: List[BigqueryColumn], - dataset_name: BigqueryTableIdentifier, - ) -> MetadataWorkUnit: - schema_metadata = SchemaMetadata( - schemaName=str(dataset_name), - platform=make_data_platform_urn(self.platform), - version=0, - hash="", - platformSchema=MySqlDDL(tableSchema=""), - # fields=[], - fields=self.gen_schema_fields(columns), - ) - - if self.config.lineage_parse_view_ddl or self.config.lineage_use_sql_parser: - self.sql_parser_schema_resolver.add_schema_metadata( - dataset_urn, schema_metadata - ) - - return MetadataChangeProposalWrapper( - entityUrn=dataset_urn, aspect=schema_metadata - ).as_workunit() - def get_report(self) -> BigQueryV2Report: return self.report - def get_tables_for_dataset( - self, - project_id: str, - dataset_name: str, - ) -> Iterable[BigqueryTable]: - # In bigquery there is no way to query all tables in a Project id - with PerfTimer() as timer: - # Partitions view throw exception if we try to query partition info for too many tables - # so we have to limit the number of tables we query partition info. - # The conn.list_tables returns table infos that information_schema doesn't contain and this - # way we can merge that info with the queried one. - # https://cloud.google.com/bigquery/docs/information-schema-partitions - max_batch_size: int = ( - self.config.number_of_datasets_process_in_batch - if not self.config.is_profiling_enabled() - else self.config.number_of_datasets_process_in_batch_if_profiling_enabled - ) - - # We get the list of tables in the dataset to get core table properties and to be able to process the tables in batches - # We collect only the latest shards from sharded tables (tables with _YYYYMMDD suffix) and ignore temporary tables - table_items = self.get_core_table_details( - dataset_name, project_id, self.config.temp_table_dataset_prefix - ) - - items_to_get: Dict[str, TableListItem] = {} - for table_item in table_items.keys(): - items_to_get[table_item] = table_items[table_item] - if len(items_to_get) % max_batch_size == 0: - yield from self.bigquery_data_dictionary.get_tables_for_dataset( - project_id, - dataset_name, - items_to_get, - with_data_read_permission=self.config.have_table_data_read_permission, - ) - items_to_get.clear() - - if items_to_get: - yield from self.bigquery_data_dictionary.get_tables_for_dataset( - project_id, - dataset_name, - items_to_get, - with_data_read_permission=self.config.have_table_data_read_permission, - ) - - self.report.metadata_extraction_sec[f"{project_id}.{dataset_name}"] = round( - timer.elapsed_seconds(), 2 - ) - - def get_core_table_details( - self, dataset_name: str, project_id: str, temp_table_dataset_prefix: str - ) -> Dict[str, TableListItem]: - table_items: Dict[str, TableListItem] = {} - # Dict to store sharded table and the last seen max shard id - sharded_tables: Dict[str, TableListItem] = {} - - for table in self.bigquery_data_dictionary.list_tables( - dataset_name, project_id - ): - table_identifier = BigqueryTableIdentifier( - project_id=project_id, - dataset=dataset_name, - table=table.table_id, - ) - - if table.table_type == "VIEW": - if ( - not self.config.include_views - or not self.config.view_pattern.allowed( - table_identifier.raw_table_name() - ) - ): - self.report.report_dropped(table_identifier.raw_table_name()) - continue - else: - if not self.config.table_pattern.allowed( - table_identifier.raw_table_name() - ): - self.report.report_dropped(table_identifier.raw_table_name()) - continue - - _, shard = BigqueryTableIdentifier.get_table_and_shard( - table_identifier.table - ) - table_name = table_identifier.get_table_name().split(".")[-1] - - # Sharded tables look like: table_20220120 - # For sharded tables we only process the latest shard and ignore the rest - # to find the latest shard we iterate over the list of tables and store the maximum shard id - # We only have one special case where the table name is a date `20220110` - # in this case we merge all these tables under dataset name as table name. - # For example some_dataset.20220110 will be turned to some_dataset.some_dataset - # It seems like there are some bigquery user who uses this non-standard way of sharding the tables. - if shard: - if table_name not in sharded_tables: - sharded_tables[table_name] = table - continue - - stored_table_identifier = BigqueryTableIdentifier( - project_id=project_id, - dataset=dataset_name, - table=sharded_tables[table_name].table_id, - ) - _, stored_shard = BigqueryTableIdentifier.get_table_and_shard( - stored_table_identifier.table - ) - # When table is none, we use dataset_name as table_name - assert stored_shard - if stored_shard < shard: - sharded_tables[table_name] = table - continue - elif str(table_identifier).startswith(temp_table_dataset_prefix): - logger.debug(f"Dropping temporary table {table_identifier.table}") - self.report.report_dropped(table_identifier.raw_table_name()) - continue - - table_items[table.table_id] = table - - # Adding maximum shards to the list of tables - table_items.update({value.table_id: value for value in sharded_tables.values()}) - - return table_items - def add_config_to_report(self): self.report.include_table_lineage = self.config.include_table_lineage self.report.use_date_sharded_audit_log_tables = ( diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_audit_log_api.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_audit_log_api.py index 75e116773df96..7d2f8ee0e1fd8 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_audit_log_api.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_audit_log_api.py @@ -66,6 +66,7 @@ def get_exported_bigquery_audit_metadata( rate_limiter = RateLimiter(max_calls=self.requests_per_min, period=60) with self.report.get_exported_log_entries as current_timer: + self.report.num_get_exported_log_entries_api_requests += 1 for dataset in bigquery_audit_metadata_datasets: logger.info( f"Start loading log entries from BigQueryAuditMetadata in {dataset}" @@ -115,6 +116,7 @@ def get_bigquery_log_entries_via_gcp_logging( ) with self.report.list_log_entries as current_timer: + self.report.num_list_log_entries_api_requests += 1 list_entries = client.list_entries( filter_=filter, page_size=log_page_size, diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_config.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_config.py index 578c9dddbd2e4..fe961dbd780f6 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_config.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_config.py @@ -24,6 +24,10 @@ logger = logging.getLogger(__name__) +DEFAULT_BQ_SCHEMA_PARALLELISM = int( + os.getenv("DATAHUB_BIGQUERY_SCHEMA_PARALLELISM", 20) +) + class BigQueryUsageConfig(BaseUsageConfig): _query_log_delay_removed = pydantic_removed_field("query_log_delay") @@ -175,12 +179,12 @@ class BigQueryV2Config( number_of_datasets_process_in_batch: int = Field( hidden_from_docs=True, - default=500, + default=10000, description="Number of table queried in batch when getting metadata. This is a low level config property which should be touched with care.", ) number_of_datasets_process_in_batch_if_profiling_enabled: int = Field( - default=200, + default=1000, description="Number of partitioned table queried in batch when getting metadata. This is a low level config property which should be touched with care. This restriction is needed because we query partitions system view which throws error if we try to touch too many tables.", ) @@ -313,6 +317,12 @@ def have_table_data_read_permission(self) -> bool: hidden_from_schema=True, ) + max_threads_dataset_parallelism: int = Field( + default=DEFAULT_BQ_SCHEMA_PARALLELISM, + description="Number of worker threads to use to parallelize BigQuery Dataset Metadata Extraction." + " Set to 1 to disable.", + ) + @root_validator(skip_on_failure=True) def profile_default_settings(cls, values: Dict) -> Dict: # Extra default SQLAlchemy option for better connection pooling and threading. diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_report.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_report.py index 8a1bf9e5f3d1d..4cfcc3922ddc3 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_report.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_report.py @@ -20,20 +20,32 @@ @dataclass class BigQuerySchemaApiPerfReport(Report): - num_list_projects: int = 0 + num_listed_projects: int = 0 num_list_projects_retry_request: int = 0 + num_list_projects_api_requests: int = 0 + num_list_datasets_api_requests: int = 0 + num_get_columns_for_dataset_api_requests: int = 0 + num_get_tables_for_dataset_api_requests: int = 0 + num_list_tables_api_requests: int = 0 + num_get_views_for_dataset_api_requests: int = 0 + num_get_snapshots_for_dataset_api_requests: int = 0 + list_projects: PerfTimer = field(default_factory=PerfTimer) list_datasets: PerfTimer = field(default_factory=PerfTimer) - get_columns_for_dataset: PerfTimer = field(default_factory=PerfTimer) - get_tables_for_dataset: PerfTimer = field(default_factory=PerfTimer) - list_tables: PerfTimer = field(default_factory=PerfTimer) - get_views_for_dataset: PerfTimer = field(default_factory=PerfTimer) - get_snapshots_for_dataset: PerfTimer = field(default_factory=PerfTimer) + + get_columns_for_dataset_sec: float = 0 + get_tables_for_dataset_sec: float = 0 + list_tables_sec: float = 0 + get_views_for_dataset_sec: float = 0 + get_snapshots_for_dataset_sec: float = 0 @dataclass class BigQueryAuditLogApiPerfReport(Report): + num_get_exported_log_entries_api_requests: int = 0 get_exported_log_entries: PerfTimer = field(default_factory=PerfTimer) + + num_list_log_entries_api_requests: int = 0 list_log_entries: PerfTimer = field(default_factory=PerfTimer) @@ -85,7 +97,6 @@ class BigQueryV2Report( num_usage_parsed_log_entries: TopKDict[str, int] = field( default_factory=int_top_k_dict ) - usage_error_count: Dict[str, int] = field(default_factory=int_top_k_dict) num_usage_resources_dropped: int = 0 num_usage_operations_dropped: int = 0 diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema.py index fe9bbc134a147..7bb9becfc9a0d 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema.py @@ -24,6 +24,7 @@ BigqueryTableType, ) from datahub.ingestion.source.sql.sql_generic import BaseColumn, BaseTable, BaseView +from datahub.utilities.perf_timer import PerfTimer from datahub.utilities.ratelimiter import RateLimiter logger: logging.Logger = logging.getLogger(__name__) @@ -163,33 +164,31 @@ def _should_retry(exc: BaseException) -> bool: return True with self.report.list_projects: - try: - # Bigquery API has limit in calling project.list request i.e. 2 request per second. - # https://cloud.google.com/bigquery/quotas#api_request_quotas - # Whenever this limit reached an exception occur with msg - # 'Quota exceeded: Your user exceeded quota for concurrent project.lists requests.' - # Hence, added the api request retry of 15 min. - # We already tried adding rate_limit externally, proving max_result and page_size - # to restrict the request calls inside list_project but issue still occured. - projects_iterator = self.bq_client.list_projects( - retry=retry.Retry( - predicate=_should_retry, initial=10, maximum=180, timeout=900 - ) + self.report.num_list_projects_api_requests += 1 + # Bigquery API has limit in calling project.list request i.e. 2 request per second. + # https://cloud.google.com/bigquery/quotas#api_request_quotas + # Whenever this limit reached an exception occur with msg + # 'Quota exceeded: Your user exceeded quota for concurrent project.lists requests.' + # Hence, added the api request retry of 15 min. + # We already tried adding rate_limit externally, proving max_result and page_size + # to restrict the request calls inside list_project but issue still occured. + projects_iterator = self.bq_client.list_projects( + retry=retry.Retry( + predicate=_should_retry, initial=10, maximum=180, timeout=900 ) - projects: List[BigqueryProject] = [ - BigqueryProject(id=p.project_id, name=p.friendly_name) - for p in projects_iterator - ] - self.report.num_list_projects = len(projects) - return projects - except Exception as e: - logger.error(f"Error getting projects. {e}", exc_info=True) - return [] + ) + projects: List[BigqueryProject] = [ + BigqueryProject(id=p.project_id, name=p.friendly_name) + for p in projects_iterator + ] + self.report.num_listed_projects = len(projects) + return projects def get_datasets_for_project_id( self, project_id: str, maxResults: Optional[int] = None ) -> List[BigqueryDataset]: with self.report.list_datasets: + self.report.num_list_datasets_api_requests += 1 datasets = self.bq_client.list_datasets(project_id, max_results=maxResults) return [ BigqueryDataset(name=d.dataset_id, labels=d.labels) for d in datasets @@ -222,50 +221,42 @@ def get_datasets_for_project_id_with_information_schema( def list_tables( self, dataset_name: str, project_id: str ) -> Iterator[TableListItem]: - with self.report.list_tables as current_timer: + with PerfTimer() as current_timer: for table in self.bq_client.list_tables(f"{project_id}.{dataset_name}"): with current_timer.pause(): yield table + self.report.num_list_tables_api_requests += 1 + self.report.list_tables_sec += current_timer.elapsed_seconds() def get_tables_for_dataset( self, project_id: str, dataset_name: str, tables: Dict[str, TableListItem], + report: BigQueryV2Report, with_data_read_permission: bool = False, - report: Optional[BigQueryV2Report] = None, ) -> Iterator[BigqueryTable]: - with self.report.get_tables_for_dataset as current_timer: + with PerfTimer() as current_timer: filter_clause: str = ", ".join(f"'{table}'" for table in tables.keys()) if with_data_read_permission: - # Tables are ordered by name and table suffix to make sure we always process the latest sharded table - # and skip the others. Sharded tables are tables with suffix _20220102 - cur = self.get_query_result( - BigqueryQuery.tables_for_dataset.format( - project_id=project_id, - dataset_name=dataset_name, - table_filter=( - f" and t.table_name in ({filter_clause})" - if filter_clause - else "" - ), - ), - ) + query_template = BigqueryQuery.tables_for_dataset else: - # Tables are ordered by name and table suffix to make sure we always process the latest sharded table - # and skip the others. Sharded tables are tables with suffix _20220102 - cur = self.get_query_result( - BigqueryQuery.tables_for_dataset_without_partition_data.format( - project_id=project_id, - dataset_name=dataset_name, - table_filter=( - f" and t.table_name in ({filter_clause})" - if filter_clause - else "" - ), + query_template = BigqueryQuery.tables_for_dataset_without_partition_data + + # Tables are ordered by name and table suffix to make sure we always process the latest sharded table + # and skip the others. Sharded tables are tables with suffix _20220102 + cur = self.get_query_result( + query_template.format( + project_id=project_id, + dataset_name=dataset_name, + table_filter=( + f" and t.table_name in ({filter_clause})" + if filter_clause + else "" ), - ) + ), + ) for table in cur: try: @@ -275,15 +266,14 @@ def get_tables_for_dataset( ) except Exception as e: table_name = f"{project_id}.{dataset_name}.{table.table_name}" - logger.warning( - f"Error while processing table {table_name}", - exc_info=True, + report.warning( + title="Failed to process table", + message="Error encountered while processing table", + context=table_name, + exc=e, ) - if report: - report.report_warning( - "metadata-extraction", - f"Failed to get table {table_name}: {e}", - ) + self.report.num_get_tables_for_dataset_api_requests += 1 + self.report.get_tables_for_dataset_sec += current_timer.elapsed_seconds() @staticmethod def _make_bigquery_table( @@ -332,7 +322,7 @@ def get_views_for_dataset( has_data_read: bool, report: BigQueryV2Report, ) -> Iterator[BigqueryView]: - with self.report.get_views_for_dataset as current_timer: + with PerfTimer() as current_timer: if has_data_read: # If profiling is enabled cur = self.get_query_result( @@ -353,14 +343,14 @@ def get_views_for_dataset( yield BigQuerySchemaApi._make_bigquery_view(table) except Exception as e: view_name = f"{project_id}.{dataset_name}.{table.table_name}" - logger.warning( - f"Error while processing view {view_name}", - exc_info=True, - ) - report.report_warning( - "metadata-extraction", - f"Failed to get view {view_name}: {e}", + report.warning( + title="Failed to process view", + message="Error encountered while processing view", + context=view_name, + exc=e, ) + self.report.num_get_views_for_dataset_api_requests += 1 + self.report.get_views_for_dataset_sec += current_timer.elapsed_seconds() @staticmethod def _make_bigquery_view(view: bigquery.Row) -> BigqueryView: @@ -416,22 +406,18 @@ def get_policy_tags_for_column( ) yield policy_tag.display_name except Exception as e: - logger.warning( - f"Unexpected error when retrieving policy tag {policy_tag_name} for column {column_name} in table {table_name}: {e}", - exc_info=True, - ) - report.report_warning( - "metadata-extraction", - f"Failed to retrieve policy tag {policy_tag_name} for column {column_name} in table {table_name} due to unexpected error: {e}", + report.warning( + title="Failed to retrieve policy tag", + message="Unexpected error when retrieving policy tag for column", + context=f"policy tag {policy_tag_name} for column {column_name} in table {table_ref}", + exc=e, ) except Exception as e: - logger.error( - f"Unexpected error retrieving schema for table {table_name} in dataset {dataset_name}, project {project_id}: {e}", - exc_info=True, - ) - report.report_warning( - "metadata-extraction", - f"Failed to retrieve schema for table {table_name} in dataset {dataset_name}, project {project_id} due to unexpected error: {e}", + report.warning( + title="Failed to retrieve policy tag for table", + message="Unexpected error retrieving policy tag for table", + context=table_ref, + exc=e, ) def get_columns_for_dataset( @@ -445,7 +431,7 @@ def get_columns_for_dataset( rate_limiter: Optional[RateLimiter] = None, ) -> Optional[Dict[str, List[BigqueryColumn]]]: columns: Dict[str, List[BigqueryColumn]] = defaultdict(list) - with self.report.get_columns_for_dataset: + with PerfTimer() as timer: try: cur = self.get_query_result( ( @@ -461,89 +447,57 @@ def get_columns_for_dataset( ), ) except Exception as e: - logger.warning(f"Columns for dataset query failed with exception: {e}") - # Error - Information schema query returned too much data. - # Please repeat query with more selective predicates. + report.warning( + title="Failed to retrieve columns for dataset", + message="Query to get columns for dataset failed with exception", + context=f"{project_id}.{dataset_name}", + exc=e, + ) return None last_seen_table: str = "" for column in cur: - if ( - column_limit - and column.table_name in columns - and len(columns[column.table_name]) >= column_limit - ): - if last_seen_table != column.table_name: - logger.warning( - f"{project_id}.{dataset_name}.{column.table_name} contains more than {column_limit} columns, only processing {column_limit} columns" - ) - last_seen_table = column.table_name - else: - columns[column.table_name].append( - BigqueryColumn( - name=column.column_name, - ordinal_position=column.ordinal_position, - field_path=column.field_path, - is_nullable=column.is_nullable == "YES", - data_type=column.data_type, - comment=column.comment, - is_partition_column=column.is_partitioning_column == "YES", - cluster_column_position=column.clustering_ordinal_position, - policy_tags=( - list( - self.get_policy_tags_for_column( - project_id, - dataset_name, - column.table_name, - column.column_name, - report, - rate_limiter, + with timer.pause(): + if ( + column_limit + and column.table_name in columns + and len(columns[column.table_name]) >= column_limit + ): + if last_seen_table != column.table_name: + logger.warning( + f"{project_id}.{dataset_name}.{column.table_name} contains more than {column_limit} columns, only processing {column_limit} columns" + ) + last_seen_table = column.table_name + else: + columns[column.table_name].append( + BigqueryColumn( + name=column.column_name, + ordinal_position=column.ordinal_position, + field_path=column.field_path, + is_nullable=column.is_nullable == "YES", + data_type=column.data_type, + comment=column.comment, + is_partition_column=column.is_partitioning_column + == "YES", + cluster_column_position=column.clustering_ordinal_position, + policy_tags=( + list( + self.get_policy_tags_for_column( + project_id, + dataset_name, + column.table_name, + column.column_name, + report, + rate_limiter, + ) ) - ) - if extract_policy_tags_from_catalog - else [] - ), + if extract_policy_tags_from_catalog + else [] + ), + ) ) - ) - - return columns - - # This is not used anywhere - def get_columns_for_table( - self, - table_identifier: BigqueryTableIdentifier, - column_limit: Optional[int], - ) -> List[BigqueryColumn]: - cur = self.get_query_result( - BigqueryQuery.columns_for_table.format(table_identifier=table_identifier), - ) - - columns: List[BigqueryColumn] = [] - last_seen_table: str = "" - for column in cur: - if ( - column_limit - and column.table_name in columns - and len(columns[column.table_name]) >= column_limit - ): - if last_seen_table != column.table_name: - logger.warning( - f"{table_identifier.project_id}.{table_identifier.dataset}.{column.table_name} contains more than {column_limit} columns, only processing {column_limit} columns" - ) - else: - columns.append( - BigqueryColumn( - name=column.column_name, - ordinal_position=column.ordinal_position, - is_nullable=column.is_nullable == "YES", - field_path=column.field_path, - data_type=column.data_type, - comment=column.comment, - is_partition_column=column.is_partitioning_column == "YES", - cluster_column_position=column.clustering_ordinal_position, - ) - ) - last_seen_table = column.table_name + self.report.num_get_columns_for_dataset_api_requests += 1 + self.report.get_columns_for_dataset_sec += timer.elapsed_seconds() return columns @@ -554,7 +508,7 @@ def get_snapshots_for_dataset( has_data_read: bool, report: BigQueryV2Report, ) -> Iterator[BigqueryTableSnapshot]: - with self.report.get_snapshots_for_dataset as current_timer: + with PerfTimer() as current_timer: if has_data_read: # If profiling is enabled cur = self.get_query_result( @@ -575,14 +529,14 @@ def get_snapshots_for_dataset( yield BigQuerySchemaApi._make_bigquery_table_snapshot(table) except Exception as e: snapshot_name = f"{project_id}.{dataset_name}.{table.table_name}" - logger.warning( - f"Error while processing view {snapshot_name}", - exc_info=True, - ) report.report_warning( - "metadata-extraction", - f"Failed to get view {snapshot_name}: {e}", + title="Failed to process snapshot", + message="Error encountered while processing snapshot", + context=snapshot_name, + exc=e, ) + self.report.num_get_snapshots_for_dataset_api_requests += 1 + self.report.get_snapshots_for_dataset_sec += current_timer.elapsed_seconds() @staticmethod def _make_bigquery_table_snapshot(snapshot: bigquery.Row) -> BigqueryTableSnapshot: diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema_gen.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema_gen.py new file mode 100644 index 0000000000000..3ffcb225db1c2 --- /dev/null +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_schema_gen.py @@ -0,0 +1,1090 @@ +import logging +import re +from collections import defaultdict +from typing import Callable, Dict, Iterable, List, Optional, Set, Type, Union, cast + +from google.cloud.bigquery.table import TableListItem + +from datahub.configuration.pattern_utils import is_schema_allowed, is_tag_allowed +from datahub.emitter.mce_builder import ( + make_data_platform_urn, + make_dataplatform_instance_urn, + make_tag_urn, +) +from datahub.emitter.mcp import MetadataChangeProposalWrapper +from datahub.emitter.mcp_builder import BigQueryDatasetKey, ContainerKey, ProjectIdKey +from datahub.ingestion.api.workunit import MetadataWorkUnit +from datahub.ingestion.glossary.classification_mixin import ( + SAMPLE_SIZE_MULTIPLIER, + ClassificationHandler, + classification_workunit_processor, +) +from datahub.ingestion.source.bigquery_v2.bigquery_audit import ( + BigqueryTableIdentifier, + BigQueryTableRef, +) +from datahub.ingestion.source.bigquery_v2.bigquery_config import BigQueryV2Config +from datahub.ingestion.source.bigquery_v2.bigquery_data_reader import BigQueryDataReader +from datahub.ingestion.source.bigquery_v2.bigquery_helper import ( + unquote_and_decode_unicode_escape_seq, +) +from datahub.ingestion.source.bigquery_v2.bigquery_report import BigQueryV2Report +from datahub.ingestion.source.bigquery_v2.bigquery_schema import ( + BigqueryColumn, + BigqueryDataset, + BigqueryProject, + BigQuerySchemaApi, + BigqueryTable, + BigqueryTableSnapshot, + BigqueryView, +) +from datahub.ingestion.source.bigquery_v2.common import ( + BQ_EXTERNAL_DATASET_URL_TEMPLATE, + BQ_EXTERNAL_TABLE_URL_TEMPLATE, +) +from datahub.ingestion.source.bigquery_v2.profiler import BigqueryProfiler +from datahub.ingestion.source.common.subtypes import ( + DatasetContainerSubTypes, + DatasetSubTypes, +) +from datahub.ingestion.source.sql.sql_utils import ( + add_table_to_schema_container, + gen_database_container, + gen_schema_container, + get_domain_wu, +) +from datahub.ingestion.source_report.ingestion_stage import ( + METADATA_EXTRACTION, + PROFILING, +) +from datahub.metadata.com.linkedin.pegasus2avro.common import ( + Status, + SubTypes, + TimeStamp, +) +from datahub.metadata.com.linkedin.pegasus2avro.dataset import ( + DatasetProperties, + ViewProperties, +) +from datahub.metadata.com.linkedin.pegasus2avro.schema import ( + ArrayType, + BooleanType, + BytesType, + DateType, + MySqlDDL, + NullType, + NumberType, + RecordType, + SchemaField, + SchemaFieldDataType, + SchemaMetadata, + StringType, + TimeType, +) +from datahub.metadata.schema_classes import ( + DataPlatformInstanceClass, + GlobalTagsClass, + TagAssociationClass, +) +from datahub.sql_parsing.schema_resolver import SchemaResolver +from datahub.utilities.file_backed_collections import FileBackedDict +from datahub.utilities.hive_schema_to_avro import ( + HiveColumnToAvroConverter, + get_schema_fields_for_hive_column, +) +from datahub.utilities.mapping import Constants +from datahub.utilities.perf_timer import PerfTimer +from datahub.utilities.ratelimiter import RateLimiter +from datahub.utilities.registries.domain_registry import DomainRegistry +from datahub.utilities.threaded_iterator_executor import ThreadedIteratorExecutor + +logger: logging.Logger = logging.getLogger(__name__) +# Handle table snapshots +# See https://cloud.google.com/bigquery/docs/table-snapshots-intro. +SNAPSHOT_TABLE_REGEX = re.compile(r"^(.+)@(\d{13})$") +CLUSTERING_COLUMN_TAG = "CLUSTERING_COLUMN" + + +class BigQuerySchemaGenerator: + # https://cloud.google.com/bigquery/docs/reference/standard-sql/data-types + # Note: We use the hive schema parser to parse nested BigQuery types. We also have + # some extra type mappings in that file. + BIGQUERY_FIELD_TYPE_MAPPINGS: Dict[ + str, + Type[ + Union[ + ArrayType, + BytesType, + BooleanType, + NumberType, + RecordType, + StringType, + TimeType, + DateType, + NullType, + ] + ], + ] = { + "BYTES": BytesType, + "BOOL": BooleanType, + "INT": NumberType, + "INT64": NumberType, + "SMALLINT": NumberType, + "INTEGER": NumberType, + "BIGINT": NumberType, + "TINYINT": NumberType, + "BYTEINT": NumberType, + "STRING": StringType, + "TIME": TimeType, + "TIMESTAMP": TimeType, + "DATE": DateType, + "DATETIME": TimeType, + "GEOGRAPHY": NullType, + "JSON": RecordType, + "INTERVAL": NullType, + "ARRAY": ArrayType, + "STRUCT": RecordType, + } + + def __init__( + self, + config: BigQueryV2Config, + report: BigQueryV2Report, + bigquery_data_dictionary: BigQuerySchemaApi, + domain_registry: Optional[DomainRegistry], + sql_parser_schema_resolver: SchemaResolver, + profiler: BigqueryProfiler, + dataset_urn_builder: Callable[[str, str, str], str], + ): + self.config = config + self.report = report + self.bigquery_data_dictionary = bigquery_data_dictionary + self.domain_registry = domain_registry + self.sql_parser_schema_resolver = sql_parser_schema_resolver + self.profiler = profiler + self.gen_dataset_urn = dataset_urn_builder + self.platform: str = "bigquery" + + self.classification_handler = ClassificationHandler(self.config, self.report) + self.data_reader: Optional[BigQueryDataReader] = None + if self.classification_handler.is_classification_enabled(): + self.data_reader = BigQueryDataReader.create( + self.config.get_bigquery_client() + ) + + # Global store of table identifiers for lineage filtering + self.table_refs: Set[str] = set() + + # Maps project -> view_ref, so we can find all views in a project + self.view_refs_by_project: Dict[str, Set[str]] = defaultdict(set) + # Maps project -> snapshot_ref, so we can find all snapshots in a project + self.snapshot_refs_by_project: Dict[str, Set[str]] = defaultdict(set) + # Maps view ref -> actual sql + self.view_definitions: FileBackedDict[str] = FileBackedDict() + # Maps snapshot ref -> Snapshot + self.snapshots_by_ref: FileBackedDict[BigqueryTableSnapshot] = FileBackedDict() + + @property + def store_table_refs(self): + return self.config.include_table_lineage or self.config.include_usage_statistics + + def get_project_workunits( + self, project: BigqueryProject + ) -> Iterable[MetadataWorkUnit]: + self.report.set_ingestion_stage(project.id, METADATA_EXTRACTION) + logger.info(f"Processing project: {project.id}") + yield from self._process_project(project) + + def get_dataplatform_instance_aspect( + self, dataset_urn: str, project_id: str + ) -> MetadataWorkUnit: + aspect = DataPlatformInstanceClass( + platform=make_data_platform_urn(self.platform), + instance=( + make_dataplatform_instance_urn(self.platform, project_id) + if self.config.include_data_platform_instance + else None + ), + ) + return MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=aspect + ).as_workunit() + + def gen_dataset_key(self, db_name: str, schema: str) -> ContainerKey: + return BigQueryDatasetKey( + project_id=db_name, + dataset_id=schema, + platform=self.platform, + env=self.config.env, + backcompat_env_as_instance=True, + ) + + def gen_project_id_key(self, database: str) -> ContainerKey: + return ProjectIdKey( + project_id=database, + platform=self.platform, + env=self.config.env, + backcompat_env_as_instance=True, + ) + + def gen_project_id_containers(self, database: str) -> Iterable[MetadataWorkUnit]: + database_container_key = self.gen_project_id_key(database) + + yield from gen_database_container( + database=database, + name=database, + sub_types=[DatasetContainerSubTypes.BIGQUERY_PROJECT], + domain_registry=self.domain_registry, + domain_config=self.config.domain, + database_container_key=database_container_key, + ) + + def gen_dataset_containers( + self, dataset: str, project_id: str, tags: Optional[Dict[str, str]] = None + ) -> Iterable[MetadataWorkUnit]: + schema_container_key = self.gen_dataset_key(project_id, dataset) + + tags_joined: Optional[List[str]] = None + if tags and self.config.capture_dataset_label_as_tag: + tags_joined = [ + f"{k}:{v}" + for k, v in tags.items() + if is_tag_allowed(self.config.capture_dataset_label_as_tag, k) + ] + + database_container_key = self.gen_project_id_key(database=project_id) + + yield from gen_schema_container( + database=project_id, + schema=dataset, + sub_types=[DatasetContainerSubTypes.BIGQUERY_DATASET], + domain_registry=self.domain_registry, + domain_config=self.config.domain, + schema_container_key=schema_container_key, + database_container_key=database_container_key, + external_url=( + BQ_EXTERNAL_DATASET_URL_TEMPLATE.format( + project=project_id, dataset=dataset + ) + if self.config.include_external_url + else None + ), + tags=tags_joined, + ) + + def _process_project( + self, bigquery_project: BigqueryProject + ) -> Iterable[MetadataWorkUnit]: + db_tables: Dict[str, List[BigqueryTable]] = {} + + project_id = bigquery_project.id + try: + bigquery_project.datasets = ( + self.bigquery_data_dictionary.get_datasets_for_project_id(project_id) + ) + except Exception as e: + + if ( + self.config.project_id or self.config.project_ids + ) and "not enabled BigQuery." in str(e): + action_mesage = ( + "The project has not enabled BigQuery API. " + "Did you mistype project id in recipe ?" + ) + else: + action_mesage = ( + "Does your service account have `bigquery.datasets.get` permission ? " + "Assign predefined role `roles/bigquery.metadataViewer` to your service account." + ) + + self.report.failure( + title="Unable to get datasets for project", + message=action_mesage, + context=project_id, + exc=e, + ) + return None + + if len(bigquery_project.datasets) == 0: + action_message = ( + "Either there are no datasets in this project or missing `bigquery.datasets.get` permission. " + "You can assign predefined roles/bigquery.metadataViewer role to your service account." + ) + if self.config.exclude_empty_projects: + self.report.report_dropped(project_id) + logger.info( + f"Excluded project '{project_id}' since no datasets were found. {action_message}" + ) + else: + yield from self.gen_project_id_containers(project_id) + self.report.warning( + title="No datasets found in project", + message=action_message, + context=project_id, + ) + return + + yield from self.gen_project_id_containers(project_id) + + self.report.num_project_datasets_to_scan[project_id] = len( + bigquery_project.datasets + ) + yield from self._process_project_datasets(bigquery_project, db_tables) + + if self.config.is_profiling_enabled(): + logger.info(f"Starting profiling project {project_id}") + self.report.set_ingestion_stage(project_id, PROFILING) + yield from self.profiler.get_workunits( + project_id=project_id, + tables=db_tables, + ) + + def _process_project_datasets( + self, + bigquery_project: BigqueryProject, + db_tables: Dict[str, List[BigqueryTable]], + ) -> Iterable[MetadataWorkUnit]: + + db_views: Dict[str, List[BigqueryView]] = {} + db_snapshots: Dict[str, List[BigqueryTableSnapshot]] = {} + project_id = bigquery_project.id + + def _process_schema_worker( + bigquery_dataset: BigqueryDataset, + ) -> Iterable[MetadataWorkUnit]: + if not is_schema_allowed( + self.config.dataset_pattern, + bigquery_dataset.name, + project_id, + self.config.match_fully_qualified_names, + ): + self.report.report_dropped(f"{bigquery_dataset.name}.*") + return + try: + # db_tables, db_views, and db_snapshots are populated in the this method + for wu in self._process_schema( + project_id, bigquery_dataset, db_tables, db_views, db_snapshots + ): + yield wu + except Exception as e: + if self.config.is_profiling_enabled(): + action_mesage = "Does your service account has bigquery.tables.list, bigquery.routines.get, bigquery.routines.list permission, bigquery.tables.getData permission?" + else: + action_mesage = "Does your service account has bigquery.tables.list, bigquery.routines.get, bigquery.routines.list permission?" + + self.report.failure( + title="Unable to get tables for dataset", + message=action_mesage, + context=f"{project_id}.{bigquery_dataset.name}", + exc=e, + ) + + for wu in ThreadedIteratorExecutor.process( + worker_func=_process_schema_worker, + args_list=[(bq_dataset,) for bq_dataset in bigquery_project.datasets], + max_workers=self.config.max_threads_dataset_parallelism, + ): + yield wu + + def _process_schema( + self, + project_id: str, + bigquery_dataset: BigqueryDataset, + db_tables: Dict[str, List[BigqueryTable]], + db_views: Dict[str, List[BigqueryView]], + db_snapshots: Dict[str, List[BigqueryTableSnapshot]], + ) -> Iterable[MetadataWorkUnit]: + dataset_name = bigquery_dataset.name + + yield from self.gen_dataset_containers( + dataset_name, project_id, bigquery_dataset.labels + ) + + columns = None + + rate_limiter: Optional[RateLimiter] = None + if self.config.rate_limit: + rate_limiter = RateLimiter( + max_calls=self.config.requests_per_min, period=60 + ) + + if ( + self.config.include_tables + or self.config.include_views + or self.config.include_table_snapshots + ): + columns = self.bigquery_data_dictionary.get_columns_for_dataset( + project_id=project_id, + dataset_name=dataset_name, + column_limit=self.config.column_limit, + run_optimized_column_query=self.config.run_optimized_column_query, + extract_policy_tags_from_catalog=self.config.extract_policy_tags_from_catalog, + report=self.report, + rate_limiter=rate_limiter, + ) + + if self.config.include_tables: + db_tables[dataset_name] = list( + self.get_tables_for_dataset(project_id, dataset_name) + ) + + for table in db_tables[dataset_name]: + table_columns = columns.get(table.name, []) if columns else [] + table_wu_generator = self._process_table( + table=table, + columns=table_columns, + project_id=project_id, + dataset_name=dataset_name, + ) + yield from classification_workunit_processor( + table_wu_generator, + self.classification_handler, + self.data_reader, + [project_id, dataset_name, table.name], + data_reader_kwargs=dict( + sample_size_percent=( + self.config.classification.sample_size + * SAMPLE_SIZE_MULTIPLIER + / table.rows_count + if table.rows_count + else None + ) + ), + ) + elif self.store_table_refs: + # Need table_refs to calculate lineage and usage + for table_item in self.bigquery_data_dictionary.list_tables( + dataset_name, project_id + ): + identifier = BigqueryTableIdentifier( + project_id=project_id, + dataset=dataset_name, + table=table_item.table_id, + ) + if not self.config.table_pattern.allowed(identifier.raw_table_name()): + self.report.report_dropped(identifier.raw_table_name()) + continue + try: + self.table_refs.add( + str(BigQueryTableRef(identifier).get_sanitized_table_ref()) + ) + except Exception as e: + logger.warning( + f"Could not create table ref for {table_item.path}: {e}" + ) + + if self.config.include_views: + db_views[dataset_name] = list( + self.bigquery_data_dictionary.get_views_for_dataset( + project_id, + dataset_name, + self.config.is_profiling_enabled(), + self.report, + ) + ) + + for view in db_views[dataset_name]: + view_columns = columns.get(view.name, []) if columns else [] + yield from self._process_view( + view=view, + columns=view_columns, + project_id=project_id, + dataset_name=dataset_name, + ) + + if self.config.include_table_snapshots: + db_snapshots[dataset_name] = list( + self.bigquery_data_dictionary.get_snapshots_for_dataset( + project_id, + dataset_name, + self.config.is_profiling_enabled(), + self.report, + ) + ) + + for snapshot in db_snapshots[dataset_name]: + snapshot_columns = columns.get(snapshot.name, []) if columns else [] + yield from self._process_snapshot( + snapshot=snapshot, + columns=snapshot_columns, + project_id=project_id, + dataset_name=dataset_name, + ) + + # This method is used to generate the ignore list for datatypes the profiler doesn't support we have to do it here + # because the profiler doesn't have access to columns + def generate_profile_ignore_list(self, columns: List[BigqueryColumn]) -> List[str]: + ignore_list: List[str] = [] + for column in columns: + if not column.data_type or any( + word in column.data_type.lower() + for word in ["array", "struct", "geography", "json"] + ): + ignore_list.append(column.field_path) + return ignore_list + + def _process_table( + self, + table: BigqueryTable, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + table_identifier = BigqueryTableIdentifier(project_id, dataset_name, table.name) + + self.report.report_entity_scanned(table_identifier.raw_table_name()) + + if not self.config.table_pattern.allowed(table_identifier.raw_table_name()): + self.report.report_dropped(table_identifier.raw_table_name()) + return + + if self.store_table_refs: + self.table_refs.add( + str(BigQueryTableRef(table_identifier).get_sanitized_table_ref()) + ) + table.column_count = len(columns) + + # We only collect profile ignore list if profiling is enabled and profile_table_level_only is false + if ( + self.config.is_profiling_enabled() + and not self.config.profiling.profile_table_level_only + ): + table.columns_ignore_from_profiling = self.generate_profile_ignore_list( + columns + ) + + if not table.column_count: + logger.warning( + f"Table doesn't have any column or unable to get columns for table: {table_identifier}" + ) + + # If table has time partitioning, set the data type of the partitioning field + if table.partition_info: + table.partition_info.column = next( + ( + column + for column in columns + if column.name == table.partition_info.field + ), + None, + ) + yield from self.gen_table_dataset_workunits( + table, columns, project_id, dataset_name + ) + + def _process_view( + self, + view: BigqueryView, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + table_identifier = BigqueryTableIdentifier(project_id, dataset_name, view.name) + + self.report.report_entity_scanned(table_identifier.raw_table_name(), "view") + + if not self.config.view_pattern.allowed(table_identifier.raw_table_name()): + self.report.report_dropped(table_identifier.raw_table_name()) + return + + if self.store_table_refs: + table_ref = str( + BigQueryTableRef(table_identifier).get_sanitized_table_ref() + ) + self.table_refs.add(table_ref) + if self.config.lineage_parse_view_ddl and view.view_definition: + self.view_refs_by_project[project_id].add(table_ref) + self.view_definitions[table_ref] = view.view_definition + + view.column_count = len(columns) + if not view.column_count: + logger.warning( + f"View doesn't have any column or unable to get columns for view: {table_identifier}" + ) + + yield from self.gen_view_dataset_workunits( + table=view, + columns=columns, + project_id=project_id, + dataset_name=dataset_name, + ) + + def _process_snapshot( + self, + snapshot: BigqueryTableSnapshot, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + table_identifier = BigqueryTableIdentifier( + project_id, dataset_name, snapshot.name + ) + + self.report.snapshots_scanned += 1 + + if not self.config.table_snapshot_pattern.allowed( + table_identifier.raw_table_name() + ): + self.report.report_dropped(table_identifier.raw_table_name()) + return + + snapshot.columns = columns + snapshot.column_count = len(columns) + if not snapshot.column_count: + logger.warning( + f"Snapshot doesn't have any column or unable to get columns for snapshot: {table_identifier}" + ) + + if self.store_table_refs: + table_ref = str( + BigQueryTableRef(table_identifier).get_sanitized_table_ref() + ) + self.table_refs.add(table_ref) + if snapshot.base_table_identifier: + self.snapshot_refs_by_project[project_id].add(table_ref) + self.snapshots_by_ref[table_ref] = snapshot + + yield from self.gen_snapshot_dataset_workunits( + table=snapshot, + columns=columns, + project_id=project_id, + dataset_name=dataset_name, + ) + + def gen_table_dataset_workunits( + self, + table: BigqueryTable, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + custom_properties: Dict[str, str] = {} + if table.expires: + custom_properties["expiration_date"] = str(table.expires) + + if table.partition_info: + custom_properties["partition_info"] = str(table.partition_info) + + if table.size_in_bytes: + custom_properties["size_in_bytes"] = str(table.size_in_bytes) + + if table.active_billable_bytes: + custom_properties["billable_bytes_active"] = str( + table.active_billable_bytes + ) + + if table.long_term_billable_bytes: + custom_properties["billable_bytes_long_term"] = str( + table.long_term_billable_bytes + ) + + if table.max_partition_id: + custom_properties["number_of_partitions"] = str(table.num_partitions) + custom_properties["max_partition_id"] = str(table.max_partition_id) + custom_properties["is_partitioned"] = str(True) + + sub_types: List[str] = [DatasetSubTypes.TABLE] + if table.max_shard_id: + custom_properties["max_shard_id"] = str(table.max_shard_id) + custom_properties["is_sharded"] = str(True) + sub_types = ["sharded table"] + sub_types + + tags_to_add = None + if table.labels and self.config.capture_table_label_as_tag: + tags_to_add = [] + tags_to_add.extend( + [ + make_tag_urn(f"""{k}:{v}""") + for k, v in table.labels.items() + if is_tag_allowed(self.config.capture_table_label_as_tag, k) + ] + ) + + yield from self.gen_dataset_workunits( + table=table, + columns=columns, + project_id=project_id, + dataset_name=dataset_name, + sub_types=sub_types, + tags_to_add=tags_to_add, + custom_properties=custom_properties, + ) + + def gen_view_dataset_workunits( + self, + table: BigqueryView, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + tags_to_add = None + if table.labels and self.config.capture_view_label_as_tag: + tags_to_add = [ + make_tag_urn(f"{k}:{v}") + for k, v in table.labels.items() + if is_tag_allowed(self.config.capture_view_label_as_tag, k) + ] + yield from self.gen_dataset_workunits( + table=table, + columns=columns, + project_id=project_id, + dataset_name=dataset_name, + tags_to_add=tags_to_add, + sub_types=[DatasetSubTypes.VIEW], + ) + + view = cast(BigqueryView, table) + view_definition_string = view.view_definition + view_properties_aspect = ViewProperties( + materialized=view.materialized, + viewLanguage="SQL", + viewLogic=view_definition_string or "", + ) + yield MetadataChangeProposalWrapper( + entityUrn=self.gen_dataset_urn(project_id, dataset_name, table.name), + aspect=view_properties_aspect, + ).as_workunit() + + def gen_snapshot_dataset_workunits( + self, + table: BigqueryTableSnapshot, + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + ) -> Iterable[MetadataWorkUnit]: + custom_properties: Dict[str, str] = {} + if table.ddl: + custom_properties["snapshot_ddl"] = table.ddl + if table.snapshot_time: + custom_properties["snapshot_time"] = str(table.snapshot_time) + if table.size_in_bytes: + custom_properties["size_in_bytes"] = str(table.size_in_bytes) + if table.rows_count: + custom_properties["rows_count"] = str(table.rows_count) + yield from self.gen_dataset_workunits( + table=table, + columns=columns, + project_id=project_id, + dataset_name=dataset_name, + sub_types=[DatasetSubTypes.BIGQUERY_TABLE_SNAPSHOT], + custom_properties=custom_properties, + ) + + def gen_dataset_workunits( + self, + table: Union[BigqueryTable, BigqueryView, BigqueryTableSnapshot], + columns: List[BigqueryColumn], + project_id: str, + dataset_name: str, + sub_types: List[str], + tags_to_add: Optional[List[str]] = None, + custom_properties: Optional[Dict[str, str]] = None, + ) -> Iterable[MetadataWorkUnit]: + dataset_urn = self.gen_dataset_urn(project_id, dataset_name, table.name) + + status = Status(removed=False) + yield MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=status + ).as_workunit() + + datahub_dataset_name = BigqueryTableIdentifier( + project_id, dataset_name, table.name + ) + + yield self.gen_schema_metadata( + dataset_urn, table, columns, datahub_dataset_name + ) + + dataset_properties = DatasetProperties( + name=datahub_dataset_name.get_table_display_name(), + description=( + unquote_and_decode_unicode_escape_seq(table.comment) + if table.comment + else "" + ), + qualifiedName=str(datahub_dataset_name), + created=( + TimeStamp(time=int(table.created.timestamp() * 1000)) + if table.created is not None + else None + ), + lastModified=( + TimeStamp(time=int(table.last_altered.timestamp() * 1000)) + if table.last_altered is not None + else None + ), + externalUrl=( + BQ_EXTERNAL_TABLE_URL_TEMPLATE.format( + project=project_id, dataset=dataset_name, table=table.name + ) + if self.config.include_external_url + else None + ), + ) + if custom_properties: + dataset_properties.customProperties.update(custom_properties) + + yield MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=dataset_properties + ).as_workunit() + + if tags_to_add: + yield self.gen_tags_aspect_workunit(dataset_urn, tags_to_add) + + yield from add_table_to_schema_container( + dataset_urn=dataset_urn, + parent_container_key=self.gen_dataset_key(project_id, dataset_name), + ) + yield self.get_dataplatform_instance_aspect( + dataset_urn=dataset_urn, project_id=project_id + ) + + subTypes = SubTypes(typeNames=sub_types) + yield MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=subTypes + ).as_workunit() + + if self.domain_registry: + yield from get_domain_wu( + dataset_name=str(datahub_dataset_name), + entity_urn=dataset_urn, + domain_registry=self.domain_registry, + domain_config=self.config.domain, + ) + + def gen_tags_aspect_workunit( + self, dataset_urn: str, tags_to_add: List[str] + ) -> MetadataWorkUnit: + tags = GlobalTagsClass( + tags=[TagAssociationClass(tag_to_add) for tag_to_add in tags_to_add] + ) + return MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=tags + ).as_workunit() + + def gen_schema_fields(self, columns: List[BigqueryColumn]) -> List[SchemaField]: + schema_fields: List[SchemaField] = [] + + # Below line affects HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR in global scope + # TODO: Refractor this such that + # converter = HiveColumnToAvroConverter(struct_type_separator=" "); + # converter.get_schema_fields_for_hive_column(...) + original_struct_type_separator = ( + HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR + ) + HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR = " " + _COMPLEX_TYPE = re.compile("^(struct|array)") + last_id = -1 + for col in columns: + # if col.data_type is empty that means this column is part of a complex type + if col.data_type is None or _COMPLEX_TYPE.match(col.data_type.lower()): + # If the we have seen the ordinal position that most probably means we already processed this complex type + if last_id != col.ordinal_position: + schema_fields.extend( + get_schema_fields_for_hive_column( + col.name, col.data_type.lower(), description=col.comment + ) + ) + + # We have to add complex type comments to the correct level + if col.comment: + for idx, field in enumerate(schema_fields): + # Remove all the [version=2.0].[type=struct]. tags to get the field path + if ( + re.sub( + r"\[.*?\]\.", + repl="", + string=field.fieldPath.lower(), + count=0, + flags=re.MULTILINE, + ) + == col.field_path.lower() + ): + field.description = col.comment + schema_fields[idx] = field + break + else: + tags = [] + if col.is_partition_column: + tags.append( + TagAssociationClass(make_tag_urn(Constants.TAG_PARTITION_KEY)) + ) + + if col.cluster_column_position is not None: + tags.append( + TagAssociationClass( + make_tag_urn( + f"{CLUSTERING_COLUMN_TAG}_{col.cluster_column_position}" + ) + ) + ) + + if col.policy_tags: + for policy_tag in col.policy_tags: + tags.append(TagAssociationClass(make_tag_urn(policy_tag))) + field = SchemaField( + fieldPath=col.name, + type=SchemaFieldDataType( + self.BIGQUERY_FIELD_TYPE_MAPPINGS.get(col.data_type, NullType)() + ), + nativeDataType=col.data_type, + description=col.comment, + nullable=col.is_nullable, + globalTags=GlobalTagsClass(tags=tags), + ) + schema_fields.append(field) + last_id = col.ordinal_position + HiveColumnToAvroConverter._STRUCT_TYPE_SEPARATOR = ( + original_struct_type_separator + ) + return schema_fields + + def gen_schema_metadata( + self, + dataset_urn: str, + table: Union[BigqueryTable, BigqueryView, BigqueryTableSnapshot], + columns: List[BigqueryColumn], + dataset_name: BigqueryTableIdentifier, + ) -> MetadataWorkUnit: + schema_metadata = SchemaMetadata( + schemaName=str(dataset_name), + platform=make_data_platform_urn(self.platform), + version=0, + hash="", + platformSchema=MySqlDDL(tableSchema=""), + # fields=[], + fields=self.gen_schema_fields(columns), + ) + + if self.config.lineage_parse_view_ddl or self.config.lineage_use_sql_parser: + self.sql_parser_schema_resolver.add_schema_metadata( + dataset_urn, schema_metadata + ) + + return MetadataChangeProposalWrapper( + entityUrn=dataset_urn, aspect=schema_metadata + ).as_workunit() + + def get_tables_for_dataset( + self, + project_id: str, + dataset_name: str, + ) -> Iterable[BigqueryTable]: + # In bigquery there is no way to query all tables in a Project id + with PerfTimer() as timer: + # Partitions view throw exception if we try to query partition info for too many tables + # so we have to limit the number of tables we query partition info. + # The conn.list_tables returns table infos that information_schema doesn't contain and this + # way we can merge that info with the queried one. + # https://cloud.google.com/bigquery/docs/information-schema-partitions + max_batch_size: int = ( + self.config.number_of_datasets_process_in_batch + if not self.config.is_profiling_enabled() + else self.config.number_of_datasets_process_in_batch_if_profiling_enabled + ) + + # We get the list of tables in the dataset to get core table properties and to be able to process the tables in batches + # We collect only the latest shards from sharded tables (tables with _YYYYMMDD suffix) and ignore temporary tables + table_items = self.get_core_table_details( + dataset_name, project_id, self.config.temp_table_dataset_prefix + ) + + items_to_get: Dict[str, TableListItem] = {} + for table_item in table_items: + items_to_get[table_item] = table_items[table_item] + if len(items_to_get) % max_batch_size == 0: + yield from self.bigquery_data_dictionary.get_tables_for_dataset( + project_id, + dataset_name, + items_to_get, + with_data_read_permission=self.config.have_table_data_read_permission, + report=self.report, + ) + items_to_get.clear() + + if items_to_get: + yield from self.bigquery_data_dictionary.get_tables_for_dataset( + project_id, + dataset_name, + items_to_get, + with_data_read_permission=self.config.have_table_data_read_permission, + report=self.report, + ) + + self.report.metadata_extraction_sec[f"{project_id}.{dataset_name}"] = round( + timer.elapsed_seconds(), 2 + ) + + def get_core_table_details( + self, dataset_name: str, project_id: str, temp_table_dataset_prefix: str + ) -> Dict[str, TableListItem]: + table_items: Dict[str, TableListItem] = {} + # Dict to store sharded table and the last seen max shard id + sharded_tables: Dict[str, TableListItem] = {} + + for table in self.bigquery_data_dictionary.list_tables( + dataset_name, project_id + ): + table_identifier = BigqueryTableIdentifier( + project_id=project_id, + dataset=dataset_name, + table=table.table_id, + ) + + if table.table_type == "VIEW": + if ( + not self.config.include_views + or not self.config.view_pattern.allowed( + table_identifier.raw_table_name() + ) + ): + self.report.report_dropped(table_identifier.raw_table_name()) + continue + else: + if not self.config.table_pattern.allowed( + table_identifier.raw_table_name() + ): + self.report.report_dropped(table_identifier.raw_table_name()) + continue + + _, shard = BigqueryTableIdentifier.get_table_and_shard( + table_identifier.table + ) + table_name = table_identifier.get_table_name().split(".")[-1] + + # Sharded tables look like: table_20220120 + # For sharded tables we only process the latest shard and ignore the rest + # to find the latest shard we iterate over the list of tables and store the maximum shard id + # We only have one special case where the table name is a date `20220110` + # in this case we merge all these tables under dataset name as table name. + # For example some_dataset.20220110 will be turned to some_dataset.some_dataset + # It seems like there are some bigquery user who uses this non-standard way of sharding the tables. + if shard: + if table_name not in sharded_tables: + sharded_tables[table_name] = table + continue + + stored_table_identifier = BigqueryTableIdentifier( + project_id=project_id, + dataset=dataset_name, + table=sharded_tables[table_name].table_id, + ) + _, stored_shard = BigqueryTableIdentifier.get_table_and_shard( + stored_table_identifier.table + ) + # When table is none, we use dataset_name as table_name + assert stored_shard + if stored_shard < shard: + sharded_tables[table_name] = table + continue + elif str(table_identifier).startswith(temp_table_dataset_prefix): + logger.debug(f"Dropping temporary table {table_identifier.table}") + self.report.report_dropped(table_identifier.raw_table_name()) + continue + + table_items[table.table_id] = table + + # Adding maximum shards to the list of tables + table_items.update({value.table_id: value for value in sharded_tables.values()}) + + return table_items diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_test_connection.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_test_connection.py new file mode 100644 index 0000000000000..3aac78c154b2e --- /dev/null +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/bigquery_test_connection.py @@ -0,0 +1,178 @@ +import logging +from datetime import datetime, timedelta +from typing import Dict, List, Union + +from google.cloud import bigquery + +from datahub.ingestion.api.source import ( + CapabilityReport, + SourceCapability, + TestConnectionReport, +) +from datahub.ingestion.source.bigquery_v2.bigquery_config import BigQueryV2Config +from datahub.ingestion.source.bigquery_v2.bigquery_report import BigQueryV2Report +from datahub.ingestion.source.bigquery_v2.bigquery_schema import BigQuerySchemaApi +from datahub.ingestion.source.bigquery_v2.lineage import BigqueryLineageExtractor +from datahub.ingestion.source.bigquery_v2.usage import BigQueryUsageExtractor +from datahub.sql_parsing.schema_resolver import SchemaResolver + +logger: logging.Logger = logging.getLogger(__name__) + + +class BigQueryTestConnection: + @staticmethod + def test_connection(config_dict: dict) -> TestConnectionReport: + test_report = TestConnectionReport() + _report: Dict[Union[SourceCapability, str], CapabilityReport] = dict() + + try: + connection_conf = BigQueryV2Config.parse_obj_allow_extras(config_dict) + client: bigquery.Client = connection_conf.get_bigquery_client() + assert client + + test_report.basic_connectivity = BigQueryTestConnection.connectivity_test( + client + ) + + connection_conf.start_time = datetime.now() + connection_conf.end_time = datetime.now() + timedelta(minutes=1) + + report: BigQueryV2Report = BigQueryV2Report() + project_ids: List[str] = [] + projects = client.list_projects() + + for project in projects: + if connection_conf.project_id_pattern.allowed(project.project_id): + project_ids.append(project.project_id) + + metadata_read_capability = ( + BigQueryTestConnection.metadata_read_capability_test( + project_ids, connection_conf + ) + ) + if SourceCapability.SCHEMA_METADATA not in _report: + _report[SourceCapability.SCHEMA_METADATA] = metadata_read_capability + + if connection_conf.include_table_lineage: + lineage_capability = BigQueryTestConnection.lineage_capability_test( + connection_conf, project_ids, report + ) + if SourceCapability.LINEAGE_COARSE not in _report: + _report[SourceCapability.LINEAGE_COARSE] = lineage_capability + + if connection_conf.include_usage_statistics: + usage_capability = BigQueryTestConnection.usage_capability_test( + connection_conf, project_ids, report + ) + if SourceCapability.USAGE_STATS not in _report: + _report[SourceCapability.USAGE_STATS] = usage_capability + + test_report.capability_report = _report + return test_report + + except Exception as e: + test_report.basic_connectivity = CapabilityReport( + capable=False, failure_reason=f"{e}" + ) + return test_report + + @staticmethod + def connectivity_test(client: bigquery.Client) -> CapabilityReport: + ret = client.query("select 1") + if ret.error_result: + return CapabilityReport( + capable=False, failure_reason=f"{ret.error_result['message']}" + ) + else: + return CapabilityReport(capable=True) + + @staticmethod + def metadata_read_capability_test( + project_ids: List[str], config: BigQueryV2Config + ) -> CapabilityReport: + for project_id in project_ids: + try: + logger.info(f"Metadata read capability test for project {project_id}") + client: bigquery.Client = config.get_bigquery_client() + assert client + bigquery_data_dictionary = BigQuerySchemaApi( + BigQueryV2Report().schema_api_perf, client + ) + result = bigquery_data_dictionary.get_datasets_for_project_id( + project_id, 10 + ) + if len(result) == 0: + return CapabilityReport( + capable=False, + failure_reason=f"Dataset query returned empty dataset. It is either empty or no dataset in project {project_id}", + ) + tables = bigquery_data_dictionary.get_tables_for_dataset( + project_id=project_id, + dataset_name=result[0].name, + tables={}, + with_data_read_permission=config.have_table_data_read_permission, + report=BigQueryV2Report(), + ) + if len(list(tables)) == 0: + return CapabilityReport( + capable=False, + failure_reason=f"Tables query did not return any table. It is either empty or no tables in project {project_id}.{result[0].name}", + ) + + except Exception as e: + return CapabilityReport( + capable=False, + failure_reason=f"Dataset query failed with error: {e}", + ) + + return CapabilityReport(capable=True) + + @staticmethod + def lineage_capability_test( + connection_conf: BigQueryV2Config, + project_ids: List[str], + report: BigQueryV2Report, + ) -> CapabilityReport: + lineage_extractor = BigqueryLineageExtractor( + connection_conf, report, lambda ref: "" + ) + for project_id in project_ids: + try: + logger.info(f"Lineage capability test for project {project_id}") + lineage_extractor.test_capability(project_id) + except Exception as e: + return CapabilityReport( + capable=False, + failure_reason=f"Lineage capability test failed with: {e}", + ) + + return CapabilityReport(capable=True) + + @staticmethod + def usage_capability_test( + connection_conf: BigQueryV2Config, + project_ids: List[str], + report: BigQueryV2Report, + ) -> CapabilityReport: + usage_extractor = BigQueryUsageExtractor( + connection_conf, + report, + schema_resolver=SchemaResolver(platform="bigquery"), + dataset_urn_builder=lambda ref: "", + ) + for project_id in project_ids: + try: + logger.info(f"Usage capability test for project {project_id}") + failures_before_test = len(report.failures) + usage_extractor.test_capability(project_id) + if failures_before_test != len(report.failures): + return CapabilityReport( + capable=False, + failure_reason="Usage capability test failed. Check the logs for further info", + ) + except Exception as e: + return CapabilityReport( + capable=False, + failure_reason=f"Usage capability test failed with: {e} for project {project_id}", + ) + return CapabilityReport(capable=True) diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/lineage.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/lineage.py index c41207ec67f62..496bd64d3b4fe 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/lineage.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/lineage.py @@ -251,11 +251,6 @@ def get_time_window(self) -> Tuple[datetime, datetime]: else: return self.config.start_time, self.config.end_time - def error(self, log: logging.Logger, key: str, reason: str) -> None: - # TODO: Remove this method. - # Note that this downgrades the error to a warning. - self.report.warning(key, reason) - def _should_ingest_lineage(self) -> bool: if ( self.redundant_run_skip_handler @@ -265,9 +260,9 @@ def _should_ingest_lineage(self) -> bool: ) ): # Skip this run - self.report.report_warning( - "lineage-extraction", - "Skip this run as there was already a run for current ingestion window.", + self.report.warning( + title="Skipped redundant lineage extraction", + message="Skip this run as there was already a run for current ingestion window.", ) return False @@ -345,12 +340,12 @@ def generate_lineage( events, sql_parser_schema_resolver ) except Exception as e: - if project_id: - self.report.lineage_failed_extraction.append(project_id) - self.error( - logger, - "lineage", - f"{project_id}: {e}", + self.report.lineage_failed_extraction.append(project_id) + self.report.warning( + title="Failed to extract lineage", + message="Unexpected error encountered", + context=project_id, + exc=e, ) lineage = {} @@ -481,98 +476,88 @@ def lineage_via_catalog_lineage_api( # Regions to search for BigQuery tables: projects/{project_id}/locations/{region} enabled_regions: List[str] = ["US", "EU"] - try: - lineage_client: lineage_v1.LineageClient = lineage_v1.LineageClient() + lineage_client: lineage_v1.LineageClient = lineage_v1.LineageClient() + + data_dictionary = BigQuerySchemaApi( + self.report.schema_api_perf, self.config.get_bigquery_client() + ) - data_dictionary = BigQuerySchemaApi( - self.report.schema_api_perf, self.config.get_bigquery_client() + # Filtering datasets + datasets = list(data_dictionary.get_datasets_for_project_id(project_id)) + project_tables = [] + for dataset in datasets: + # Enables only tables where type is TABLE, VIEW or MATERIALIZED_VIEW (not EXTERNAL) + project_tables.extend( + [ + table + for table in data_dictionary.list_tables(dataset.name, project_id) + if table.table_type in ["TABLE", "VIEW", "MATERIALIZED_VIEW"] + ] ) - # Filtering datasets - datasets = list(data_dictionary.get_datasets_for_project_id(project_id)) - project_tables = [] - for dataset in datasets: - # Enables only tables where type is TABLE, VIEW or MATERIALIZED_VIEW (not EXTERNAL) - project_tables.extend( + lineage_map: Dict[str, Set[LineageEdge]] = {} + curr_date = datetime.now() + for project_table in project_tables: + # Convert project table to .. format + table = f"{project_table.project}.{project_table.dataset_id}.{project_table.table_id}" + + if not is_schema_allowed( + self.config.dataset_pattern, + schema_name=project_table.dataset_id, + db_name=project_table.project, + match_fully_qualified_schema_name=self.config.match_fully_qualified_names, + ) or not self.config.table_pattern.allowed(table): + self.report.num_skipped_lineage_entries_not_allowed[ + project_table.project + ] += 1 + continue + + logger.info("Creating lineage map for table %s", table) + upstreams = set() + downstream_table = lineage_v1.EntityReference() + # fully_qualified_name in format: "bigquery:.." + downstream_table.fully_qualified_name = f"bigquery:{table}" + # Searches in different regions + for region in enabled_regions: + location_request = lineage_v1.SearchLinksRequest( + target=downstream_table, + parent=f"projects/{project_id}/locations/{region.lower()}", + ) + response = lineage_client.search_links(request=location_request) + upstreams.update( [ - table - for table in data_dictionary.list_tables( - dataset.name, project_id + str(lineage.source.fully_qualified_name).replace( + "bigquery:", "" ) - if table.table_type in ["TABLE", "VIEW", "MATERIALIZED_VIEW"] + for lineage in response ] ) - lineage_map: Dict[str, Set[LineageEdge]] = {} - curr_date = datetime.now() - for project_table in project_tables: - # Convert project table to .. format - table = f"{project_table.project}.{project_table.dataset_id}.{project_table.table_id}" - - if not is_schema_allowed( - self.config.dataset_pattern, - schema_name=project_table.dataset_id, - db_name=project_table.project, - match_fully_qualified_schema_name=self.config.match_fully_qualified_names, - ) or not self.config.table_pattern.allowed(table): - self.report.num_skipped_lineage_entries_not_allowed[ - project_table.project - ] += 1 - continue - - logger.info("Creating lineage map for table %s", table) - upstreams = set() - downstream_table = lineage_v1.EntityReference() - # fully_qualified_name in format: "bigquery:.." - downstream_table.fully_qualified_name = f"bigquery:{table}" - # Searches in different regions - for region in enabled_regions: - location_request = lineage_v1.SearchLinksRequest( - target=downstream_table, - parent=f"projects/{project_id}/locations/{region.lower()}", - ) - response = lineage_client.search_links(request=location_request) - upstreams.update( - [ - str(lineage.source.fully_qualified_name).replace( - "bigquery:", "" - ) - for lineage in response - ] - ) - - # Downstream table identifier - destination_table_str = str( - BigQueryTableRef( - table_identifier=BigqueryTableIdentifier(*table.split(".")) - ) + # Downstream table identifier + destination_table_str = str( + BigQueryTableRef( + table_identifier=BigqueryTableIdentifier(*table.split(".")) ) + ) - # Only builds lineage map when the table has upstreams - logger.debug("Found %d upstreams for table %s", len(upstreams), table) - if upstreams: - lineage_map[destination_table_str] = { - LineageEdge( - table=str( - BigQueryTableRef( - table_identifier=BigqueryTableIdentifier.from_string_name( - source_table - ) + # Only builds lineage map when the table has upstreams + logger.debug("Found %d upstreams for table %s", len(upstreams), table) + if upstreams: + lineage_map[destination_table_str] = { + LineageEdge( + table=str( + BigQueryTableRef( + table_identifier=BigqueryTableIdentifier.from_string_name( + source_table ) - ), - column_mapping=frozenset(), - auditStamp=curr_date, - ) - for source_table in upstreams - } - return lineage_map - except Exception as e: - self.error( - logger, - "lineage-exported-catalog-lineage-api", - f"Error: {e}", - ) - raise e + ) + ), + column_mapping=frozenset(), + auditStamp=curr_date, + ) + for source_table in upstreams + } + return lineage_map def _get_parsed_audit_log_events(self, project_id: str) -> Iterable[QueryEvent]: # We adjust the filter values a bit, since we need to make sure that the join diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/profiler.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/profiler.py index 8c393d1e8a436..582c312f99098 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/profiler.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/profiler.py @@ -227,8 +227,9 @@ def get_profile_request( if partition is None and bq_table.partition_info: self.report.report_warning( - "profile skipped as partitioned table is empty or partition id or type was invalid", - profile_request.pretty_name, + title="Profile skipped for partitioned table", + message="profile skipped as partitioned table is empty or partition id or type was invalid", + context=profile_request.pretty_name, ) return None if ( diff --git a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/usage.py b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/usage.py index 1b95cbf505016..6824d630a2277 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/usage.py +++ b/metadata-ingestion/src/datahub/ingestion/source/bigquery_v2/usage.py @@ -358,9 +358,9 @@ def _should_ingest_usage(self) -> bool: ) ): # Skip this run - self.report.report_warning( - "usage-extraction", - "Skip this run as there was already a run for current ingestion window.", + self.report.warning( + title="Skipped redundant usage extraction", + message="Skip this run as there was already a run for current ingestion window.", ) return False @@ -410,8 +410,7 @@ def _get_workunits_internal( ) usage_state.report_disk_usage(self.report) except Exception as e: - logger.error("Error processing usage", exc_info=True) - self.report.report_warning("usage-ingestion", str(e)) + self.report.warning(message="Error processing usage", exc=e) self.report_status("usage-ingestion", False) def generate_read_events_from_query( @@ -477,10 +476,12 @@ def _ingest_events( ) except Exception as e: - logger.warning( - f"Unable to store usage event {audit_event}", exc_info=True + self.report.warning( + message="Unable to store usage event", + context=f"{audit_event}", + exc=e, ) - self._report_error("store-event", e) + logger.info(f"Total number of events aggregated = {num_aggregated}.") if self.report.num_view_query_events > 0: @@ -500,11 +501,11 @@ def _generate_operational_workunits( yield operational_wu self.report.num_operational_stats_workunits_emitted += 1 except Exception as e: - logger.warning( - f"Unable to generate operation workunit for event {audit_event}", - exc_info=True, + self.report.warning( + message="Unable to generate operation workunit", + context=f"{audit_event}", + exc=e, ) - self._report_error("operation-workunit", e) def _generate_usage_workunits( self, usage_state: BigQueryUsageState @@ -541,11 +542,11 @@ def _generate_usage_workunits( ) self.report.num_usage_workunits_emitted += 1 except Exception as e: - logger.warning( - f"Unable to generate usage workunit for bucket {entry.timestamp}, {entry.resource}", - exc_info=True, + self.report.warning( + message="Unable to generate usage statistics workunit", + context=f"{entry.timestamp}, {entry.resource}", + exc=e, ) - self._report_error("statistics-workunit", e) def _get_usage_events(self, projects: Iterable[str]) -> Iterable[AuditEvent]: if self.config.use_exported_bigquery_audit_metadata: @@ -559,12 +560,12 @@ def _get_usage_events(self, projects: Iterable[str]) -> Iterable[AuditEvent]: ) yield from self._get_parsed_bigquery_log_events(project_id) except Exception as e: - logger.error( - f"Error getting usage events for project {project_id}", - exc_info=True, - ) self.report.usage_failed_extraction.append(project_id) - self.report.report_warning(f"usage-extraction-{project_id}", str(e)) + self.report.warning( + message="Failed to get some or all usage events for project", + context=project_id, + exc=e, + ) self.report_status(f"usage-extraction-{project_id}", False) self.report.usage_extraction_sec[project_id] = round( @@ -898,12 +899,10 @@ def _get_parsed_bigquery_log_events( self.report.num_usage_parsed_log_entries[project_id] += 1 yield event except Exception as e: - logger.warning( - f"Unable to parse log entry `{entry}` for project {project_id}", - exc_info=True, - ) - self._report_error( - f"log-parse-{project_id}", e, group="usage-log-parse" + self.report.warning( + message="Unable to parse usage log entry", + context=f"`{entry}` for project {project_id}", + exc=e, ) def _generate_filter(self, corrected_start_time, corrected_end_time): @@ -946,13 +945,6 @@ def get_tables_from_query( return parsed_table_refs - def _report_error( - self, label: str, e: Exception, group: Optional[str] = None - ) -> None: - """Report an error that does not constitute a major failure.""" - self.report.usage_error_count[label] += 1 - self.report.report_warning(group or f"usage-{label}", str(e)) - def test_capability(self, project_id: str) -> None: for entry in self._get_parsed_bigquery_log_events(project_id, limit=1): logger.debug(f"Connection test got one {entry}") diff --git a/metadata-ingestion/src/datahub/ingestion/source/snowflake/snowflake_schema_gen.py b/metadata-ingestion/src/datahub/ingestion/source/snowflake/snowflake_schema_gen.py index e604ed96b8eb6..25442492b1eb4 100644 --- a/metadata-ingestion/src/datahub/ingestion/source/snowflake/snowflake_schema_gen.py +++ b/metadata-ingestion/src/datahub/ingestion/source/snowflake/snowflake_schema_gen.py @@ -1,7 +1,5 @@ -import concurrent.futures import itertools import logging -import queue from typing import Callable, Dict, Iterable, List, Optional, Union from datahub.configuration.pattern_utils import is_schema_allowed @@ -101,6 +99,7 @@ from datahub.metadata.com.linkedin.pegasus2avro.tag import TagProperties from datahub.sql_parsing.sql_parsing_aggregator import SqlParsingAggregator from datahub.utilities.registries.domain_registry import DomainRegistry +from datahub.utilities.threaded_iterator_executor import ThreadedIteratorExecutor logger = logging.getLogger(__name__) @@ -318,41 +317,22 @@ def _process_db_schemas( snowflake_db: SnowflakeDatabase, db_tables: Dict[str, List[SnowflakeTable]], ) -> Iterable[MetadataWorkUnit]: - q: "queue.Queue[MetadataWorkUnit]" = queue.Queue(maxsize=100) - - def _process_schema_worker(snowflake_schema: SnowflakeSchema) -> None: + def _process_schema_worker( + snowflake_schema: SnowflakeSchema, + ) -> Iterable[MetadataWorkUnit]: for wu in self._process_schema( snowflake_schema, snowflake_db.name, db_tables ): - q.put(wu) - - with concurrent.futures.ThreadPoolExecutor( - max_workers=SCHEMA_PARALLELISM - ) as executor: - futures = [] - for snowflake_schema in snowflake_db.schemas: - f = executor.submit(_process_schema_worker, snowflake_schema) - futures.append(f) - - # Read from the queue and yield the work units until all futures are done. - while True: - if not q.empty(): - while not q.empty(): - yield q.get_nowait() - else: - try: - yield q.get(timeout=0.2) - except queue.Empty: - pass - - # Filter out the done futures. - futures = [f for f in futures if not f.done()] - if not futures: - break - - # Yield the remaining work units. This theoretically should not happen, but adding it just in case. - while not q.empty(): - yield q.get_nowait() + yield wu + + for wu in ThreadedIteratorExecutor.process( + worker_func=_process_schema_worker, + args_list=[ + (snowflake_schema,) for snowflake_schema in snowflake_db.schemas + ], + max_workers=SCHEMA_PARALLELISM, + ): + yield wu def fetch_schemas_for_database( self, snowflake_db: SnowflakeDatabase, db_name: str diff --git a/metadata-ingestion/src/datahub/utilities/threaded_iterator_executor.py b/metadata-ingestion/src/datahub/utilities/threaded_iterator_executor.py new file mode 100644 index 0000000000000..216fa155035d3 --- /dev/null +++ b/metadata-ingestion/src/datahub/utilities/threaded_iterator_executor.py @@ -0,0 +1,52 @@ +import concurrent.futures +import contextlib +import queue +from typing import Any, Callable, Generator, Iterable, Tuple, TypeVar + +T = TypeVar("T") + + +class ThreadedIteratorExecutor: + """ + Executes worker functions of type `Callable[..., Iterable[T]]` in parallel threads, + yielding items of type `T` as they become available. + """ + + @classmethod + def process( + cls, + worker_func: Callable[..., Iterable[T]], + args_list: Iterable[Tuple[Any, ...]], + max_workers: int, + ) -> Generator[T, None, None]: + + out_q: queue.Queue[T] = queue.Queue() + + def _worker_wrapper( + worker_func: Callable[..., Iterable[T]], *args: Any + ) -> None: + for item in worker_func(*args): + out_q.put(item) + + with concurrent.futures.ThreadPoolExecutor(max_workers=max_workers) as executor: + futures = [] + for args in args_list: + future = executor.submit(_worker_wrapper, worker_func, *args) + futures.append(future) + # Read from the queue and yield the work units until all futures are done. + while True: + if not out_q.empty(): + while not out_q.empty(): + yield out_q.get_nowait() + else: + with contextlib.suppress(queue.Empty): + yield out_q.get(timeout=0.2) + + # Filter out the done futures. + futures = [f for f in futures if not f.done()] + if not futures: + break + + # Yield the remaining work units. This theoretically should not happen, but adding it just in case. + while not out_q.empty(): + yield out_q.get_nowait() diff --git a/metadata-ingestion/tests/integration/bigquery_v2/test_bigquery.py b/metadata-ingestion/tests/integration/bigquery_v2/test_bigquery.py index a24b6174eb925..762c73d2a55c6 100644 --- a/metadata-ingestion/tests/integration/bigquery_v2/test_bigquery.py +++ b/metadata-ingestion/tests/integration/bigquery_v2/test_bigquery.py @@ -11,7 +11,6 @@ DynamicTypedClassifierConfig, ) from datahub.ingestion.glossary.datahub_classifier import DataHubClassifierConfig -from datahub.ingestion.source.bigquery_v2.bigquery import BigqueryV2Source from datahub.ingestion.source.bigquery_v2.bigquery_data_reader import BigQueryDataReader from datahub.ingestion.source.bigquery_v2.bigquery_schema import ( BigqueryColumn, @@ -19,6 +18,9 @@ BigQuerySchemaApi, BigqueryTable, ) +from datahub.ingestion.source.bigquery_v2.bigquery_schema_gen import ( + BigQuerySchemaGenerator, +) from tests.test_helpers import mce_helpers from tests.test_helpers.state_helpers import run_and_get_pipeline @@ -39,7 +41,7 @@ def random_email(): @freeze_time(FROZEN_TIME) @patch.object(BigQuerySchemaApi, "get_tables_for_dataset") -@patch.object(BigqueryV2Source, "get_core_table_details") +@patch.object(BigQuerySchemaGenerator, "get_core_table_details") @patch.object(BigQuerySchemaApi, "get_datasets_for_project_id") @patch.object(BigQuerySchemaApi, "get_columns_for_dataset") @patch.object(BigQueryDataReader, "get_sample_data_for_table") diff --git a/metadata-ingestion/tests/unit/test_bigquery_source.py b/metadata-ingestion/tests/unit/test_bigquery_source.py index b58f35c0deef5..ea32db0ef2757 100644 --- a/metadata-ingestion/tests/unit/test_bigquery_source.py +++ b/metadata-ingestion/tests/unit/test_bigquery_source.py @@ -32,6 +32,9 @@ BigqueryTableSnapshot, BigqueryView, ) +from datahub.ingestion.source.bigquery_v2.bigquery_schema_gen import ( + BigQuerySchemaGenerator, +) from datahub.ingestion.source.bigquery_v2.lineage import ( LineageEdge, LineageEdgeColumnMapping, @@ -231,8 +234,9 @@ def test_get_dataplatform_instance_aspect_returns_project_id(get_bq_client_mock) config = BigQueryV2Config.parse_obj({"include_data_platform_instance": True}) source = BigqueryV2Source(config=config, ctx=PipelineContext(run_id="test")) + schema_gen = source.bq_schema_extractor - data_platform_instance = source.get_dataplatform_instance_aspect( + data_platform_instance = schema_gen.get_dataplatform_instance_aspect( "urn:li:test", project_id ) metadata = data_platform_instance.metadata @@ -246,8 +250,9 @@ def test_get_dataplatform_instance_aspect_returns_project_id(get_bq_client_mock) def test_get_dataplatform_instance_default_no_instance(get_bq_client_mock): config = BigQueryV2Config.parse_obj({}) source = BigqueryV2Source(config=config, ctx=PipelineContext(run_id="test")) + schema_gen = source.bq_schema_extractor - data_platform_instance = source.get_dataplatform_instance_aspect( + data_platform_instance = schema_gen.get_dataplatform_instance_aspect( "urn:li:test", "project_id" ) metadata = data_platform_instance.metadata @@ -395,8 +400,9 @@ def test_gen_table_dataset_workunits(get_bq_client_mock, bigquery_table): source: BigqueryV2Source = BigqueryV2Source( config=config, ctx=PipelineContext(run_id="test") ) + schema_gen = source.bq_schema_extractor - gen = source.gen_table_dataset_workunits( + gen = schema_gen.gen_table_dataset_workunits( bigquery_table, [], project_id, dataset_name ) mcp = cast(MetadataChangeProposalClass, next(iter(gen)).metadata) @@ -710,9 +716,10 @@ def test_table_processing_logic(get_bq_client_mock, data_dictionary_mock): data_dictionary_mock.get_tables_for_dataset.return_value = None source = BigqueryV2Source(config=config, ctx=PipelineContext(run_id="test")) + schema_gen = source.bq_schema_extractor _ = list( - source.get_tables_for_dataset( + schema_gen.get_tables_for_dataset( project_id="test-project", dataset_name="test-dataset" ) ) @@ -784,9 +791,10 @@ def test_table_processing_logic_date_named_tables( data_dictionary_mock.get_tables_for_dataset.return_value = None source = BigqueryV2Source(config=config, ctx=PipelineContext(run_id="test")) + schema_gen = source.bq_schema_extractor _ = list( - source.get_tables_for_dataset( + schema_gen.get_tables_for_dataset( project_id="test-project", dataset_name="test-dataset" ) ) @@ -882,7 +890,9 @@ def test_get_views_for_dataset( assert list(views) == [bigquery_view_1, bigquery_view_2] -@patch.object(BigqueryV2Source, "gen_dataset_workunits", lambda *args, **kwargs: []) +@patch.object( + BigQuerySchemaGenerator, "gen_dataset_workunits", lambda *args, **kwargs: [] +) @patch.object(BigQueryV2Config, "get_bigquery_client") def test_gen_view_dataset_workunits( get_bq_client_mock, bigquery_view_1, bigquery_view_2 @@ -897,8 +907,9 @@ def test_gen_view_dataset_workunits( source: BigqueryV2Source = BigqueryV2Source( config=config, ctx=PipelineContext(run_id="test") ) + schema_gen = source.bq_schema_extractor - gen = source.gen_view_dataset_workunits( + gen = schema_gen.gen_view_dataset_workunits( bigquery_view_1, [], project_id, dataset_name ) mcp = cast(MetadataChangeProposalClass, next(iter(gen)).metadata) @@ -908,7 +919,7 @@ def test_gen_view_dataset_workunits( viewLogic=bigquery_view_1.view_definition, ) - gen = source.gen_view_dataset_workunits( + gen = schema_gen.gen_view_dataset_workunits( bigquery_view_2, [], project_id, dataset_name ) mcp = cast(MetadataChangeProposalClass, next(iter(gen)).metadata) @@ -990,8 +1001,9 @@ def test_gen_snapshot_dataset_workunits(get_bq_client_mock, bigquery_snapshot): source: BigqueryV2Source = BigqueryV2Source( config=config, ctx=PipelineContext(run_id="test") ) + schema_gen = source.bq_schema_extractor - gen = source.gen_snapshot_dataset_workunits( + gen = schema_gen.gen_snapshot_dataset_workunits( bigquery_snapshot, [], project_id, dataset_name ) mcp = cast(MetadataChangeProposalWrapper, list(gen)[2].metadata) diff --git a/metadata-ingestion/tests/unit/utilities/test_threaded_iterator_executor.py b/metadata-ingestion/tests/unit/utilities/test_threaded_iterator_executor.py new file mode 100644 index 0000000000000..35c44c7b4a847 --- /dev/null +++ b/metadata-ingestion/tests/unit/utilities/test_threaded_iterator_executor.py @@ -0,0 +1,14 @@ +from datahub.utilities.threaded_iterator_executor import ThreadedIteratorExecutor + + +def test_threaded_iterator_executor(): + def table_of(i): + for j in range(1, 11): + yield f"{i}x{j}={i*j}" + + assert { + res + for res in ThreadedIteratorExecutor.process( + table_of, [(i,) for i in range(1, 30)], max_workers=2 + ) + } == {x for i in range(1, 30) for x in table_of(i)}