Source code for airflow.providers.amazon.aws.operators.appflow

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
from __future__ import annotations

from datetime import datetime, timedelta
from time import sleep
from typing import TYPE_CHECKING, cast

from airflow.compat.functools import cached_property
from airflow.exceptions import AirflowException
from airflow.models import BaseOperator
from airflow.operators.python import ShortCircuitOperator
from airflow.providers.amazon.aws.hooks.appflow import AppflowHook
from airflow.providers.amazon.aws.utils import datetime_to_epoch_ms

if TYPE_CHECKING:
    from mypy_boto3_appflow.type_defs import (
        DescribeFlowExecutionRecordsResponseTypeDef,
        ExecutionRecordTypeDef,
        TaskTypeDef,
    )

    from airflow.utils.context import Context

[docs]SUPPORTED_SOURCES = {"salesforce", "zendesk"}
[docs]MANDATORY_FILTER_DATE_MSG = "The filter_date argument is mandatory for {entity}!"
[docs]NOT_SUPPORTED_SOURCE_MSG = "Source {source} is not supported for {entity}!"
[docs]class AppflowBaseOperator(BaseOperator): """ Amazon Appflow Base Operator class (not supposed to be used directly in DAGs). :param source: The source name (Supported: salesforce, zendesk) :param flow_name: The flow name :param flow_update: A boolean to enable/disable a flow update before the run :param source_field: The field name to apply filters :param filter_date: The date value (or template) to be used in filters. :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """
[docs] ui_color = "#2bccbd"
[docs] UPDATE_PROPAGATION_TIME: int = 15
def __init__( self, source: str, flow_name: str, flow_update: bool, source_field: str | None = None, filter_date: str | None = None, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: super().__init__(**kwargs) if source not in SUPPORTED_SOURCES: raise ValueError(f"{source} is not a supported source (options: {SUPPORTED_SOURCES})!") self.filter_date = filter_date self.flow_name = flow_name self.source = source self.source_field = source_field self.poll_interval = poll_interval self.aws_conn_id = aws_conn_id self.region = region self.flow_update = flow_update self.wait_for_completion = wait_for_completion @cached_property
[docs] def hook(self) -> AppflowHook: """Create and return an AppflowHook.""" return AppflowHook(aws_conn_id=self.aws_conn_id, region_name=self.region)
[docs] def execute(self, context: Context) -> None: self.filter_date_parsed: datetime | None = ( datetime.fromisoformat(self.filter_date) if self.filter_date else None ) self.connector_type = self._get_connector_type() if self.flow_update: self._update_flow() # while schedule flows will pick up the update right away, on-demand flows might use out of date # info if triggered right after an update, so we need to wait a bit for the DB to be consistent. sleep(AppflowBaseOperator.UPDATE_PROPAGATION_TIME) self._run_flow(context)
def _get_connector_type(self) -> str: response = self.hook.conn.describe_flow(flowName=self.flow_name) connector_type = response["sourceFlowConfig"]["connectorType"] if self.source != connector_type.lower(): raise ValueError(f"Incompatible source ({self.source} and connector type ({connector_type})!") return connector_type def _update_flow(self) -> None: self.hook.update_flow_filter(flow_name=self.flow_name, filter_tasks=[], set_trigger_ondemand=True) def _run_flow(self, context) -> str: execution_id = self.hook.run_flow( flow_name=self.flow_name, poll_interval=self.poll_interval, wait_for_completion=self.wait_for_completion, ) task_instance = context["task_instance"] task_instance.xcom_push("execution_id", execution_id) return execution_id
[docs]class AppflowRunOperator(AppflowBaseOperator): """ Execute a Appflow run with filters as is. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRunOperator` :param source: The source name (Supported: salesforce, zendesk) :param flow_name: The flow name :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """ def __init__( self, source: str, flow_name: str, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: if source not in {"salesforce", "zendesk"}: raise ValueError(NOT_SUPPORTED_SOURCE_MSG.format(source=source, entity="AppflowRunOperator")) super().__init__( source=source, flow_name=flow_name, flow_update=False, source_field=None, filter_date=None, poll_interval=poll_interval, aws_conn_id=aws_conn_id, region=region, wait_for_completion=wait_for_completion, **kwargs,
)
[docs]class AppflowRunFullOperator(AppflowBaseOperator): """ Execute a Appflow full run removing any filter. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRunFullOperator` :param source: The source name (Supported: salesforce, zendesk) :param flow_name: The flow name :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """ def __init__( self, source: str, flow_name: str, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: if source not in {"salesforce", "zendesk"}: raise ValueError(NOT_SUPPORTED_SOURCE_MSG.format(source=source, entity="AppflowRunFullOperator")) super().__init__( source=source, flow_name=flow_name, flow_update=True, source_field=None, filter_date=None, poll_interval=poll_interval, aws_conn_id=aws_conn_id, region=region, wait_for_completion=wait_for_completion, **kwargs,
)
[docs]class AppflowRunBeforeOperator(AppflowBaseOperator): """ Execute a Appflow run after updating the filters to select only previous data. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRunBeforeOperator` :param source: The source name (Supported: salesforce) :param flow_name: The flow name :param source_field: The field name to apply filters :param filter_date: The date value (or template) to be used in filters. :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """
[docs] template_fields = ("filter_date",)
def __init__( self, source: str, flow_name: str, source_field: str, filter_date: str, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: if not filter_date: raise ValueError(MANDATORY_FILTER_DATE_MSG.format(entity="AppflowRunBeforeOperator")) if source != "salesforce": raise ValueError( NOT_SUPPORTED_SOURCE_MSG.format(source=source, entity="AppflowRunBeforeOperator") ) super().__init__( source=source, flow_name=flow_name, flow_update=True, source_field=source_field, filter_date=filter_date, poll_interval=poll_interval, aws_conn_id=aws_conn_id, region=region, wait_for_completion=wait_for_completion, **kwargs, ) def _update_flow(self) -> None: if not self.filter_date_parsed: raise ValueError(f"Invalid filter_date argument parser value: {self.filter_date_parsed}") if not self.source_field: raise ValueError(f"Invalid source_field argument value: {self.source_field}") filter_task: TaskTypeDef = { "taskType": "Filter", "connectorOperator": {self.connector_type: "LESS_THAN"}, # type: ignore "sourceFields": [self.source_field], "taskProperties": { "DATA_TYPE": "datetime", "VALUE": str(datetime_to_epoch_ms(self.filter_date_parsed)), }, # NOT inclusive } self.hook.update_flow_filter( flow_name=self.flow_name, filter_tasks=[filter_task], set_trigger_ondemand=True
)
[docs]class AppflowRunAfterOperator(AppflowBaseOperator): """ Execute a Appflow run after updating the filters to select only future data. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRunAfterOperator` :param source: The source name (Supported: salesforce, zendesk) :param flow_name: The flow name :param source_field: The field name to apply filters :param filter_date: The date value (or template) to be used in filters. :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """
[docs] template_fields = ("filter_date",)
def __init__( self, source: str, flow_name: str, source_field: str, filter_date: str, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: if not filter_date: raise ValueError(MANDATORY_FILTER_DATE_MSG.format(entity="AppflowRunAfterOperator")) if source not in {"salesforce", "zendesk"}: raise ValueError(NOT_SUPPORTED_SOURCE_MSG.format(source=source, entity="AppflowRunAfterOperator")) super().__init__( source=source, flow_name=flow_name, flow_update=True, source_field=source_field, filter_date=filter_date, poll_interval=poll_interval, aws_conn_id=aws_conn_id, region=region, wait_for_completion=wait_for_completion, **kwargs, ) def _update_flow(self) -> None: if not self.filter_date_parsed: raise ValueError(f"Invalid filter_date argument parser value: {self.filter_date_parsed}") if not self.source_field: raise ValueError(f"Invalid source_field argument value: {self.source_field}") filter_task: TaskTypeDef = { "taskType": "Filter", "connectorOperator": {self.connector_type: "GREATER_THAN"}, # type: ignore "sourceFields": [self.source_field], "taskProperties": { "DATA_TYPE": "datetime", "VALUE": str(datetime_to_epoch_ms(self.filter_date_parsed)), }, # NOT inclusive } self.hook.update_flow_filter( flow_name=self.flow_name, filter_tasks=[filter_task], set_trigger_ondemand=True
)
[docs]class AppflowRunDailyOperator(AppflowBaseOperator): """ Execute a Appflow run after updating the filters to select only a single day. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRunDailyOperator` :param source: The source name (Supported: salesforce) :param flow_name: The flow name :param source_field: The field name to apply filters :param filter_date: The date value (or template) to be used in filters. :param poll_interval: how often in seconds to check the query status :param aws_conn_id: aws connection to use :param region: aws region to use :param wait_for_completion: whether to wait for the run to end to return """
[docs] template_fields = ("filter_date",)
def __init__( self, source: str, flow_name: str, source_field: str, filter_date: str, poll_interval: int = 20, aws_conn_id: str = "aws_default", region: str | None = None, wait_for_completion: bool = True, **kwargs, ) -> None: if not filter_date: raise ValueError(MANDATORY_FILTER_DATE_MSG.format(entity="AppflowRunDailyOperator")) if source != "salesforce": raise ValueError(NOT_SUPPORTED_SOURCE_MSG.format(source=source, entity="AppflowRunDailyOperator")) super().__init__( source=source, flow_name=flow_name, flow_update=True, source_field=source_field, filter_date=filter_date, poll_interval=poll_interval, aws_conn_id=aws_conn_id, region=region, wait_for_completion=wait_for_completion, **kwargs, ) def _update_flow(self) -> None: if not self.filter_date_parsed: raise ValueError(f"Invalid filter_date argument parser value: {self.filter_date_parsed}") if not self.source_field: raise ValueError(f"Invalid source_field argument value: {self.source_field}") start_filter_date = self.filter_date_parsed - timedelta(milliseconds=1) end_filter_date = self.filter_date_parsed + timedelta(days=1) filter_task: TaskTypeDef = { "taskType": "Filter", "connectorOperator": {self.connector_type: "BETWEEN"}, # type: ignore "sourceFields": [self.source_field], "taskProperties": { "DATA_TYPE": "datetime", "LOWER_BOUND": str(datetime_to_epoch_ms(start_filter_date)), # NOT inclusive "UPPER_BOUND": str(datetime_to_epoch_ms(end_filter_date)), # NOT inclusive }, } self.hook.update_flow_filter( flow_name=self.flow_name, filter_tasks=[filter_task], set_trigger_ondemand=True
)
[docs]class AppflowRecordsShortCircuitOperator(ShortCircuitOperator): """ Short-circuit in case of a empty Appflow's run. .. seealso:: For more information on how to use this operator, take a look at the guide: :ref:`howto/operator:AppflowRecordsShortCircuitOperator` :param flow_name: The flow name :param appflow_run_task_id: Run task ID from where this operator should extract the execution ID :param ignore_downstream_trigger_rules: Ignore downstream trigger rules :param aws_conn_id: aws connection to use :param region: aws region to use """
[docs] ui_color = "#33ffec" # Light blue
def __init__( self, *, flow_name: str, appflow_run_task_id: str, ignore_downstream_trigger_rules: bool = True, aws_conn_id: str = "aws_default", region: str | None = None, **kwargs, ) -> None: super().__init__( python_callable=self._has_new_records_func, op_kwargs={ "flow_name": flow_name, "appflow_run_task_id": appflow_run_task_id, }, ignore_downstream_trigger_rules=ignore_downstream_trigger_rules, **kwargs, ) self.aws_conn_id = aws_conn_id self.region = region @staticmethod def _get_target_execution_id( records: list[ExecutionRecordTypeDef], execution_id: str ) -> ExecutionRecordTypeDef | None: for record in records: if record.get("executionId") == execution_id: return record return None @cached_property
[docs] def hook(self) -> AppflowHook: """Create and return an AppflowHook.""" return AppflowHook(aws_conn_id=self.aws_conn_id, region_name=self.region)
def _has_new_records_func(self, **kwargs) -> bool: appflow_task_id = kwargs["appflow_run_task_id"] self.log.info("appflow_task_id: %s", appflow_task_id) flow_name = kwargs["flow_name"] self.log.info("flow_name: %s", flow_name) af_client = self.hook.conn task_instance = kwargs["task_instance"] execution_id = task_instance.xcom_pull(task_ids=appflow_task_id, key="execution_id") # type: ignore if not execution_id: raise AirflowException(f"No execution_id found from task_id {appflow_task_id}!") self.log.info("execution_id: %s", execution_id) args = {"flowName": flow_name, "maxResults": 100} response: DescribeFlowExecutionRecordsResponseTypeDef = cast( "DescribeFlowExecutionRecordsResponseTypeDef", {} ) record = None while not record: if "nextToken" in response: response = af_client.describe_flow_execution_records(nextToken=response["nextToken"], **args) else: response = af_client.describe_flow_execution_records(**args) record = AppflowRecordsShortCircuitOperator._get_target_execution_id( response["flowExecutions"], execution_id ) if not record and "nextToken" not in response: raise AirflowException(f"Flow ({execution_id}) without recordsProcessed info.") execution = record.get("executionResult", {}) if "recordsProcessed" not in execution: raise AirflowException(f"Flow ({execution_id}) without recordsProcessed info!") records_processed = execution["recordsProcessed"] self.log.info("records_processed: %d", records_processed) task_instance.xcom_push("records_processed", records_processed) # type: ignore return records_processed > 0

Was this entry helpful?