AWS のサンプル非同期リモートサービス¶

import json
import time
import boto3

HTTP_METHOD_STRING = "httpMethod"
HEADERS_STRING = "headers"
BATCH_ID_STRING = "sf-external-function-query-batch-id"
DATA_STRING = "data"
REGION_NAME = "us-east-2"

TABLE_NAME = "Jobs"
IN_PROGRESS_STATUS = "IN_PROGRESS"
SUCCESS_STATUS = "SUCCESS"

def lambda_handler(event, context):
    # this is called from either the GET or POST
    if (HTTP_METHOD_STRING in event):
        method = event[HTTP_METHOD_STRING]
        if method == "POST":
            return initiate(event, context)
        elif method == "GET":
            return poll(event, context)
        else:
            return create_response(400, "Function called from invalid method")

    # if not called from GET or POST, then this lambda was called to
    # process data
    else:
        return process_data(event, context)


# Reads batch_ID and data from the request, marks the batch_ID as being processed, and
# starts the processing service.
def initiate(event, context):
    batch_id = event[HEADERS_STRING][BATCH_ID_STRING]
    data = json.loads(event["body"])[DATA_STRING]

    lambda_name = context.function_name

    write_to_storage(batch_id, IN_PROGRESS_STATUS, "NULL")
    lambda_response = invoke_process_lambda(batch_id, data, lambda_name)

    # lambda response returns 202, because we are invoking it with
    # InvocationType = 'Event'
    if lambda_response["StatusCode"] != 202:
        response = create_response(400, "Error in inititate: processing lambda not started")
    else:
        response = {
            'statusCode': lambda_response["StatusCode"]
        }

    return response


# Processes the data passed to it from the POST handler. In this example,
# the processing is to perform sentiment analysis on text.
def process_data(event, context):
    data = event[DATA_STRING]
    batch_id = event[BATCH_ID_STRING]

    def process_data_impl(data):
        comprehend = boto3.client(service_name='comprehend', region_name=REGION_NAME)
        # create return rows
        ret = []
        for i in range(len(data)):
            text = data[i][1]
            sentiment_response = comprehend.detect_sentiment(Text=text, LanguageCode='en')
            sentiment_score = json.dumps(sentiment_response['SentimentScore'])
            ret.append([i, sentiment_score])
        return ret

    processed_data = process_data_impl(data)
    write_to_storage(batch_id, SUCCESS_STATUS, processed_data)

    return create_response(200, "No errors in process")


# Repeatedly checks on the status of the batch_ID, and returns the result after the
# processing has been completed.
def poll(event, context):
    batch_id = event[HEADERS_STRING][BATCH_ID_STRING]
    processed_data = read_data_from_storage(batch_id)

    def parse_processed_data(response):
        # in this case, the response is the response from DynamoDB
        response_metadata = response['ResponseMetadata']
        status_code = response_metadata['HTTPStatusCode']

        # Take action depending on item status
        item = response['Item']
        job_status = item['status']
        if job_status == SUCCESS_STATUS:
            # the row number is stored at index 0 as a Decimal object,
            # we need to convert it into a normal int to be serialized to JSON
            data = [[int(row[0]), row[1]] for row in item['data']]
            return {
                'statusCode': 200,
                'body': json.dumps({
                    'data': data
                })
            }
        elif job_status == IN_PROGRESS_STATUS:
            return {
                'statusCode': 202,
                "body": "{}"
            }
        else:
            return create_response(500, "Error in poll: Unknown item status.")

    return parse_processed_data(processed_data)


def create_response(code, msg):
    return {
        'statusCode': code,
        'body': msg
    }


def invoke_process_lambda(batch_id, data, lambda_name):
    # Create payload to be sent to processing lambda
    invoke_payload = json.dumps({
        BATCH_ID_STRING: batch_id,
        DATA_STRING: data
    })

    # Invoke processing lambda asynchronously by using InvocationType='Event'.
    # This allows the processing to continue while the POST handler returns HTTP 202.
    lambda_client = boto3.client('lambda', region_name=REGION_NAME,)
    lambda_response = lambda_client.invoke(
        FunctionName=lambda_name,
        InvocationType='Event',
        Payload=invoke_payload
    )
    # returns 202 on success if InvocationType = 'Event'
    return lambda_response


def write_to_storage(batch_id, status, data):
    # we assume that the table has already been created
    client = boto3.resource('dynamodb')
    table = client.Table(TABLE_NAME)

    # Put in progress item in table
    item_to_store = {
        'batch_id': batch_id,
        'status': status,
        'data': data,
        'timestamp': "{}".format(time.time())
    }
    db_response = table.put_item(
        Item=item_to_store
    )


def read_data_from_storage(batch_id):
    # we assume that the table has already been created
    client = boto3.resource('dynamodb')
    table = client.Table(TABLE_NAME)

    response = table.get_item(Key={'batch_id': batch_id},
                          ConsistentRead=True)
    return response