如何使用Boto3 get_query_results方法从AWS Athena创建数据帧

Question

我使用AWS Athena查询S3的原始数据。因为雅典娜将查询输出写到了我以前做的S3输出bucket中：

df = pd.read_csv(OutputLocation)

但这似乎是一种昂贵的方式。最近我注意到了boto3的get_query_results方法，它返回结果的复杂字典。

client = boto3.client('athena')
response = client.get_query_results(
        QueryExecutionId=res['QueryExecutionId']
        )

我面临两个主要问题：

1. 如何将get_query_results的结果格式化为pandas数据帧？
2. get_query_results只返回1000行。我怎么能用它得到两百万行呢？

Answer 1

get_query_results只返回1000行。如何使用它将200万行放入熊猫数据帧？

如果您尝试添加：

client.get_query_results(QueryExecutionId=res['QueryExecutionId'], MaxResults=2000)

您将获得下一个错误：

An error occurred (InvalidRequestException) when calling the GetQueryResults operation: MaxResults is more than maximum allowed length 1000.

如果直接从bucket s3获取文件（在下一个示例中是Pandas数据帧），则可以获得数百万行：

def obtain_data_from_s3(self):
    self.resource = boto3.resource('s3', 
                          region_name = self.region_name, 
                          aws_access_key_id = self.aws_access_key_id,
                          aws_secret_access_key= self.aws_secret_access_key)

    response = self.resource \
    .Bucket(self.bucket) \
    .Object(key= self.folder + self.filename + '.csv') \
    .get()

    return pd.read_csv(io.BytesIO(response['Body'].read()), encoding='utf8')   

self.filename可以是：

self.filename = response['QueryExecutionId'] + ".csv"

因为雅典娜将文件命名为QueryExecutionId。我将为您编写所有代码，这些代码接受一个查询并返回一个包含所有行和列的数据框。

import time
import boto3
import pandas as pd
import io

class QueryAthena:

    def __init__(self, query, database):
        self.database = database
        self.folder = 'my_folder/'
        self.bucket = 'my_bucket'
        self.s3_input = 's3://' + self.bucket + '/my_folder_input'
        self.s3_output =  's3://' + self.bucket + '/' + self.folder
        self.region_name = 'us-east-1'
        self.aws_access_key_id = "my_aws_access_key_id"
        self.aws_secret_access_key = "my_aws_secret_access_key"
        self.query = query

    def load_conf(self, q):
        try:
            self.client = boto3.client('athena', 
                              region_name = self.region_name, 
                              aws_access_key_id = self.aws_access_key_id,
                              aws_secret_access_key= self.aws_secret_access_key)
            response = self.client.start_query_execution(
                QueryString = q,
                    QueryExecutionContext={
                    'Database': self.database
                    },
                    ResultConfiguration={
                    'OutputLocation': self.s3_output,
                    }
            )
            self.filename = response['QueryExecutionId']
            print('Execution ID: ' + response['QueryExecutionId'])

        except Exception as e:
            print(e)
        return response                

    def run_query(self):
        queries = [self.query]
        for q in queries:
            res = self.load_conf(q)
        try:              
            query_status = None
            while query_status == 'QUEUED' or query_status == 'RUNNING' or query_status is None:
                query_status = self.client.get_query_execution(QueryExecutionId=res["QueryExecutionId"])['QueryExecution']['Status']['State']
                print(query_status)
                if query_status == 'FAILED' or query_status == 'CANCELLED':
                    raise Exception('Athena query with the string "{}" failed or was cancelled'.format(self.query))
                time.sleep(10)
            print('Query "{}" finished.'.format(self.query))

            df = self.obtain_data()
            return df

        except Exception as e:
            print(e)      

    def obtain_data(self):
        try:
            self.resource = boto3.resource('s3', 
                                  region_name = self.region_name, 
                                  aws_access_key_id = self.aws_access_key_id,
                                  aws_secret_access_key= self.aws_secret_access_key)

            response = self.resource \
            .Bucket(self.bucket) \
            .Object(key= self.folder + self.filename + '.csv') \
            .get()

            return pd.read_csv(io.BytesIO(response['Body'].read()), encoding='utf8')   
        except Exception as e:
            print(e)  


if __name__ == "__main__":       
    query = "SELECT * FROM bucket.folder"
    qa = QueryAthena(query=query, database='myAthenaDb')
    dataframe = qa.run_query()

Answer 2

我有第一个问题的解决方案，使用以下函数

def results_to_df(results):

    columns = [
        col['Label']
        for col in results['ResultSet']['ResultSetMetadata']['ColumnInfo']
    ]

    listed_results = []
    for res in results['ResultSet']['Rows'][1:]:
         values = []
         for field in res['Data']:
            try:
                values.append(list(field.values())[0]) 
            except:
                values.append(list(' '))

        listed_results.append(
            dict(zip(columns, values))
        )

    return listed_results

然后：

t = results_to_df(response)
pd.DataFrame(t)

至于我的第二个问题和@EricBellet的请求，我还添加了分页方法，与在S3中加载雅典娜输出的结果相比，我发现这种方法效率低、耗时长：

def run_query(query, database, s3_output):
    ''' 
    Function for executing Athena queries and return the query ID 
    '''
    client = boto3.client('athena')
    response = client.start_query_execution(
        QueryString=query,
        QueryExecutionContext={
            'Database': database
            },
        ResultConfiguration={
            'OutputLocation': s3_output,
            }
        )
    print('Execution ID: ' + response['QueryExecutionId'])
    return response



def format_result(results):
    '''
    This function format the results toward append in the needed format.
    '''
    columns = [
        col['Label']
        for col in results['ResultSet']['ResultSetMetadata']['ColumnInfo']
    ]

    formatted_results = []

    for result in results['ResultSet']['Rows'][0:]:
        values = []
        for field in result['Data']:
            try:
                values.append(list(field.values())[0]) 
            except:
                values.append(list(' '))

        formatted_results.append(
            dict(zip(columns, values))
        )
    return formatted_results



res = run_query(query_2, database, s3_ouput) #query Athena



import sys
import boto3

marker = None
formatted_results = []
query_id = res['QueryExecutionId']
i = 0
start_time = time.time()

while True:
    paginator = client.get_paginator('get_query_results')
    response_iterator = paginator.paginate( 
        QueryExecutionId=query_id,
        PaginationConfig={
            'MaxItems': 1000,
            'PageSize': 1000,
            'StartingToken': marker})

    for page in response_iterator:
        i = i + 1
        format_page = format_result(page)
        if i == 1:
            formatted_results = pd.DataFrame(format_page)
        elif i > 1:
            formatted_results = formatted_results.append(pd.DataFrame(format_page))

    try:
        marker = page['NextToken']
    except KeyError:
        break

print ("My program took", time.time() - start_time, "to run")

它的格式不是很好，但我认为它能完成任务。。。