Filtering DynamicFrame with AWS Glue or PySpark

问题

I have a table in my AWS Glue Data Catalog called 'mytable'. This table is in an on-premises Oracle database connection 'mydb'.

I'd like to filter the resulting DynamicFrame to only rows where the X_DATETIME_INSERT column (which is a timestamp) is greater than a certain time (in this case, '2018-05-07 04:00:00'). Afterwards, I'm trying to count the rows to ensure that the count is low (the table is about 40,000 rows, but only a few rows should meet the filter criteria).

Here is my current code:

import boto3
from datetime import datetime
import logging
import os
import pg8000
import pytz
import sys
from awsglue.context import GlueContext
from awsglue.job import Job
from awsglue.transforms import *
from awsglue.utils import getResolvedOptions
from base64 import b64decode
from pyspark.context import SparkContext
from pyspark.sql.functions import lit
## @params: [TempDir, JOB_NAME]
args = getResolvedOptions(sys.argv, ['TempDir','JOB_NAME'])

sc = SparkContext()
glueContext = GlueContext(sc)
spark = glueContext.spark_session
job = Job(glueContext)
job.init(args['JOB_NAME'], args)

datasource0 = glueContext.create_dynamic_frame.from_catalog(database = "mydb", table_name = "mytable", transformation_ctx = "datasource0")

# Try Glue native filtering    
filtered_df = Filter.apply(frame = datasource0, f = lambda x: x["X_DATETIME_INSERT"] > '2018-05-07 04:00:00')
filtered_df.count()

This code runs for 20 minutes and times out. I've tried other variations:

df = datasource0.toDF()
df.where(df.X_DATETIME_INSERT > '2018-05-07 04:00:00').collect()

And

df.filter(df["X_DATETIME_INSERT"].gt(lit("'2018-05-07 04:00:00'")))

Which have failed. What am I doing wrong? I'm experienced in Python but new to Glue and PySpark.

回答1:

AWS Glue loads entire dataset from your JDBC source into temp s3 folder and applies filtering afterwards. If your data was in s3 instead of Oracle and partitioned by some keys (ie. /year/month/day) then you could use pushdown-predicate feature to load a subset of data:

val partitionPredicate = s"to_date(concat(year, '-', month, '-', day)) BETWEEN '${fromDate}' AND '${toDate}'"

val df = glueContext.getCatalogSource(
   database = "githubarchive_month",
   tableName = "data",
   pushDownPredicate = partitionPredicate).getDynamicFrame()

Unfortunately, this doesn't work for JDBC data sources yet.

来源：https://stackoverflow.com/questions/50256995/filtering-dynamicframe-with-aws-glue-or-pyspark