我遇到一个问题，即我正在使用python创建数据流模板，并且该模板在启动新的数据流作业时需要接受3个用户定义的参数。
beam.io.gcp.pubsub.WriteToPubSub（）出现问题，我尝试从ValueProvider提供主题名称，根据Google文档，创建模板时需要该主题名称：
https://cloud.google.com/dataflow/docs/guides/templates/creating-templates
源beam.io.ReadFromPubSub（）成功地接受了订阅值的值提供者，而变换beam.io.gcp.bigquery.WriteToBigQuery（）也是如此。
显然共享我的代码会有所帮助：）
首先是通常的进口：
from __future__ import absolute_import
import argparse
import apache_beam as beam
from apache_beam.options.pipeline_options import PipelineOptions
from apache_beam.options.pipeline_options import SetupOptions
from apache_beam.options.pipeline_options import StandardOptions
from apache_beam.options.value_provider import StaticValueProvider
import json
import time
from datetime import datetime
import dateutil.parser
import sys
接下来是我为提供给模板的输入参数定义的类：
class userOptions(PipelineOptions):
@classmethod
def _add_argparse_args(cls, parser):
parser.add_value_provider_argument(
'--subscription',
default='projects/MYPROJECT/subscrpiptions/subscription',
help='PubSub subscription to listen on')
parser.add_value_provider_argument(
'--bqtable',
default='dataset.table',
help='Big Query Table Name in the format project:dataset.table')
parser.add_value_provider_argument(
'--topic',
default='projects/MYPROJECT/subscrpiptions/subscription',
help='PubSub topic to write failed messages to')
并且管道本身定义为（请注意，我省略了map函数）
def run():
user_options = PipelineOptions().view_as(userOptions)
pipeline_options = PipelineOptions()
pipeline_options.view_as(SetupOptions).save_main_session = True
pipeline_options.view_as(StandardOptions).streaming = True
with beam.Pipeline(options=pipeline_options) as p:
records = (
p | 'Read from PubSub'
>> beam.io.ReadFromPubSub(
subscription=str(user_options.subscription),
id_label='Message_ID',
with_attributes=True)
| 'Format Message' >>
beam.Map(format_message_element)
| 'Transform null records to empty list' >>
beam.Map(transform_null_records)
| 'Transform Dates' >>
beam.Map(format_dates)
| 'Write to Big Query' >>
beam.io.gcp.bigquery.WriteToBigQuery(
table=user_options.bqtable,
create_disposition='CREATE_IF_NEEDED',
write_disposition='WRITE_APPEND',
insert_retry_strategy='RETRY_NEVER'
)
| 'Write Failures to Pub Sub' >>
beam.io.gcp.pubsub.WriteToPubSub(user_options.topic)
)
现在，当我尝试使用powershell命令生成模板时：
python profiles-pipeline.py --project xxxx-xxxxxx-xxxx `
--subscription projects/xxxx-xxxxxx-xxxx/subscriptions/sub-xxxx-xxxxxx-xxxx-dataflow `
--bqtable xxxx-xxxxxx-xxxx:dataset.table `
--topic projects/xxxx-xxxxxx-xxxx/topics/top-xxxx-xxxxxx-xxxx-failures `
--runner DataflowRunner `
--temp_location gs://xxxx-xxxxxx-xxxx/temp/ `
--staging_location gs://xxxx-xxxxxx-xxxx/staging/ `
--template_location gs://xxxx-xxxxxx-xxxx/template
我收到此错误：
File "pipeline.py", line 193, in <module>
run()
File "pipeline.py", line 183, in run
beam.io.gcp.pubsub.WriteToPubSub(user_options.topic)
File "C:\github\pipeline-dataflow-jobs\dataflow\lib\site-packages\apache_beam\io\gcp\pubsub.py", line 292, in __init__
topic, id_label, with_attributes, timestamp_attribute)
File "C:\github\pipeline-dataflow-jobs\dataflow\lib\site-packages\apache_beam\io\gcp\pubsub.py", line 430, in __init__
self.project, self.topic_name = parse_topic(topic)
File "C:\github\pipeline-dataflow-jobs\dataflow\lib\site-packages\apache_beam\io\gcp\pubsub.py", line 325, in parse_topic
match = re.match(TOPIC_REGEXP, full_topic)
File "c:\program files\python37\lib\re.py", line 173, in match
return _compile(pattern, flags).match(string)
TypeError: expected string or bytes-like object
我在尝试使用beam.io.WriteToBigQuery（）之前遇到此错误，但是一旦更改为beam.io.gcp.bigquery.WriteToBigQuery（），该错误就得到解决，因为它接受表名的ValueProvider。但是对于pubsub来说，我找不到可行的替代方法。
任何帮助将不胜感激。
我已部分解决了该问题，因为我的管道错误地尝试将失败的插入发布到Big Query，但是仍然存在无法将pubsub主题名称作为输入参数传递的问题。但是，如果主题名称是硬编码的，它将起作用]
#################################################################
# Import the libraries required by the pipeline #
#################################################################
from __future__ import absolute_import
import argparse
import apache_beam as beam
from apache_beam.options.pipeline_options import PipelineOptions
from apache_beam.options.pipeline_options import SetupOptions
from apache_beam.options.pipeline_options import StandardOptions
from apache_beam.options.value_provider import RuntimeValueProvider
import json
import time
from datetime import datetime
import dateutil.parser
import sys
import logging
#################################################################
# Create a class for the user defined settings provided at job
# creation
#################################################################
class userOptions(PipelineOptions):
@classmethod
def _add_argparse_args(cls, parser):
parser.add_value_provider_argument(
'--subscription',
default='projects/MYPROJECT/subscrpiptions/subscription',
help='PubSub subscription to listen on')
parser.add_value_provider_argument(
'--bqtable',
default='dataset.table',
help='Big Query Table Name in the format project:dataset.table')
parser.add_value_provider_argument(
'--topic',
default='projects/MYPROJECT/topics/subscription',
help='Pubsub topic to write failures to')
##############################################################################
# Format failure message
##############################################################################
def format_failed_message(data):
try:
message=json.dumps(data)
except:
print("customError in function format_failed_message occured.", sys.exc_info(), "Message contents: ", data)
return message
#################################################################
# create a function called run #
#################################################################
def run():
##############################################################
# Setup the pipeline options with both passed in arguments
# and streaming options
##############################################################
user_options = PipelineOptions().view_as(userOptions)
pipeline_options = PipelineOptions()
pipeline_options.view_as(SetupOptions).save_main_session = True
pipeline_options.view_as(StandardOptions).streaming = True
##############################################################
# Define the pipeline
##############################################################
with beam.Pipeline(options=pipeline_options) as p:
# First we create a PCollection which will contain the messages read from Pubsub
records = (
p | 'Read from PubSub'
>> beam.io.ReadFromPubSub(
subscription=str(user_options.subscription),
id_label='Message_ID',
with_attributes=True)
# Transform the message and its attributes to a dict.
| 'Format Message' >>
beam.Map(format_message_element)
# Transform the empty arrays defined as element:null to element:[].
| 'Transform null records to empty list' >>
beam.Map(transform_null_records)
# Transform the dateCreated and DateModified to a big query compatible timestamp format.
| 'Transform Dates' >>
beam.Map(format_dates)
# Attempt to write the rows to BQ
| 'Write to Big Query' >>
beam.io.gcp.bigquery.WriteToBigQuery(
table=user_options.bqtable,
create_disposition='CREATE_IF_NEEDED',
write_disposition='WRITE_APPEND',
insert_retry_strategy='RETRY_NEVER'
)
)
#For any rows that failed to write to BQ
failed_data = (records[beam.io.gcp.bigquery.BigQueryWriteFn.FAILED_ROWS]
#Format the dictionary to a string
| 'Format the dictionary as a string for publishing' >>
beam.Map(format_failed_message)
#Encode the string to utf8 bytes
| 'Encode the message' >>
beam.Map(lambda x: x.encode('utf-8')).with_output_types(bytes)
)
#Published the failed rows to pubsub
failed_data | beam.io.gcp.pubsub.WriteToPubSub(topic='projects/xxxx-xxxxx-xxxxxx/topics/top-xxxxx-failures')
#failed_data | beam.io.gcp.pubsub.WriteToPubSub(topic=user_options.topic)
# As this is a streaming pipeline it will run continuosly till either we
# stop the pipeline or it fails.
result = p.run()
result.wait_until_finish()
#At the main entry point call the run function
if __name__ == '__main__':
#logging.getLogger().setLevel(logging.INFO)
run()