应用this或this示例来构建程序,每次尝试插入Big Query时,都会出现此错误:
OverflowError:[在运行“格式”时]日期值超出范围
我的光束管道是这样的:
Bigquery = (transformation
| 'Format' >> beam.ParDo(FormatBigQueryoFn())
| 'Write to BigQuery' >> beam.io.Write(beam.io.BigQuerySink(
'XXXX',
schema=TABLE_SCHEMA,
create_disposition=beam.io.BigQueryDisposition.CREATE_IF_NEEDED,
write_disposition=beam.io.BigQueryDisposition.WRITE_APPEND
)))
在FormatBigQueryoFn类中,它应该是窗口数据时间的逻辑
示例1的代码:
def timestamp2str(t, fmt='%Y-%m-%d %H:%M:%S.000'):
"""Converts a unix timestamp into a formatted string."""
return datetime.fromtimestamp(t).strftime(fmt)
class TeamScoresDict(beam.DoFn):
"""Formats the data into a dictionary of BigQuery columns with their values
Receives a (team, score) pair, extracts the window start timestamp, and
formats everything together into a dictionary. The dictionary is in the format
{'bigquery_column': value}
"""
def process(self, team_score, window=beam.DoFn.WindowParam):
team, score = team_score
start = timestamp2str(int(window.start))
yield {
'team': team,
'total_score': score,
'window_start': start,
'processing_time': timestamp2str(int(time.time()))
}
示例2的代码:
class FormatDoFn(beam.DoFn):
def process(self, element, window=beam.DoFn.WindowParam):
ts_format = '%Y-%m-%d %H:%M:%S.%f UTC'
window_start = window.start.to_utc_datetime().strftime(ts_format)
window_end = window.end.to_utc_datetime().strftime(ts_format)
return [{'word': element[0],
'count': element[1],
'window_start':window_start,
'window_end':window_end}]
我的管道有什么问题?
编辑:
例如,如果我打印window.start,我将得到:
Timestamp(-9223372036860)
答案 0 :(得分:1)
问题是我要先从文件中读取数据,然后才能通过Google Pub / Sub进行测试。
当我从文件中读取数据时,元素没有时间戳。
必须在元素中添加时间戳。
发布/订阅会自动附加此时间戳。
最简单的窗口形式是使用固定时间的窗口:给定时间戳的PCollection可能会不断更新,每个窗口可能会捕获(例如)所有时间戳落在5分钟间隔内的元素。