使用间接架构更改更新BigQuery视图

时间:2019-01-04 11:22:42

标签: python google-bigquery

在更新视图时,似乎不会采用间接模式更改。

复制步骤

  • 使用字段view1(例如field1)创建SELECT 1 AS field1
  • 创建view2,从view1中选择所有字段
  • 更新view1使其也包含field2(例如SELECT 1 AS field1, 2 AS field2
  • 使用与以前相同的查询更新view2(由于documented limitation

所需结果

  • view1view2的架构,包括field1field2
  • 查看更新应该是原子的

实际结果

  • view1的架构已正确更新(包括field1field2
  • view2的架构仅包括field1
  • view2中进行选择实际上会返回field1field2

我可以删除view2并重新创建它,但这不是原子的,并且有时会出现视图不可用的情况。

我还尝试更新view2的schema属性,但是Cannot add fields (field: field2)拒绝了该属性:

  

google.api_core.exceptions.BadRequest:400 PATCH https://www.googleapis.com/bigquery/v2/projects/<project-id>/datasets/dataset1/tables/view2:提供的架构与表:dataset1.view2不匹配。无法添加字段(字段:field2)

问题

有什么方法可以自动更新视图,同时还可以更新间接更改的模式(视图从中选择的表/视图)。

注意:当然,我的view2将添加其他字段,并且我目前可以通过创建新的临时视图来确定其架构。

注意:该模式很重要,因为诸如Data Studio的BigQuery连接器之类的工具正在检查该模式。

代码以重现步骤

# Python 3.6+
import google.api_core.exceptions
from google.cloud import bigquery


def delete_table_if_exists(client: bigquery.Client, table: bigquery.Table):
    try:
        client.delete_table(table)
    except google.api_core.exceptions.NotFound:
        pass


def full_table_id(table: bigquery.Table) -> str:
    # Note: the documentation says it should be separated by a dot but uses a colon
    return table.full_table_id.replace(':', '.')


def view_test():
    client = bigquery.Client()

    dataset_ref = client.dataset('dataset1')
    try:
        client.create_dataset(dataset_ref)
    except google.api_core.exceptions.Conflict:
        pass

    view1 = bigquery.Table(dataset_ref.table('view1'))
    view2 = bigquery.Table(dataset_ref.table('view2'))
    delete_table_if_exists(client, view1)
    delete_table_if_exists(client, view2)

    view1.view_query = 'SELECT 1 AS field1'
    view1 = client.create_table(view1)

    view2.view_query = f'SELECT * FROM `{full_table_id(view1)}`'
    client.create_table(view2)

    view1.view_query = 'SELECT 1 AS field1, 2 AS field2'
    client.update_table(view1, ['view_query'])

    client.update_table(view2, ['view_query'])
    print('view2 schema:', client.get_table(view2).schema)

    # trying to update the schema fails with 'Cannot add fields (field: field2)'
    view2.schema = client.get_table(view1).schema
    client.update_table(view2, ['schema'])


if __name__ == '__main__':
    view_test()

Bash示例执行相同操作

#!/bin/bash

set -e

project_id=$(gcloud config list --format 'value(core.project)' 2>/dev/null)

bq mk -f dataset1

bq rm -f dataset1.view1
bq rm -f dataset1.view2

bq mk --use_legacy_sql=false --view 'SELECT 1 AS field1' dataset1.view1
bq mk --use_legacy_sql=false --view 'SELECT * FROM `'$project_id'.dataset1.view1`' dataset1.view2

bq update --use_legacy_sql=false --view 'SELECT 1 AS field1, 2 AS field2' dataset1.view1
bq update --use_legacy_sql=false --view 'SELECT * FROM `'$project_id'.dataset1.view1`' dataset1.view2

bq show dataset1.view2

更新:带有可接受答案的代码

Python代码

def get_create_or_replace_view_query(view: bigquery.Table) -> str:
    return f'CREATE OR REPLACE VIEW {view.dataset_id}.{view.table_id} AS {view.view_query}'


def view_test():
    # ...
    query_job = client.query(get_create_or_replace_view_query(view2))
    query_job.result()
    print('view2 schema:', client.get_table(view2).schema)

打击魔术

bq query --use_legacy_sql=false 'CREATE OR REPLACE VIEW dataset1.view2 AS SELECT * FROM `'$project_id'.dataset1.view1`'

1 个答案:

答案 0 :(得分:2)

您应该使用CREATE OR REPLACE VIEW语句;请参阅related documentation。 BigQuery为所有执行表修改的查询提供ACID语义,CREATE OR REPLACE VIEW也不例外,因此这原子地替换了视图的定义和架构。