如何使用结构数据类型映射选择 Azure Databricks 中增量表列中的记录？

Question

我在 Databricks 中有一个名为

prod.silver.platform_ctl_ingestion

的增量表。它有几列，包括具有字符串数据类型的

table_name

和具有以下结构的

transform_options

：

 |-- transform_options: map (nullable = true)
 |    |-- key: string
 |    |-- value: struct (valueContainsNull = true)
 |    |    |-- col_name_mappings: map (nullable = true)
 |    |    |    |-- key: string
 |    |    |    |-- value: string (valueContainsNull = true)
 |    |    |-- type_mappings: map (nullable = true)
 |    |    |    |-- key: string
 |    |    |    |-- value: string (valueContainsNull = true)
 |    |    |-- partition_duplicates_by: array (nullable = true)
 |    |    |    |-- element: string (containsNull = true)
 |    |    |-- order_duplicates_by: array (nullable = true)
 |    |    |    |-- element: string (containsNull = true)

例如，当

table_name

为

prod.silver.weather

时，

transform_options

为：

{
"prod.bronze.weather_source_a":{"col_name_mappings":{"col_a_old":"col_a_new","col_b_old":"col_b_new"},"type_mappings":{"col_a_new":"INT","col_b_new":"TIMESTAMP"},"partition_duplicates_by":["col_a_new"],"order_duplicates_by":["_commit_version"]},
"prod.bronze.weather_source_b":{"col_name_mappings":{"col_c_old":"col_c_new","col_d_old":"col_d_new"},"type_mappings":{"col_c_new":"INT","col_d_new":"TIMESTAMP"},"partition_duplicates_by":["col_c_new"],"order_duplicates_by":["ingestion_timestamp","_commit_version"]}
}

我想选择

order_duplicates_by

中

transform_options

包含

_commit_version

的行。

我尝试了以下查询，但它返回所有行。

SELECT 
    *,
    transform_values(transform_options, (k, v) -> 
    CASE
    WHEN array_contains(v.order_duplicates_by, '_commit_version') THEN
    STRUCT(v.col_name_mappings, v.type_mappings, v.partition_duplicates_by, v.order_duplicates_by)
    ELSE NULL
    END
    ) AS updated_transform_options
FROM prod.silver.platform_ctl_ingestion;

知道该怎么做吗？

Answer 1

如果您正在过滤，那么很容易删除数组中的

_commit_version

。

命令：

%sql
SELECT 
    *,
    transform_values(transform_options, (k, v) -> 
    STRUCT(v.col_name_mappings, v.type_mappings, v.partition_duplicates_by, array_append(array_remove(v.order_duplicates_by, '_commit_version'), 'commit_version') as order_duplicates_by)
    ) AS updated_transform_options
FROM control_table
WHERE array_contains(map_values(transform_options)[0]['order_duplicates_by'],'_commit_version');

无需检查

_commit_version

中的

transform_values

。