BigQuery - Select 多列并希望排除两个双嵌套列

BigQuery - Select multiple columns and want to exclude two double nested columns

您好,我正在处理 lt-pcf-analytics-exp.90676036.ga_sessions_* table,我需要提取不同的变量,包括嵌套命中列中的所有变量,hits.customDimensions.value 和 [=14= 列除外].我认为 hits 和 hits.customDimensions 都是 ARRAY。我如何在标准 SQL 中执行此操作?

我发现了一个关于类似问题的问题 (),但在我的例子中,我有一个双嵌套数组列,我无法调整代码。

基本上,这就是我要提取的内容。如何修改它以便排除 hits.customDimensions.valuehits.customDimensions.index?谢谢。

SELECT fullVisitorId,
   visitId,
   visitNumber,
   cd.value as PCF_CUST_ID,
   date,
   TIMESTAMP_SECONDS(visitStartTime) as visitStartTime,
   totals.visits as visits,
   totals.hits as total_hits,
   hits.* (EXCEPT hits.customDimensions.value and hits.customDimensions.index)
FROM `lt-pcf-analytics-exp.90676036.ga_sessions_*` as t
          left join unnest(customDimensions) as cd
          left join unnest(hits) as hits
WHERE _TABLE_SUFFIX between '20210101' and '20210131' 
   and cd.index = 4 and cd.value is not null
ORDER BY PCF_CUST_ID, visitStartTime, hitNumber

如果你检查 BigQuery documentation for except,这不是好的语法:

SELECT [ AS { typename | STRUCT | VALUE } ] [{ ALL | DISTINCT }]
    { [ expression. ]* [ EXCEPT ( column_name [, ...] ) ]
        [ REPLACE ( expression [ AS ] column_name [, ...] ) ]
    | expression [ [ AS ] alias ] } [, ...]

所以,像这样使用它:

SELECT hits.* EXCEPT (value, index)

就像@martinus 注意到的那样,您的 except 语法不正确。如果您看一下 BigQuery Documentation,您会发现 运行 带有例外的查询的正确方法是:

SELECT 
   field.* EXCEPT (nested_field1, nested_field2)
FROM `my_table`

但是,您不能直接在嵌套字段上直接使用 EXCEPT。作为解决方法,您可以从 hits.* 中排除所有 hits.customDimensions 值,然后仅 hits.customDimensions.* 排除 SELECT,然后排除您需要删除的嵌套元素,例如 indexvalue.

像下面这样的查询应该有效:

SELECT fullVisitorId,
   visitId,
   visitNumber,
   cd.value as PCF_CUST_ID,
   date,
   TIMESTAMP_SECONDS(visitStartTime) as visitStartTime,
   totals.visits as visits,
   totals.hits as total_hits,
   hits.* EXCEPT (hits.customDimensions),
   hits.customDimensions.* EXCEPT (index, value)
FROM `lt-pcf-analytics-exp.90676036.ga_sessions_*` as t
          left join unnest(customDimensions) as cd
          left join unnest(hits) as hits
WHERE _TABLE_SUFFIX between '20210101' and '20210131' 
   and cd.index = 4 and cd.value is not null
ORDER BY PCF_CUST_ID, visitStartTime, hitNumber