如何将嵌套展平转换为标准 SQL

Question

目前使用来自 google 的 BigQuery。我的旧查询如下所示：

Select * FROM FLATTEN(FLATTEN([DB.table1], record.atr), record.atr.value) 
WHERE record.atr.value IN "1"

我了解BigQuery now automatically flattens 表，所以我认为这可以解决问题：

SELECT * FROM `DB.table1`
WHERE record.atr.value IN "1" 

但是我收到一个错误错误：无法访问类型为 ARRAY、sv ARRAY、...>> 在 [2:17] 的值的字段值

如何将其转换为标准 SQL 查询？

Answer 1

如何将其转换为标准 SQL 查询？
选择 * FROM `DB.table1`
WHERE record.atr.value IN "1"

以下是 BigQuery 标准 SQL

#standardSQL
SELECT *
FROM `DB.table1`
WHERE (SELECT COUNT(1) 
        FROM UNNEST(record.atr) AS atr, 
              UNNEST(atr.value) AS val 
        WHERE val IN ('1')) > 0  

您可以使用虚拟数据（部分借鉴 Elliott 的示例）对其进行测试

#standardSQL
WITH `DB.table1` AS (
  SELECT 1 AS id, STRUCT([STRUCT<x STRING, value ARRAY<STRING>>('foo', ['1', '2']), ('bar', ['3', '4', '5'])] AS atr) AS record UNION ALL
  SELECT 2, STRUCT([STRUCT<x STRING, value ARRAY<STRING>>('baz', ['6', '7'])])
)
SELECT *
FROM `DB.table1`
WHERE (SELECT COUNT(1) 
        FROM UNNEST(record.atr) AS atr, 
              UNNEST(atr.value) AS val 
        WHERE val IN ('1')) > 0

Answer 2

BigQuery 在使用标准 SQL 时不会自动展平数组；您提供的链接表明相反。在您的情况下，查询将是这样的：

#standardSQL
SELECT * EXCEPT(value), value
FROM `Db.table1`
CROSS JOIN UNNEST(record.atr) AS atr
CROSS JOIN UNNEST(atr.value) AS value;

如果您想为空数组获取一行，您可以改用LEFT JOIN：

#standardSQL
SELECT * EXCEPT(value), value
FROM `Db.table1`
LEFT JOIN UNNEST(record.atr) AS atr
LEFT JOIN UNNEST(atr.value) AS value;

这是一个您可以尝试运行的示例查询：

#standardSQL
WITH Input AS (
  SELECT STRUCT(ARRAY<STRUCT<x STRING, value ARRAY<INT64>>>[('foo', [1, 2]), ('bar', [3, 4, 5])] AS atr) AS record UNION ALL
  SELECT STRUCT(ARRAY<STRUCT<x STRING, value ARRAY<INT64>>>[('baz', [6, 7])])
)
SELECT * EXCEPT (value)
FROM Input
CROSS JOIN UNNEST(record.atr) AS atr
CROSS JOIN UNNEST(atr.value) AS value;