雪花中不支持的子查询答案

【问题标题】：Unsupported subquery in snowflake雪花中不支持的子查询
【发布时间】：2021-03-12 12:59:52
【问题描述】：

我正在尝试在雪花中运行查询：


SELECT DISTINCT RPU.RID,
RA.DATE,
COALESCE(RTH.CT, 0) as RCNT,
COALESCE (R.COMMENT_SITE, PR.COMMENT_SITE) as COMMENT_SITE,

##########################
COALESCE ((SELECT SPD FROM "DB"."SCHEMA"."PRD" WHERE
           COMMENT_SITE = R.COMMENT_SITE AND CLOTH_ID = RPU.GPR
           ORDER BY (IS_ACTIVE = false) LIMIT 1), PR.SPD) as SPD
#########################
FROM "DB"."SCHEMA"."CLOTH_USED" RPU JOIN "DB"."SCHEMA"."CAGR" RA ON RPU.RID = RA.RID 
LEFT OUTER JOIN "DB"."SCHEMA"."PNTR_IMP" PR ON RPU.COMMENT_ID[1] = PR.ID
left outer join "DB"."SCHEMA"."COMMENTS" R ON RPU.COMMENT_ID[1] = R.ID
LEFT OUTER JOIN CLOTH_COUNT RTH ON RTH.CLOTH_ID = RPU.GPR 
ORDER BY RPU.RID DESC, COALESCE(RTH.CT, 0) ASC, RA.DATE DESC LIMIT 1;

我得到的错误是：

SQL 编译错误：无法评估不受支持的子查询类型

以下查询工作正常：

SELECT DISTINCT RPU.RID,
RA.DATE,
COALESCE(RTH.CT, 0) as RCNT,
COALESCE (R.COMMENT_SITE, PR.COMMENT_SITE) as COMMENT_SITE,
FROM "DB"."SCHEMA"."CLOTH_USED" RPU JOIN "DB"."SCHEMA"."CAGR" RA ON RPU.RID = RA.RID 
LEFT OUTER JOIN "DB"."SCHEMA"."PNTR_IMP" PR ON RPU.COMMENT_ID[1] = PR.ID
left outer join "DB"."SCHEMA"."COMMENTS" R ON RPU.COMMENT_ID[1] = R.ID
LEFT OUTER JOIN CLOTH_COUNT RTH ON RTH.CLOTH_ID = RPU.GPR 
ORDER BY RPU.RID DESC, COALESCE(RTH.CT, 0) ASC, RA.DATE DESC LIMIT 1;

SQL 编译错误：无法评估不支持的子查询类型，添加时出现：

COALESCE ((SELECT SPD FROM "DB"."SCHEMA"."PRD" WHERE
           COMMENT_SITE = R.COMMENT_SITE AND CLOTH_ID = RPU.GPR
           ORDER BY (IS_ACTIVE = false) LIMIT 1), PR.SPD) as SPD

如何更正查询？问题在于 where 子句，而没有 where 可以正常工作。如何重写查询？

【问题讨论】：

你能不能用 COLAESCE 函数运行子查询，看看是否有效。
@hkandpal 如果我删除 where 条件，它可以工作，但在 where 条件下它不起作用。
问题是当我对这些值进行硬编码时，它无法引用 R.COMMENT_SITE 和 RPU.GPR

标签： sql snowflake-cloud-data-platform

【解决方案1】：

因此，在您的情况下，相关子查询通常可以转换为某种形式的连接，您正在做的主要事情是在两列上匹配并选择随机出现的内容（如果您有超过 1 行处于 active=false 状态）行，这可以转换为 row_number() 并使用 QUALITY 作为 LIMIT 的等价物。

因此：

WITH sub_query AS (
    SELECT 
        spd
        comment_site
        cloth_id
    FROM db.schema.prd
    QUALIFY row_number() OVER (PARTITION BY comment_site, cloth_id ORDER BY IS_ACTIVE = false) = 1
)
SELECT DISTINCT 
    rpu.rid,
    ra.date,
    COALESCE(rth.ct, 0) AS rcnt,
    COALESCE (r.comment_site, pr.comment_site) AS comment_site,

    COALESCE ( sq.spd, pr.spd) as SPD

FROM DB.schema.cloth_used AS rpu 
JOIN db.schema.cagr  AS ra 
    ON rpu.rid = ra.rid 
LEFT OUTER JOIN db.schema.pntr_imp AS pr 
    ON rpu.comment_id[1] = pr.id
LEFT OUTER JOIN db.schema.comments AS r 
    ON rpu.comment_id[1] = r.id
LEFT OUTER JOIN sub_query AS sq
    ON sq.comment_site = r.comment_site AND sq.cloth_id = rpu.gpr
LEFT OUTER JOIN cloth_count AS rth 
    ON RTH.CLOTH_ID = RPU.GPR 
ORDER BY rpu.rid DESC, COALESCE(rth.ct, 0) ASC, ra.date DESC 
LIMIT 1;

应该满足您的需求，现在我使用了 CTE，但如果您愿意，可以将其移至 JOIN 区域，或者像这样需要：

SELECT DISTINCT 
    rpu.rid,
    ra.date,
    COALESCE(rth.ct, 0) AS rcnt,
    COALESCE (r.comment_site, pr.comment_site) AS comment_site,

    COALESCE ( sq.spd, pr.spd) as SPD

FROM DB.schema.cloth_used AS rpu 
JOIN db.schema.cagr  AS ra 
    ON rpu.rid = ra.rid 
LEFT OUTER JOIN db.schema.pntr_imp AS pr 
    ON rpu.comment_id[1] = pr.id
LEFT OUTER JOIN db.schema.comments AS r 
    ON rpu.comment_id[1] = r.id
LEFT OUTER JOIN (
    SELECT 
        spd
        comment_site
        cloth_id
    FROM db.schema.prd
    QUALIFY row_number() OVER (PARTITION BY comment_site, cloth_id ORDER BY IS_ACTIVE = false) = 1
) AS sq
    ON sq.comment_site = r.comment_site AND sq.cloth_id = rpu.gpr
LEFT OUTER JOIN cloth_count AS rth 
    ON RTH.CLOTH_ID = RPU.GPR 
ORDER BY rpu.rid DESC, COALESCE(rth.ct, 0) ASC, ra.date DESC 
LIMIT 1;

【讨论】：