Hive数据仓库行转列

查了很多资料发现网上很多文章都是转发和抄袭,有些问题。这里分享一个自己项目中使用的行转列例子,供大家参考。代码如下:

sql 复制代码
SELECT
  my_id,
  nm_cd_map['A'] AS my_cd_a,
  nm_cd_map['B'] AS my_cd_b,
  nm_cd_map['C'] AS my_cd_c,
  nm_num_map['A'] AS my_num_a,
  nm_num_map['B'] AS my_num_b,
  nm_num_map['C'] AS my_num_c
FROM
  (
    SELECT
      t.my_id,
      STR_TO_MAP(my_nm_cds,';',':') AS nm_cd_map,
      STR_TO_MAP(my_nm_nums,';',':') AS nm_num_map
    FROM
      (
        SELECT
          my_id,
          CONCAT_WS(';',COLLECT_LIST(CONCAT(my_nm,':',my_cd))) AS my_nm_cds,
          CONCAT_WS(';',COLLECT_LIST(CONCAT(my_nm,':',my_num))) AS my_nm_nums
        FROM
          (
            SELECT '1' AS my_id,'A' AS my_nm,'D01' AS my_cd,19 AS my_num
            UNION ALL
            SELECT '1' AS my_id,'B' AS my_nm,'D04' AS my_cd,18 AS my_num
            UNION ALL
            SELECT '1' AS my_id,'C' AS my_nm,'D02' AS my_cd,17 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'A' AS my_nm,'D03' AS my_cd,16 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'B' AS my_nm,'D05' AS my_cd,15 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'C' AS my_nm,'D06' AS my_cd,14 AS my_num
          )
        GROUP BY my_id
      ) t
  ) t
WHERE 1=1;

如果是在SparkSQL或Presto平台,或者阿里云的MaxCompute平台,还可使用如下方式:

sql 复制代码
-- 其实也可使用CONCAT然后STR_TO_MAP的方式,或者用MAP_FROM_ARRAYS,再或者用数组排序后ARRAY[n] AS的方式
SELECT
  my_id,
  nm_cd_map['A'] AS my_cd_a,
  nm_cd_map['B'] AS my_cd_b,
  nm_cd_map['C'] AS my_cd_c,
  nm_num_map['A'] AS my_num_a,
  nm_num_map['B'] AS my_num_b,
  nm_num_map['C'] AS my_num_c
FROM
  (
    SELECT
      t.my_id,
      MAP_FROM_ENTRIES(COLLECT_LIST(nm_cd)) AS nm_cd_map,
      MAP_FROM_ENTRIES(COLLECT_LIST(nm_num)) AS nm_num_map
    FROM
      (
        SELECT
          my_id,
          my_nm,
          my_cd,
          my_num,
          STRUCT(my_nm,my_cd) AS nm_cd,
          STRUCT(my_nm,my_num) AS nm_num
        FROM
          (
            SELECT '1' AS my_id,'A' AS my_nm,'D01' AS my_cd,19 AS my_num
            UNION ALL
            SELECT '1' AS my_id,'B' AS my_nm,'D04' AS my_cd,18 AS my_num
            UNION ALL
            SELECT '1' AS my_id,'C' AS my_nm,'D02' AS my_cd,17 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'A' AS my_nm,'D03' AS my_cd,16 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'B' AS my_nm,'D05' AS my_cd,15 AS my_num
            UNION ALL
            SELECT '2' AS my_id,'C' AS my_nm,'D06' AS my_cd,14 AS my_num
          )
      ) t
    GROUP BY my_id
  ) t
WHERE 1=1;
相关推荐
skywalk81631 天前
hive 常驻集群:像 OSPF 一样在线,像 git 一样可信
运维·人工智能·hive
Francek Chen2 天前
【大数据处理与分析】数据仓库Hive:09 Hive 编程实践
大数据·数据仓库·hive·hadoop·sql
skywalk81635 天前
卡不等于分支:我用 TRAE 宝配合 Hive 跑三张任务卡的一天
运维·人工智能·hive·trae
freejackman6 天前
【架构篇】双遍历 + 多栈协同:手撸 SQL 血缘引擎的架构思考
java·大数据·数据库·hive·sql·架构·spark
陈年老古董6 天前
ClickHouse 单机搭建记录
hive·笔记·学习·clickhouse·硬件架构
陈年老古董6 天前
Spark 3.1.3 集群搭建 + Spark-SQL 整合 Hive 记录
hive·sql·spark
FserSuN6 天前
Hive 分区表存储与查询优化学习总结
hive·hadoop·学习
步行cgn7 天前
为什么既要 web.xml 又要 springmvc.xml?能不能只配一个?
xml·前端·hive
@小匠8 天前
Flink + Iceberg 批量失败复盘:一次超时如何变成 Hive Metastore 锁泄漏
大数据·hive·flink
skywalk81639 天前
hive —— 多机 AI Agent 协同的任务总线:分布式,多节点同时运行,保障不会宕机
人工智能·hive·分布式