Hive 常用 SQL 备忘

last modify

Keywords: Hive

• 构造临时数据

• 集合操作

  • collect + sort + truncate

• References

构造临时数据

-- 法1
SELECT inline(array(
          struct('A', 10, date '2015-01-01')
        , struct('B', 20, date '2016-02-02')
    )) AS (col1, col2, col3);

-- 法2 (推荐)
SELECT stack(2 -- 表示下面有两条数据
        , 'A', 10, date '2015-01-01'
        , 'B', 20, date '2016-01-01'
    ) AS (col0, col1, col2);

集合操作

collect + sort + truncate

ADD JAR hdfs://path_to_brickhouse.jar;
CREATE TEMPORARY FUNCTION truncate_array AS 'brickhouse.udf.collect.TruncateArrayUDF';

-- truncate_array 截取数组时, 长度不够会补 NULL
SELECT pkey, IF(size(items_sort) > 5, truncate_array(items_sort, 5), items_sort) AS items_sort
FROM (
    SELECT pkey, sort_array(collect_set(struct(-score, item))).col2 AS items_sort
    FROM (
        SELECT pkey, item, score
        FROM ...
    ) A
    GROUP BY pkey
) A
;

References