带有ORDER BY的HIVE GROUP_CONCAT

Question

我有一张像

这样的表格

我希望输出像这样（group将结果连接到一个记录中，group_concat应该按值DESC对结果进行排序）。

这是我试过的查询，

SELECT id,
       CONCAT('{',CONCAT_WS(',',GROUP_CONCAT(CONCAT('"',key, '":"',value, '"'))), '}') AS value
FROM
    table_name
GROUP BY id

我希望目标表中的值应按源表值排序（降序）。

为此，我尝试了 GROUP_CONCAT（... ORDER BY值）。

看起来Hive不支持此功能。有什么其他方法可以在蜂巢中实现这一目标吗？

Answer 1

试试这个查询。

Hive不支持GROUP_CONCAT函数，但您可以使用collect_list函数来实现类似的功能。此外，您将需要使用分析窗口函数，因为Hive不支持collect_list函数内的ORDER BY子句

select
  id,
  -- since we have a duplicate group_concat values against the same key
  -- we can pick any one value by using the min() function
  -- and grouping by the key 'id'
  -- Finally, we can use the concat and concat_ws functions to 
  -- add the commas and the open/close braces for the json object
  concat('{', concat_ws(',', min(g)), '}') 
from
  (
    select
      s.id,
      -- The window function collect_list is run against each row with 
      -- the partition key of 'id'. This will create a value which is 
      -- similar to the value obtained for group_concat, but this 
      -- same/duplicate value will be appended for each row for the 
      -- same key 'id'
      collect_list(s.c) over (partition by s.id 
        order by s.v desc 
      rows between unbounded preceding and unbounded following) g 
    from
      (
        -- First, form the key/value pairs from the original table.
        -- Also, bring along the value column 'v', so that we can use
        -- it further for ordering
        select
          id,
          v,
          concat('"', k, '":"', v, '"') as c 
        from
          table_name -- This it th
      )
      s
  )
  gs 
-- Need to group by 'id' since we have duplicate collect_list values
group by
  id