Group By with Case 语句不计算零

Question

我有以下 SQL：

WITH MYFILTER AS (
SELECT
CASE WHEN STATE IN ('A', 'B') then
case
   when (NUM/DENOM - 1.0) < -0.3 then 'Less than -30%'
           when (NUM/DENOM - 1.0) >= -0.3
           and (NUM/DENOM - 1.0) < -0.2  then '-30% to -20%'
           when (NUM/DENOM - 1.0) >= -0.2
           and (NUM/DENOM - 1.0) < -0.1  then '-20% to -10%'
           when (NUM/DENOM - 1.0) >= -0.1
           and (NUM/DENOM - 1.0) < 0.0   then '-10% to 0%'
           when (NUM/DENOM - 1.0) >= 0.0
           and (NUM/DENOM - 1.0) < 0.1    then '0% to 10%'
           when (NUM/DENOM - 1.0) >= 0.1
           and (NUM/DENOM - 1.0) < 0.2    then '10% to 20%'
           when (NUM/DENOM - 1.0) >= 0.2
           and (NUM/DENOM - 1.0) < 0.3    then '20% to 30%'
           when (NUM/DENOM - 1.0) >= 0.3  THEN 'At least 30%'
           end
ELSE case
   when (NUM/DENOM < -0.3 then 'Less than -30%'
           when (NUM/DENOM >= -0.3
           and (NUM/DENOM < -0.2  then '-30% to -20%'
           when (NUM/DENOM >= -0.2
           and (NUM/DENOM < -0.1  then '-20% to -10%'
           when (NUM/DENOM >= -0.1
           and (NUM/DENOM < 0.0   then '-10% to 0%'
           when (NUM/DENOM >= 0.0
           and (NUM/DENOM < 0.1    then '0% to 10%'
           when (NUM/DENOM >= 0.1
           and (NUM/DENOM < 0.2    then '10% to 20%'
           when (NUM/DENOM >= 0.2
           and (NUM/DENOM < 0.3    then '20% to 30%'
           when (NUM/DENOM >= 0.3  THEN 'At least 30%'
           end
END AS indrange
FROM MYTABLE
WHERE DENOM <> 0 AND
YEAR = 2020 AND
MONTH = 11
)
SELECT
indrange,
count (*) AS total
FROM FILTER
GROUP BY indrange

只要比率不属于某个范围（例如，我的表中没有 NUM/DENOM - 1 > 0.3 的行），那么我最终得到的结果不包括“至少 30% " 作为具有 0 值的行。相反，该行根本不存在。我将如何更改代码以使其仍包含具有相应 0 值的“至少 30%”行？换句话说，我明白了：

INDRANGE        TOTAL
Less than -30%  285
-30% to -20%    1,608
-20% to -10%    7,409
-10% to 0%      164,212
0% to 10%       169,665
10% to 20%      1

但我想要这个：

INDRANGE        TOTAL
Less than -30%  285
-30% to -20%    1,608
-20% to -10%    7,409
-10% to 0%      164,212
0% to 10%       169,665
10% to 20%      1
20% to 30%      0
At Least 30%    0

我该怎么做？这是 DB2。

Answer 1

您可以在范围和现有 CTE 之间使用 LEFT JOIN。例如：

with myranges (indrange) as (
  select 'Less than -30%'         from sysibm.sysdummy1
  union all select '-30% to -20%' from sysibm.sysdummy1
  union all select '-20% to -10%' from sysibm.sysdummy1
  union all select '-10% to 0%'   from sysibm.sysdummy1
  union all select '0% to 10%'    from sysibm.sysdummy1
  union all select '10% to 20%'   from sysibm.sysdummy1
  union all select '20% to 30%'   from sysibm.sysdummy1
  union all select 'At Least 30%' from sysibm.sysdummy1
),
myfilter as (
  -- here add you existing CTE
)
select r.indrange, count(f.indrange) AS total
from myranges r
left join myfilter f on f.indrange = r.indrange
group by r.indrange

Answer 2

未经测试：但基于我之前的 cmets 的类似内容。

• 添加了一个 MyRanges cte 列出每个范围 1 次，从而确保每个范围都包含在最终查询中。
• 在最终查询中使用 Union all 将其与来自 MyFilter 的结果结合起来
• 从 count(*) 结果中减去 1，因为我们将 myRanges union 的结果增加了 1

.

WITH  MYRANGES AS(
  SELECT 'Less than -30%' as INDRANGE FROM DUAL UNION ALL
  SELECT '-30% to -20%'    as INDRANGE FROM DUAL UNION ALL
  SELECT '-20% to -10%'    as INDRANGE FROM DUAL UNION ALL
  SELECT '-10% to 0%'      as INDRANGE FROM DUAL UNION ALL
  SELECT '0% to 10%'       as INDRANGE FROM DUAL UNION ALL
  SELECT '10% to 20%'      as INDRANGE FROM DUAL UNION ALL
  SELECT '20% to 30%'      as INDRANGE FROM DUAL UNION ALL
  SELECT 'At Least 30%'    as INDRANGE FROM DUAL),
MYFILTER AS (SELECT
  CASE WHEN STATE IN ('A', 'B') then
  case
   when (NUM/DENOM - 1.0) < -0.3 then 'Less than -30%'
           when (NUM/DENOM - 1.0) >= -0.3
           and (NUM/DENOM - 1.0) < -0.2  then '-30% to -20%'
           when (NUM/DENOM - 1.0) >= -0.2
           and (NUM/DENOM - 1.0) < -0.1  then '-20% to -10%'
           when (NUM/DENOM - 1.0) >= -0.1
           and (NUM/DENOM - 1.0) < 0.0   then '-10% to 0%'
           when (NUM/DENOM - 1.0) >= 0.0
           and (NUM/DENOM - 1.0) < 0.1    then '0% to 10%'
           when (NUM/DENOM - 1.0) >= 0.1
           and (NUM/DENOM - 1.0) < 0.2    then '10% to 20%'
           when (NUM/DENOM - 1.0) >= 0.2
           and (NUM/DENOM - 1.0) < 0.3    then '20% to 30%'
           when (NUM/DENOM - 1.0) >= 0.3  THEN 'At least 30%'
           end
ELSE case
   when (NUM/DENOM < -0.3 then 'Less than -30%'
           when (NUM/DENOM >= -0.3
           and (NUM/DENOM < -0.2  then '-30% to -20%'
           when (NUM/DENOM >= -0.2
           and (NUM/DENOM < -0.1  then '-20% to -10%'
           when (NUM/DENOM >= -0.1
           and (NUM/DENOM < 0.0   then '-10% to 0%'
           when (NUM/DENOM >= 0.0
           and (NUM/DENOM < 0.1    then '0% to 10%'
           when (NUM/DENOM >= 0.1
           and (NUM/DENOM < 0.2    then '10% to 20%'
           when (NUM/DENOM >= 0.2
           and (NUM/DENOM < 0.3    then '20% to 30%'
           when (NUM/DENOM >= 0.3  THEN 'At least 30%'
           end
END AS indrange
FROM MYTABLE
WHERE DENOM <> 0 AND
  YEAR = 2020 AND
  MONTH = 11
),

SELECT indrange, count (*)-1 AS total
FROM   (SELECT * FROM MyFilter UNION ALL
        SELECT * FROM MyRanges) as MyFilterandAllRanges
GROUP BY indrange

好吧，既然我已经做了一些事情，我已经考虑过了，我可以做得更好......

所以将最终查询修改为....，我们不必弄乱数字和数学。我们基本上采用范围的“主列表”并将其外部连接到我们的总数中，以便始终包含所有范围，并且只有匹配的范围才会显示总数。我们使用合并来处理一些可能不存在的事实并替换所需的 0。

SELECT MyRanges.indrange, coalesce(total,0) as total
FROM   MyRanges
LEFT JOIN (SELECT count(*) total, Indrange 
           FROM MyFilter
           GROUP BY indrange) as sub
     on MyRanges.indrange = sub.indrange