多案例声明无法在Postgres中正常工作

Question

这是查询：

SELECT DISTINCT 
         completed_phases,
         CAST(completed_phases::bit(8) AS VARCHAR),
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=1 THEN 'FT' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=2 THEN 'ED' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=3 THEN 'MC' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=4 THEN 'HC' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=5 THEN 'UV' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=6 THEN 'TT' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=7 THEN 'RX' ELSE '' END ||
         CASE WHEN STRPOS(CAST(completed_phases::bit(8) AS VARCHAR),'1')=8 THEN 'PI' ELSE '' END
FROM rx_sales_order

如果completed_phase为129 ，我最后一栏的输出应为FTPI。 但这仅显示FT。 即使所有情况都是不同的，似乎只有第一个case语句仍然有效。

Answer 1

STRPOS()将始终返回搜索到的字符串的第一个匹配项。 因此，对strpos()所有调用都将为输入值129返回1。

您可以改用substring() ：

CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),1,1)='1' THEN 'FT' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),2,1)='1' THEN 'ED' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),3,1)='1' THEN 'MC' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),4,1)='1' THEN 'HC' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),5,1)='1' THEN 'UV' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),6,1)='1' THEN 'TT' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),7,1)='1' THEN 'RX' ELSE '' END ||
CASE WHEN substring(CAST(completed_phases::bit(8) AS VARCHAR),8,1)='1' THEN 'PI' ELSE '' END

另一个选择是使用get_bit()分别测试每个位：

case when get_bit(completed_phases::bit(8), 0) = 1 then 'FT' else '' END||
case when get_bit(completed_phases::bit(8), 1) = 1 then 'ED' else '' END||
case when get_bit(completed_phases::bit(8), 2) = 1 then 'MC' else '' END||
case when get_bit(completed_phases::bit(8), 3) = 1 then 'HC' else '' END||
case when get_bit(completed_phases::bit(8), 4) = 1 then 'UV' else '' END||
case when get_bit(completed_phases::bit(8), 5) = 1 then 'TT' else '' END||
case when get_bit(completed_phases::bit(8), 6) = 1 then 'RX' else '' END||
case when get_bit(completed_phases::bit(8), 7) = 1 then 'PI' else '' END

---

一种更灵活的方法是将这些位转换为行并将数组用作查找。 就像是：

with lookup (codes) as (
   values (array['FT','ED','MC','HC','UV','TT','RX','PI'])
)
SELECT  completed_phases, 
        completed_phases::bit(8),
        x.code
FROM rx_sales_order
  join lateral (
    select string_agg(codes[i],'') as code
    from lookup, unnest(string_to_array(completed_phases::bit(8)::text, null)) with ordinality as t(b,i) 
    where b = '1'
 ) as x on true

regexp_split_to_table(completed_phases::bit(8)::text, '') with ordinality as t(b,i)将返回以下值129：

b | i
--+--
1 | 1
0 | 2
0 | 3
0 | 4
0 | 5
0 | 6
0 | 7
1 | 8

code[i]使用索引查找匹配的代码，然后string_agg()将所有选定的代码再次放到一个字符串中。 where b = '1'的条件仅选择设置的位。

该解决方案将比硬编码的case表达式慢得多（因为它增加了行数，只是为了再次减少行数）-但它更加灵活并且易于维护。

---

如果您需要大量资源，最好的选择是将case表达式放入函数中并在查询中使用该函数。

create or replace function get_codes(p_phases integer)
  returns text
as 
$$
  select
    case when get_bit(p_phases::bit(8), 0) = 1 then 'FT' else '' END||
    case when get_bit(p_phases::bit(8), 1) = 1 then 'ED' else '' END||
    case when get_bit(p_phases::bit(8), 2) = 1 then 'MC' else '' END||
    case when get_bit(p_phases::bit(8), 3) = 1 then 'HC' else '' END||
    case when get_bit(p_phases::bit(8), 4) = 1 then 'UV' else '' END||
    case when get_bit(p_phases::bit(8), 5) = 1 then 'TT' else '' END||
    case when get_bit(p_phases::bit(8), 6) = 1 then 'RX' else '' END||
    case when get_bit(p_phases::bit(8), 7) = 1 then 'PI' else '' END
$$
language sql;

然后使用：

SELECT DISTINCT 
         completed_phases,
         get_codes(completed_phases) as codes
FROM rx_sales_order

Answer 2

正如a_horse_with_no_name答案中指出的那样 ， strpos将返回搜索到的字符串的第一个匹配项。 无论如何，最好使用get_bit而不是强制转换为VARCHAR来检查是否设置了位。 另外，代替|| ，您可以使用concat ，它将处理null s作为空白字符串。 您的查询随后可以重写为：

SELECT DISTINCT 
     completed_phases,
     CAST(completed_phases::bit(8) AS VARCHAR),
     concat(CASE get_bit(completed_phases::bit(8), 0) WHEN 1 THEN 'FT' END,
            CASE get_bit(completed_phases::bit(8), 1) WHEN 1 THEN 'ED' END,
            CASE get_bit(completed_phases::bit(8), 2) WHEN 1 THEN 'MC' END,
            CASE get_bit(completed_phases::bit(8), 3) WHEN 1 THEN 'HC' END,
            CASE get_bit(completed_phases::bit(8), 4) WHEN 1 THEN 'UV' END,
            CASE get_bit(completed_phases::bit(8), 5) WHEN 1 THEN 'TT' END,
            CASE get_bit(completed_phases::bit(8), 6) WHEN 1 THEN 'RX' END,
            CASE get_bit(completed_phases::bit(8), 7) WHEN 1 THEN 'PI' END)
FROM rx_sales_order;

附带说明一下，如果您可以选择这样做，建议您更改数据库架构，以将阶段存储为单独的boolean列，而不要使用位图。 请参见数据库列中位标志的任何缺点？ 很好地讨论了为什么。