PostgreSQL 中别名列的总和

Question

我有一个聚合三个结果列的查询：

pending_with_documents
pending_without_documents
inprocessing

SELECT COUNT(DISTINCT (CASE
                           WHEN status_code IN ('ACK', 'INT') THEN
                               CASE
                                   WHEN ack_no IN
                                        (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
    END))        AS pending_with_documents,
       COUNT(DISTINCT (CASE
                           WHEN status_code IN ('ACK', 'INT') THEN
                               CASE
                                   WHEN ack_no NOT IN
                                        (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
           END)) AS pending_without_documents,
       COUNT(CASE
                 WHEN status_code NOT IN ('ACK', 'INT', 'APR', 'REJ', 'SBK', 'OBJ') THEN
                     ack_no
           END)  AS Inprocessing
FROM application_ht_install
WHERE service_code IN (36);

现在我需要一个第四列total来总结这三个。所以我尝试了：

SELECT COUNT(DISTINCT (CASE
                           WHEN status_code IN ('ACK', 'INT') THEN
                               CASE
                                   WHEN ack_no IN
                                        (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
    END))                             AS pending_with_documents,
       COUNT(DISTINCT (CASE
                           WHEN status_code IN ('ACK', 'INT') THEN
                               CASE
                                   WHEN ack_no NOT IN
                                        (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
           END))                      AS pending_without_documents,
       COUNT(CASE
                 WHEN status_code NOT IN ('ACK', 'INT', 'APR', 'REJ', 'SBK', 'OBJ')
                     THEN ack_no END) AS Inprocessing
    SUM((COUNT(DISTINCT(CASE WHEN status_code IN ('ACK','INT') THEN
    CASE WHEN ack_no IN
    (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
    END)) AS pending_with_documents )+ ( COUNT(DISTINCT(CASE WHEN status_code  IN ('ACK','INT') THEN
    CASE WHEN ack_no NOT IN
    (SELECT ack_no FROM bescom_appl_upload_doc) THEN ack_no END
    END)) AS pending_without_documents)+ ( COUNT(CASE WHEN status_code NOT IN('ACK','INT','APR','REJ','SBK','OBJ') THEN
    ack_no
    END) AS Inprocessing))
FROM application_ht_install
WHERE service_code IN (36);

但我收到一个错误：

ERROR:  syntax error at or near "sum"
LINE 13: sum((count(distinct(case when status_code in ('ACK','INT') t...
              ^ SQL state: 42601 Character: 473

这三个派生列如何相加？

Answer 1

您的语句中有几个语法错误。您正在为不需要别名的事物添加别名。 这是纠正了语法错误的代码。注意：有一列没有别名，所以我称它为“do_not_know_name”，因为我不知道你的意思。

select 
count(distinct(case when status_code in ('ACK','INT') then 
case when ack_no in
(select ack_no from bescom_appl_upload_doc) then ack_no end 
end)) as pending_with_documents ,
count(distinct(case when status_code  in ('ACK','INT') then 
case when ack_no not in
(select ack_no from bescom_appl_upload_doc) then ack_no end 
end)) as pending_without_documents ,
count(case when status_code not in('ACK','INT','APR','REJ','SBK','OBJ') 
then ack_no end) as Inprocessing,
sum((count(distinct(case when status_code in ('ACK','INT') then 
case when ack_no in
(select ack_no from bescom_appl_upload_doc) then ack_no end 
end))  )+ ( count(distinct(case when status_code  in ('ACK','INT') then 
case when ack_no not in
(select ack_no from bescom_appl_upload_doc) then ack_no end 
end)) )+ ( count(case when status_code not in('ACK','INT','APR','REJ','SBK','OBJ') then 
ack_no
end) )) as do_not_know_name
from application_ht_install where service_code in (36) 

Answer 2

基本上，您可以重复冗长的表达式，或者将所有内容放在子查询中并在外部SELECT 中求和。我建议后者：

SELECT pending_with_documents
     , pending - pending_with_documents AS pending_without_documents
     , inprocessing
     , pending + inprocessing AS total
FROM  (
   SELECT count(DISTINCT a.ack_no) FILTER (WHERE status_code IN ('ACK', 'INT'))              AS pending
        , count(DISTINCT b.ack_no) FILTER (WHERE status_code IN ('ACK', 'INT'))              AS pending_with_documents
        , count(ack_no) FILTER (WHERE status_code <> ALL ('{ACK, INT, APR, REJ, SBK, OBJ}')) AS inprocessing
   FROM   application_ht_install a
   LEFT   JOIN (SELECT DISTINCT ack_no FROM bescom_appl_upload_doc) b USING (ack_no)
   WHERE  service_code = 36
   ) sub;

这个可能做你想做的事。取决于实际的表定义。如果任一表中的ack_no 或status_code 可以是NULL，您可能需要做更多。见：

• Select rows which are not present in other table

此外，子查询的LEFT JOIN 可能更有效，也可能不会更有效，具体取决于未公开的架构详细信息和基数。

<> ALL (<array>) 等价于 NOT IN (<list>)，但对于较长的列表来说更短，通常表现更好。当涉及 NULL 时，两者都失败...

关于更高效的FILTER 子句替换嵌套的CASE 表达式：

• Aggregate columns with additional (distinct) filters

注意在内部SELECT 列表中使用a.ack_no、b.ack_no 和ack_no。 a.ack_no 与 USING 子句左连接后实际上等同于 ack_no。我只是为了对称才拼出来的。

count(DISTINCT b.ack_no) 有效，因为 count() 不计算 NULL 值 - 如果 bescom_appl_upload_doc 中不存在 b.ack_no，您将获得这些值。