Question

今天刚刚测试了大型数据集。我的查询确实表现不佳，我怀疑是因为子查询。这是我的查询（MYSQL PDO：）

SELECT  a.data_id,a.data_name,a.data_parent_id,a.data_type,a.data_return_text,a.data_description,
IF ((length(a.data_id)<=1), 0,
CASE a.data_type
WHEN '0' 
THEN 
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '1' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '2' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '3' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '4' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '5' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '6' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '7' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '8' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '9' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '10' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '11' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '12' THEN 
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
WHEN '13' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '14' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND client_id=:client_id),0)
WHEN '15' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '0' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('1','0') AND (DATE(trans_date) <= DATE(NOW())) AND client_id=:client_id),0)
WHEN '16' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND (DATE(trans_date) <= DATE(NOW())) AND client_id=:client_id),0)
WHEN '17' THEN
IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)
END) as total 
FROM data_list a
LEFT JOIN data_transaction_log b
ON b.trans_data_id = a.data_id
WHERE a.client_id=:client_id
GROUP   BY a.data_name
ORDER BY a.data_id asc

[说明]

有表格：

表：DATA_LIST

包含大约24种仪器，仪器在使用时会创建一个事务日志。它将产生（值）及其（+ ve或-ve符号）及其日期戳

表：data_transaction_log

此表将记录data_list中仪器的每个结果。

table：data_list 的结果将在此表中记录：

（value）=＆gt; trans_value
（+ ve符号将记录为0＆amp; -ve符号将记录为1）=＆gt; TRANS_TYPE
（日期）=＆gt; TRANS_DATE

[意向]

我想得到+ ve＆amp;的总和 - 每种工具的每笔交易的价值：

IFNULL（（SELECT SUM（trans_value * case trans_type when'1'，然后是-1 else 1 end）FROM data_transaction_log WHERE trans_data_id = a.data_id和trans_type in（'0'，'1'）AND client_id =：client_id），0）

某些工具有特殊条件（案例15和16），需要日期条件。

IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND (DATE(trans_date) <= DATE(NOW())) AND client_id=:client_id),0)

[问题}

在低数据集大约100（在data_transaction_log中）它可以正常工作。但是超过1k它真的很慢！

请指导我。

谢谢！

Answer 1

对于data_list数据类型字段，多个值具有相同的精确结果。例如，WHEN a.data_type IN ('0', '1', '2', '3', '7', '9', '10', '12', '17')

你有相同的结果：

IFNULL((SELECT SUM(trans_value * case trans_type when '1' then -1 else 1 end) FROM data_transaction_log WHERE trans_data_id = a.data_id and trans_type in ('0','1') AND client_id=:client_id),0)

我认为它不会使代码更快，但编写的代码更少，更容易调试：

SELECT a.data_id,a.data_name,a.data_parent_id,a.data_type
      ,a.data_return_text,a.data_description
      ,CASE WHEN length(a.data_id)<=1) 
             THEN 0

            WHEN a.data_type IN ('0', '1', '2', '3', '7', '9', '10', '12', '17')
             THEN IFNULL((SELECT SUM(trans_value * case WHEN trans_type = '1' THEN -1 else 1 end) 
                            FROM data_transaction_log 
                           WHERE trans_data_id = a.data_id and trans_type in ('0','1') 
                             AND client_id=:client_id)
                           ,0)            

            WHEN a.data_type IN ('4', '5', '6', '8', '11', '13', '14')
             THEN IFNULL((SELECT SUM(trans_value * case WHEN trans_type = '0' then -1 else 1 end) 
                            FROM data_transaction_log 
                           WHERE trans_data_id = a.data_id and trans_type in ('0','1') 
                             AND client_id=:client_id)
                           ,0)

            WHEN a.data_type IN ('15', '16') 
             THEN IFNULL((SELECT SUM(trans_value * case WHEN trans_type = '0' then -1 else 1 end) 
                            FROM data_transaction_log 
                           WHERE trans_data_id = a.data_id and trans_type in ('0','1') 
                             AND (DATE(trans_date) <= DATE(NOW())) AND client_id=:client_id)
                           ,0)

       END as total 
  FROM data_list a LEFT JOIN data_transaction_log b ON b.trans_data_id = a.data_id
 WHERE a.client_id=:client_id
GROUP BY a.data_name
ORDER BY a.data_id asc

-

第二次采取：

为了解决这个问题，最好知道为什么要检查IFNULL。如果由于左连接返回NULL而连接条件不匹配，那么您可以通过检查左连接表中的主键data_transaction_log，再一步简单地算法一步，对于NULL ...主键不应为null，如果为null，则左键必须为NULL。在原始算法中，您给NULL结果值为0：

SELECT a.data_id,a.data_name,a.data_parent_id,a.data_type
      ,a.data_return_text,a.data_description
      ,CASE WHEN b.<PRIMARYKEY> IS NULL
             THEN 0
            WHEN length(a.data_id)<=1) 
             THEN 0    
            WHEN a.data_type IN ('0', '1', '2', '3', '7', '9', '10', '12', '17')
             THEN (SELECT SUM(trans_value * case WHEN trans_type = '1' THEN -1 else 1 end) 
                     FROM data_transaction_log 
                    WHERE trans_data_id = a.data_id and trans_type in ('0','1')                             AND client_id=:client_id)                      

            WHEN a.data_type IN ('4', '5', '6', '8', '11', '13', '14')
             THEN (SELECT SUM(trans_value * case WHEN trans_type = '0' then -1 else 1 end) 
                     FROM data_transaction_log 
                    WHERE trans_data_id = a.data_id and trans_type in ('0','1') 
                      AND client_id=:client_id)                   

            WHEN a.data_type IN ('15', '16') 
             THEN (SELECT SUM(trans_value * case WHEN trans_type = '0' then -1 else 1 end) 
                     FROM data_transaction_log 
                    WHERE trans_data_id = a.data_id and trans_type in ('0','1') 
                      AND (DATE(trans_date) <= DATE(NOW())) AND client_id=:client_id)                  

       END as total 
  FROM data_list a LEFT JOIN data_transaction_log b ON b.trans_data_id = a.data_id
 WHERE a.client_id=:client_id
GROUP BY a.data_name
ORDER BY a.data_id asc

Mysql查询真的很慢

1 个答案: