mysql:用户自定义变量关联失效

自定义变量的属性和限制

1. 使用自定义变量的查询，无法使用查询缓存。
2. 不能在使用常量或者标识列的地方使用自定义变量，例如表名、列明和LIMIT子句中。
3. 用户自定义变量的生命周期是在一个连接中有效，所以不能使用它们来做连接间的通信。
4. 如果使用连接池或者持久化连接，自定义变量可能让看起来毫无关系的代码发生交互。
5. 在5.0版本之前，是大小写敏感的，所以要注意代码在不同版本之间的兼容性问题。
6. 不能显示的声明自定义变量的类型。它是一个动态类型。整数初始化为0，浮点型初始化为0.0，字符串初始化为’’。
7. MySQL优化器在某些场景下可能会将这些变量优化掉，这可能导致代码不按预想的方式运行。
8. 赋值的顺序和赋值的时间点并不总是固定的，这依赖于优化器的决定。
9. 赋值符号:=的优先级非常低，所有要注意赋值表达式应该使用明确的括号。
10. 使用未定义变量不会产生任何错误，如果没有意识到这一点，非常容易犯错。

SET @applykey = 0, @loanRank = 0;

SELECT applyId, custId, ruleVersion, rejectRule, STATUS
    , extra, createTime, updateTime, mobile
    , IF(STATUS = 1, IF(@applykey <> custId, @loanRank := 1, @loanRank := @loanRank + 1), '') AS applyPassRank
    , @applykey := custId
FROM (
    SELECT applyId, custId, ruleVersion, rejectRule, STATUS
        , extra, createTime, updateTime, mobile
    FROM sync.credit_apply
    WHERE updateTime < '2019-11-27'
    ORDER BY custId, updateTime
) t;

SET @cfkkey = 0, @cfkRank = 0;

SELECT cust_id, loan_amt_quota
    , IF(@cfkkey <> cust_id, @cfkRank := 1, @cfkRank := @cfkRank + 1) AS cfkRank
    , @cfkkey := cust_id
FROM (
    SELECT cust_id, loan_amt AS loan_amt_quota
    FROM sync.loandb_cfk_loan
    ORDER BY cust_id, create_time
) t

SET @applykey = 0, @loanRank = 0;
SET @cfkkey = 0, @cfkRank = 0;

CREATE TABLE test2
AS
SELECT applyId, custId, ruleVersion, rejectRule, STATUS
    , extra, loan_amt_quota
    , IF(first_finish_loan_Time IS NULL, 1, IF(updateTime < first_finish_loan_Time, 1, 0)) AS is_new
    , createTime, updateTime, mobile, applyPassRank, cfk_Rank
FROM (
    SELECT applyId, custId, ruleVersion, rejectRule, STATUS
        , extra, createTime, updateTime, mobile
        , IF(STATUS = 1, IF(@applykey <> custId, @loanRank := 1, @loanRank := @loanRank + 1), '') AS applyPassRank
        , @applykey := custId
    FROM (
        SELECT applyId, custId, ruleVersion, rejectRule, STATUS
            , extra, createTime, updateTime, mobile
        FROM sync.credit_apply
        WHERE updateTime < '2019-11-27'
        ORDER BY custId, updateTime
    ) t
) apply
    LEFT JOIN (
        SELECT cust_id
            , CASE
                WHEN locate('|', updateTime) > 0 THEN SUBSTR(updateTime, 1, INSTR(updateTime, '|') - 1)
                ELSE updateTime
            END AS first_finish_loan_Time
        FROM (
            SELECT cust_id, GROUP_CONCAT(update_Time ORDER BY update_Time ASC SEPARATOR '|') AS updateTime
            FROM sync.loandb_cfk_loan
            WHERE (substr(update_Time, 1, 10) < '2019-11-27'
                AND loan_type = 'xxx'
                AND loan_status = 'yyy')
            GROUP BY cust_id
        ) t
    ) cfk_loan
    ON apply.custId = cfk_loan.cust_id
    LEFT JOIN (
        SELECT cust_id, loan_amt_quota
            , IF(@cfkkey <> cust_id, @cfkRank := 1, @cfkRank := @cfkRank + 1) AS cfk_Rank
            , @cfkkey := cust_id
        FROM (
            SELECT cust_id, loan_amt AS loan_amt_quota
            FROM sync.loandb_cfk_loan
            ORDER BY cust_id, create_time
        ) t
    ) quota
    ON apply.custId = quota.cust_id
        AND apply.applyPassRank = quota.cfk_Rank;

后来发现不是这个问题而是多加上一个排序字段就OK了：

SET @applykey := ,
@loanRank = ;

SELECT applyId, custId, ruleVersion, rejectRule, STATUS
        , extra, createTime, updateTime, mobile
        , IF(STATUS = , IF(@applykey <> custId, @loanRank := , @loanRank := @loanRank + ), '') AS applyPassRank
        , @applykey := custId
FROM (
        SELECT applyId, custId, ruleVersion, rejectRule, STATUS
                , extra, createTime, updateTime, mobile
        FROM sync.credit_apply
        WHERE updateTime < '2019-11-28'
        ORDER BY custId,STATUS, updateTime
) t HAVING custId=xxx33089