Question

我的表是这样的：

CREATE TABLE USER_TRANSACTIONS (
  START_TIME        BIGINT UNSIGNED  NOT NULL,
  APPLICATION_ID    CHAR(64) BINARY  NOT NULL,
  ENTRY_POINT       CHAR(255) BINARY NOT NULL,
  USER_ID           CHAR(64) BINARY  NOT NULL,
  ERROR_VIOLATION   BIT(1) NOT NULL,
  LATENCY_VIOLATION BIT(1) NOT NULL,
  PRIMARY KEY (START_TIME, APPLICATION_ID, ENTRY_POINT, USER_ID)
)

我想要实现的总结如下：每个entry_point我想看看有多少独特用户以及有多少用户有错误和延迟问题。

例如：

ENTRY_POINT | TOTAL_USERS | TOTAL_ERRORS | TOTAL_LATENCY
page1       | 2           | 2            | 1
page2       | 1           | 1            | 1

我可以通过此查询实现此目标：

SELECT UT.ENTRY_POINT, COUNT(USER_ID) AS TOTAL_USERS, SUM(EXP_ERRORS) AS TOTAL_ERRORS, SUM(EXP_LATENCY) AS TOTAL_LATENCY
FROM (
    SELECT ENTRY_POINT, USER_ID,
    BIT_OR(ERROR_VIOLATION) AS EXP_ERRORS,
    BIT_OR(LATENCY_VIOLATION) AS EXP_LATENCY
    FROM user_transactions
    GROUP BY ENTRY_POINT, USER_ID
) AS UT
GROUP BY UT.ENTRY_POINT;

嵌套查询用于总结用户是否遇到过错误或延迟问题，但在包含大量数据的表上我遇到了性能问题。

我的问题是如何优化此查询以避免使用内部子查询？

Answer 1

使用count(distinct)。以下是编写查询的一种方法：

SELECT ENTRY_POINT, COUNT(DISTINCT USER_ID),
       SUM(ERROR_VIOLATION > 0) AS TOTAL_ERRORS,
       SUM(LATENCY_VIOLATION > 0) AS TOTAL_LATENCY
FROM user_transactions
GROUP BY ENTRY_POINT;

如果您希望用户有错误而不是总错误：

SELECT ENTRY_POINT, COUNT(DISTINCT USER_ID),
       COUNT(DISTINCT CASE WHEN ERROR_VIOLATION > 0 THEN USER_ID END) AS TOTAL_ERRORS,
       COUNT(DISTINCT CASE WHEN LATENCY_VIOLATION > 0 THEN USER_ID END) AS TOTAL_LATENCY
FROM user_transactions
GROUP BY ENTRY_POINT;

Answer 2

你不能只使用这样的东西：

SELECT
    ENTRY_POINT
        ,COUNT(USER_ID) AS TOTAL_USERS
        ,SUM(EXP_ERRORS) AS TOTAL_ERRORS
        ,SUM(EXP_LATENCY) AS TOTAL_LATENCY
    FROM user_transactions
    GROUP BY ENTRY_POINT

如何在没有嵌套查询的情况下在MySQL中执行组聚合？

2 个答案: