mysql - 查询时间太长

Question

我有以下查询:

SELECT *
FROM document d
INNER JOIN users_documents ud ON d.id = ud.document_id 
WHERE (d.document_state_id=2
       AND ud.sharing_type_id=1
       AND ud.user_id=:id
       AND CURRENT_TIMESTAMP() BETWEEN d.upload_datetime AND d.approval_end_time
       AND ud.approval=2)
OR(d.document_state_id=1
   AND ud.sharing_type_id=2
   AND ud.user_id=:id
   AND CURRENT_TIMESTAMP() BETWEEN d.active_start_time AND d.active_end_time 
   AND ud.approval=2)
ORDER BY ud.application_date;

问题是，它需要大约 800 毫秒。当我将查询结尾更改为:

ORDER BY ud.uploadDatetime;

我添加了索引:upload_datetime(表文档)，现在查询需要大约 70 毫秒。问题是我真的需要根据 ud.application_date 对数据进行排序。 我尝试添加索引:application_date(在表 users_documents 上)，但它没有帮助(仍然需要 800 毫秒)。这次应该用哪个指标来降低呢？或者我应该做什么？我正在使用 MySQL。

非常感谢您的帮助!

编辑: 查询说明:查询返回具有某个 id 的给定用户的新文档。它返回“批准”(sharing_type=1)和阅读(sharing_type=2)的文档，approval=2(用户尚未看到该文档)，approval=1(用户批准了该文档)，0=未批准； upload_datetime=文档上传到系统的时间戳； applyr_end_time=文档在此日期之前可以被用户批准； activeStartTime=批准后，文档将从该日期到日期(=activeEndTime)向其他用户显示。 结构

-- -----------------------------------------------------
-- Table `po_db`.`document_state`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `po_db`.`document_state` (
  `id` BIGINT(20) NOT NULL,
  `state` VARCHAR(255) NULL DEFAULT NULL,
  PRIMARY KEY (`id`))
ENGINE = InnoDB
DEFAULT CHARACTER SET = utf8mb4
COLLATE = utf8mb4_0900_ai_ci;

-- -----------------------------------------------------
-- Table `po_db`.`document`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `po_db`.`document` (
  `id` BIGINT(20) NOT NULL,
  `active_end_time` DATETIME NOT NULL,
  `active_start_time` DATETIME NOT NULL,
  `approval_end_time` DATETIME NOT NULL,
  `description` VARCHAR(255) NOT NULL,
  `name` VARCHAR(60) NOT NULL,
  `resource_path` VARCHAR(255) NOT NULL,
  `title` VARCHAR(15) NOT NULL,
  `upload_datetime` DATETIME NOT NULL,
  `document_state_id` BIGINT(20) NOT NULL,
  `user_id` BIGINT(20) NULL DEFAULT NULL,
  PRIMARY KEY (`id`),
  UNIQUE INDEX `UK_36vs45u76s1n950kwxfa5lyhc` (`name` ASC),
  INDEX `FK1rwpvwxw2bvldt30kwfcbf57l` (`document_state_id` ASC),
  INDEX `FKjhdxdv9sijhujiynqbb5jc010` (`user_id` ASC),
  INDEX `title` (`title` ASC),
  INDEX `upload` USING BTREE (`upload_datetime`),
  CONSTRAINT `FK1rwpvwxw2bvldt30kwfcbf57l`
    FOREIGN KEY (`document_state_id`)
    REFERENCES `po_db`.`document_state` (`id`),
  CONSTRAINT `FKjhdxdv9sijhujiynqbb5jc010`
    FOREIGN KEY (`user_id`)
    REFERENCES `po_db`.`user` (`id`))
ENGINE = InnoDB
DEFAULT CHARACTER SET = utf8mb4
COLLATE = utf8mb4_0900_ai_ci;


-- -----------------------------------------------------
-- Table `po_db`.`sharing_type`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `po_db`.`sharing_type` (
  `id` BIGINT(20) NOT NULL,
  `name` VARCHAR(255) NULL DEFAULT NULL,
  PRIMARY KEY (`id`))
ENGINE = InnoDB
DEFAULT CHARACTER SET = utf8mb4
COLLATE = utf8mb4_0900_ai_ci;

和

-- -----------------------------------------------------
-- Table `po_db`.`users_documents`
-- -----------------------------------------------------
CREATE TABLE IF NOT EXISTS `po_db`.`users_documents` (
  `document_id` BIGINT(20) NOT NULL,
  `sharing_type_id` BIGINT(20) NOT NULL,
  `user_id` BIGINT(20) NOT NULL,
  `application_date` DATETIME NOT NULL,
  `approval` INT(11) NOT NULL,
  `email_sent` BIT(1) NOT NULL,
  PRIMARY KEY (`document_id`, `sharing_type_id`, `user_id`),
  INDEX `FK4mt7odsubst5269c4djnj4jip` (`sharing_type_id` ASC),
  INDEX `FKtn1i27tltfpy3n4pe306vtwu9` (`user_id` ASC),
  INDEX `historie_dok` (`approval` ASC, `sharing_type_id` ASC, `user_id` ASC, `application_date` ASC),
  CONSTRAINT `FK4mt7odsubst5269c4djnj4jip`
    FOREIGN KEY (`sharing_type_id`)
    REFERENCES `po_db`.`sharing_type` (`id`),
  CONSTRAINT `FK6abpoybb0f2ydy6ufla1wo80x`
    FOREIGN KEY (`document_id`)
    REFERENCES `po_db`.`document` (`id`),
  CONSTRAINT `FKtn1i27tltfpy3n4pe306vtwu9`
    FOREIGN KEY (`user_id`)
    REFERENCES `po_db`.`user` (`id`))
ENGINE = InnoDB
DEFAULT CHARACTER SET = utf8mb4
COLLATE = utf8mb4_0900_ai_ci;

编辑2: 我将查询的结尾更改为:

ORDER BY CASE WHEN d.document_state_id=2 THEN d.upload_datetime ELSE d.active_start_time END;

它与 application_date 基本相同，但仍然很慢(300 毫秒，但比 800 毫秒要好)。

我尝试添加索引(表文档):document_state_id、upload_datetime、active_start_time

但它再次没有任何帮助......:/

Answer 1

A 计划:将 OR 转换为 UNION:

    ( SELECT  *
            FROM  document d
            INNER JOIN  users_documents ud  ON d.id = ud.document_id
            WHERE  d.document_state_id=2
              AND  ud.sharing_type_id=1
              AND  ud.user_id=:id
              AND  CURRENT_TIMESTAMP() BETWEEN d.upload_datetime
                                           AND d.approval_end_time
              AND  ud.approval=2
    )
    UNION ALL
    ( SELECT  *
            FROM  document d
            INNER JOIN  users_documents ud  ON d.id = ud.document_id
            WHERE  d.document_state_id=1
              AND  ud.sharing_type_id=2
              AND  ud.user_id=:id
              AND  CURRENT_TIMESTAMP() BETWEEN d.active_start_time
                                           AND d.active_end_time
              AND  ud.approval=2 
    )
    ORDER BY  application_date;

这样，每个 SELECT 都可以充分利用以approval ASC、sharing_type_id ASC、user_id ASC 开头的索引。 (第四列没有帮助。)

B 计划:使用“覆盖索引”:

SELECT d2.*, ud2.*
  FROM (
    ( SELECT  id, document_id, sharing_type_id, user_id
            FROM  document d
            INNER JOIN  users_documents ud  ON d.id = ud.document_id
            WHERE  d.document_state_id=2
              AND  ud.sharing_type_id=1
              AND  ud.user_id=:id
              AND  CURRENT_TIMESTAMP() BETWEEN d.upload_datetime 
                                           AND d.approval_end_time
              AND  ud.approval=2
    )
    UNION ALL
    ( SELECT  id, document_id, sharing_type_id, user_id
            FROM  document d
            INNER JOIN  users_documents ud  ON d.id = ud.document_id
            WHERE  d.document_state_id=1
              AND  ud.sharing_type_id=2
              AND  ud.user_id=:id
              AND  CURRENT_TIMESTAMP() BETWEEN d.active_start_time
                                           AND d.active_end_time
              AND  ud.approval=2 
    )
       )
  JOIN document d2  USING(id)
  JOIN users_documents ud2  USING(document_id, sharing_type_id, user_id)
  ORDER BY  ud2.application_date;

加上这些新索引；顺序很重要:

UD: INDEX(approval, sharing_type_id, user_id, document_id)  -- document_id last
D: INDEX(document_state_id, approval_end_time, approval_start_time, document_id)
D: INDEX(document_state_id, active_end_time, active_start_time, document_id)