合并 MySQL 查询答案

【问题标题】：Consolidating MySQL queries合并 MySQL 查询
【发布时间】：2014-08-19 06:58:49
【问题描述】：

我想知道如何将我使用的 4 个查询合并为一个查询。我能够查询我的数据集中表示的所有日期以及每个特定日期的点击次数。理想情况下，我还可以从该查询中获得安装次数（每天）、总成本（每天）和备注（每天）。

我已经考虑过如何实现这一点，但由于我是 MySQL 新手，我还没有提出任何完整的解决方案。有没有从多个表中提取数据的好方法？

$days = mysqli_query($link, "
    SELECT
      t1.date,
      t1.clicks
    FROM (SELECT
        date_format(date_sub(ic.click_utc, INTERVAL 7 HOUR), '%Y-%m-%d') as date,
        count(distinct ic.txid) as clicks
          FROM users_clicks ic
          GROUP BY date_format(date_sub(click_utc, INTERVAL 7 HOUR), '%Y-%m-%d')
          ORDER BY date DESC) t1
    WHERE date >= '2014-06-28'");
while ($day = mysqli_fetch_assoc($days)) {
    $date = $day['date'];
    $day_clicks = $day['clicks'];
    $day_installs = mysqli_fetch_row(mysqli_query($link, sprintf("SELECT count(txid) FROM (SELECT txid FROM users_installs WHERE date_format(date_sub(click_utc, INTERVAL 7 HOUR), '%%Y-%%m-%%d') = '%s' GROUP BY txid) table1", $day['date'])));
    $day_cost = mysqli_fetch_row(mysqli_query($link, sprintf("SELECT sum(earnings) FROM (SELECT max(cost) as earnings FROM users_clicks WHERE date_format(date_sub(click_utc, INTERVAL 7 HOUR), '%%Y-%%m-%%d') = '%s' GROUP BY txid) table1", $day['date'])));
    $note = mysqli_fetch_assoc(mysqli_query($link, sprintf("SELECT * FROM reporting_notes WHERE date_format(date_sub(timestamp, INTERVAL 7 HOUR), '%%Y-%%m-%%d') = '%s' LIMIT 1", $day['date'])));
}

以下是表格的结构：

用户点击次数：

   txid  |       click_utc        |  cost  
   12t3  |   2014-08-19 07:08:47  |  0.50
   27a5  |   2014-08-18 03:28:03  |  0.25
   48a5  |   2014-08-17 12:55:23  |  0.25

用户安装次数：

   txid  |       click_utc 
   1o23  |   2014-08-19 07:08:47
   1ee3  |   2014-08-17 11:10:53

Reporting_notes：

      timestamp        |  note  
  2014-08-19 07:08:47  |  "hey"
  2014-08-17 03:02:41  |  "hey"

期望的输出：数据集中每一天的点击次数、安装次数、总成本和任何注释

【问题讨论】：

这不会有帮助。请发布您的表格结构和所需的输出，然后我们将能够提供帮助。
没人愿意看到那个。正如 raheel 所说，让我们从头开始
我已经为每个涉及的表添加了结构

标签： php mysql sql mysqli

【解决方案1】：

首先：MySQL 能够处理命名时区，但您需要先填充时区信息表。请参阅 CONVERT_TZ 的文档作为起点。

由于您的表似乎不相关（即没有包含所有 txids 的主表，...）您可能最好使用在开始时接近触发单独查询的子查询。

点击次数和每日费用：

SELECT
  DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date,
  COUNT(*)                                        AS clicks,
  SUM(cost)                                       AS costs
FROM users_clicks
GROUP BY DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00'));

只计算每个 txid 一次

SELECT
  DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00')) AS date,
  COUNT(*)                                            AS clicks,
  SUM(max_costs)                                      AS costs
FROM (SELECT
        txid,
        MIN(click_utc) AS min_click_utc,
        MAX(cost)      AS max_costs
      FROM users_clicks
      GROUP BY txid) distinct_txids
GROUP BY DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00'))

每天安装次数：

SELECT
  DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date,
  COUNT(*)                                        AS installs
FROM users_installs
GROUP BY DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00'));

只计算每个 txid 一次

SELECT
  DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00')) AS date,
  COUNT(*)                                            AS installs
FROM (SELECT
        txid,
        MIN(click_utc) AS min_click_utc
      FROM users_installs
      GROUP BY txid) distinct_txids
GROUP BY DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00'));

每日笔记

（注意GROUP_CONCAT 默认只返回 1024 个字符。这可以通过将group_concat_max_len 设置为更高的值来更改）：

SELECT
  DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date,
  COUNT(*)                                        AS note_count,
  GROUP_CONCAT(note SEPARATOR ', ')               AS notes
FROM reporting_notes
GROUP BY DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00'));

如果你想一次获得所有东西，你需要先获得一个公共日期池因为 MySQL 只有LEFT 或RIGHT OUTER JOINs，但没有FULL OUTER JOINs：

SELECT DISTINCT
  date
FROM (
       SELECT
         DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
       FROM users_clicks
       UNION SELECT
               DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
             FROM users_installs
       UNION SELECT
               DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date
             FROM reporting_notes) dates;

然后你可以将它们全部组合成

一个查询

SELECT
  dates.date,
  clicks.clicks,
  clicks.costs,
  installs.installs,
  notes.note_count,
  notes.notes
FROM (SELECT DISTINCT
        date
      FROM (
             SELECT
               DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
             FROM users_clicks
             UNION SELECT
                     DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
                   FROM users_installs
             UNION SELECT
                     DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date
                   FROM reporting_notes) data) dates
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date,
               COUNT(*)                                        AS clicks,
               SUM(cost)                                       AS costs
             FROM users_clicks
             GROUP BY DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00'))) clicks ON clicks.date = dates.date
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date,
               COUNT(*)                                        AS installs
             FROM users_installs
             GROUP BY DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00'))) installs ON installs.date = dates.date
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date,
               COUNT(*)                                        AS note_count,
               GROUP_CONCAT(note SEPARATOR ', ')               AS notes
             FROM reporting_notes
             GROUP BY DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00'))) notes ON notes.date = dates.date;

如果您知道，在发生安装的每一天，还会有一次点击，查询会变得更容易，因为您可以将点击查询也用作日期池。无论如何，我可能会运行一个单独的查询来获取每天的笔记，以便能够对文本采取行动。

只计算一次 txid

并且此查询将确保 users_installs 和 users_clicks 中的每个 txid 仅在最早发生的日期计算一次。由于有两个额外的子查询，性能会（至少）稍微差一些。如果 txids 已经不同，我会建议不要这样做。如果您开始在子查询中过滤日期，则可以提高性能。

SELECT
  dates.date,
  clicks.clicks,
  clicks.costs,
  installs.installs,
  notes.note_count,
  notes.notes
FROM (SELECT DISTINCT
        date
      FROM (
             SELECT
               DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
             FROM users_clicks
             UNION SELECT
                     DATE(CONVERT_TZ(click_utc, '+00:00', '-07:00')) AS date
                   FROM users_installs
             UNION SELECT
                     DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date
                   FROM reporting_notes) data) dates
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00')) AS date,
               COUNT(*)                                            AS clicks,
               SUM(max_costs)                                      AS costs
             FROM (SELECT
                     txid,
                     MIN(click_utc) AS min_click_utc,
                     MAX(cost)      AS max_costs
                   FROM users_clicks
                   GROUP BY txid) distinct_txids
             GROUP BY DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00'))) clicks ON clicks.date = dates.date
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00')) AS date,
               COUNT(*)                                            AS installs
             FROM (SELECT
                     txid,
                     MIN(click_utc) AS min_click_utc
                   FROM users_installs
                   GROUP BY txid) distinct_txids
             GROUP BY DATE(CONVERT_TZ(min_click_utc, '+00:00', '-07:00'))) installs ON installs.date = dates.date
  LEFT JOIN (SELECT
               DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00')) AS date,
               COUNT(*)                                        AS note_count,
               GROUP_CONCAT(note SEPARATOR ', ')               AS notes
             FROM reporting_notes
             GROUP BY DATE(CONVERT_TZ(timestamp, '+00:00', '-07:00'))) notes ON notes.date = dates.date;

【讨论】：

感谢您将这些放在一起。我遇到的一个问题是点击次数、安装次数和成本应该基于不同的txid 值。如果有多个点击条目具有相同的txid 和不同的cost 值，则应选择最大的cost。
@585connor 这些条目可以在不同的日期吗？应该怎么处理？
txid 在所有日子里都应该是不同的。如果一次点击或安装在不同的日期使用相同的 txid 记录，则应计为一次（最好是最早的日期）。谢谢！
@585connor 昨天很忙。这就是你要找的吗？
谢谢！感谢您抽出宝贵时间帮助我。

【解决方案2】：

认为这需要加入子查询。

一个子查询获取使用的各种日期，然后将其与一个子查询连接以获取点击次数和子项，另一个获取安装次数：-

SELECT sub0.aDate, click_count, click_cost, install_count, GROUP_CONCAT(reporting_notes.note)
FROM
(
    SELECT DATE(click_utc) AS aDate
    FROM users_clicks
    UNION
    SELECT DATE(click_utc)
    FROM users_installs
    UNION
    SELECT DATE(timestamp)
    FROM reporting_notes
) sub0
LEFT OUTER JOIN
(
    SELECT DATE(click_utc) AS aDate, COUNT(txid) AS click_count, SUM(cost) AS click_cost
    FROM users_clicks
    GROUP BY aDate
) sub1
ON sub0.aDate = sub1.aDate
LEFT OUTER JOIN
(
    SELECT DATE(click_utc) AS aDate, COUNT(txid) AS install_count
    FROM users_installs
    GROUP BY aDate
) sub2
ON sub0.aDate = sub2.aDate
LEFT OUTER JOIN reporting_notes
ON sub0.aDate = DATE(reporting_notes.timestamp)
GROUP BY sub0.aDate, click_count, click_cost, install_count

如果您只想要唯一的注释并且不需要总和，则可以在没有子查询的情况下完成计数。

【讨论】：