Question

我正在尝试查询即将到来的数据，在一个特殊的情况下，我可能不得不按多个值对数据进行分组。但是大多数时候它根本不需要分组。因此，我在GROUP BY中使用了多个CASE WHEN {...}，并且所有WHEN基本上都具有相同的条件。问题是，如果满足条件，则一切正常。但是，如果为假，则GROUP BY部分为空，并且查询仅返回第一行。

我基本上想方设法以各种方式对查询进行重组，似乎没有任何效果，并且我在互联网上没有发现任何结论性的信息。

我正在使用MySql 5.7。

SELECT
    {element I want to select}
FROM
    {tables}
WHERE
    {conditions}
GROUP BY
    CASE WHEN (condition) THEN [table].[column] END,
    CASE WHEN (condition) THEN [table].[column] END,
    CASE WHEN (condition) THEN [table].[column] END
ORDER BY
    {...}

完整查询：

SELECT
    tx.code,
    IFNULL(hr.label,'')                         AS rh_label,
    IFNULL(cli.label,'')                        AS client_label,
    DATE(FROM_UNIXTIME(created.value / 1000))   AS Created,
    IFNULL(item_enfant.label,'')                As Parasite,
    IFNULL(item_parent.label,'')                As Zone,
    CASE 
        WHEN :perWeek = 'week' THEN SUM(qte.value)
        ELSE qte.value
    END AS Quantite,
    CEILING(DATEDIFF(DATE(FROM_UNIXTIME(created.value / 1000)), DATE(FROM_UNIXTIME(:from / 1000))) / 7) AS Weeks
FROM tx
    LEFT JOIN tx_type           AS tt ON tt.id = tx.tx_type_id
    LEFT JOIN human_resource    AS hr ON hr.id = tx.human_resource_id
    LEFT JOIN client            AS cli ON cli.id = tx.client_id
    LEFT JOIN tx_state          AS ts ON ts.id = tx.current_tx_state_id
    LEFT JOIN workflow_step     AS ws ON ws.id = ts.workflow_step_id
    LEFT JOIN item              AS item_enfant ON item_enfant.item_list_id = tx.item_list_id
    JOIN item_type              AS ite ON ite.id = item_enfant.item_type_id
    LEFT JOIN item_meta         AS qte ON qte.item_id = item_enfant.id AND qte.name = 'qtePourRapport'
    LEFT JOIN item_prop         AS created ON created.item_id = item_enfant.id AND created.name = 'visite.timestamp'
    JOIN item                   AS item_parent ON item_parent.id = item_enfant.parent_item_id 
    JOIN item_type              AS itp ON itp.id = item_parent.item_type_id
WHERE
    ite.name = 'parasite' AND
    item_enfant.product_id IN (:parasiteIds) AND
    itp.name = 'zone' AND
    item_parent.product_id IN (:zoneIds) AND
    cli.id = (:clientId) AND 
    ws.logic_id = 600 AND 
    created.value BETWEEN :from AND :to AND 
    created.value IS NOT NULL AND qte.value IS NOT NULL
GROUP BY
    CASE WHEN :perWeek = 'week' THEN item_enfant.label END, #Parasite
    CASE WHEN :perWeek = 'week' THEN item_parent.label END, #Zone
    CASE WHEN :perWeek = 'week' THEN CEILING(DATEDIFF(DATE(FROM_UNIXTIME(created.value / 1000)), DATE(FROM_UNIXTIME(:from / 1000))) / 7) END #Weeks
ORDER BY 
    Created;

我仅获得第一行的数据。而且我实际上不知道如何在不满足条件的情况下不进行分组。

Answer 1

对于聚合或两个单独的查询，您需要一个唯一的值。最简单的方法可能是union all：

select . . .
from t
where <conditions not to group by>
union all
select . . .
from t
where <conditions to group by>
group by . . .;

您需要确保每个子查询返回兼容的列。

Answer 2

SELECT
    {element I want to select}
FROM
    {tables}
WHERE
    {conditions}
GROUP BY
    CASE WHEN (condition) THEN [table].[column] ELSE [some unique value of same data-type as column] END,
    CASE WHEN (condition) THEN [table].[column] ELSE [some unique value of same data-type as column] END,
    CASE WHEN (condition) THEN [table].[column] ELSE [some unique value of same data-type as column] END
ORDER BY
    {...}

我猜缺少的ELSE子句将评估为NULL。这是恒定的，因此所有行将在同一组中，因此该组仅返回一行。为避免分组，您需要在分组元素的组合中所有返回的行上具有唯一的值（而不是如先前所述，不是在每个分组元素中）。

编辑

因此，从注释中获得灵感可能更容易：只需添加另一个分组元素CASE WHEN !(condition) THEN CONCAT([different elements making it unique]) END

在GROUP BY中是否可以有多个CASE可以使它完全不分组？

2 个答案: