从 SQL 中的 LEFT JOIN 创建矩阵/表

发布于 2024-07-27 12:36:00 字数 556 浏览 3 评论 0原文

我想知道将 LEFT JOIN 转换为表/矩阵的好解决方案。

例如,给定这个模式:

objects
   id (integer)
   name (string)

attributes
   id (integer)
   object_id (integer)
   name (string)
   value (string)

和这些值:

1,ball
2,box

1,1,colour,red
2,1,shape,sphere
3,1,material,rubber
4,2,colour,brown
5,2,shape,cube

我想得到这个:

object_id | object_name | colour | shape  | material

1         | ball        | red    | sphere | rubber
2         | box         | brown  | cube   | NULL

理想情况下,每个对象都有未知数量的属性,并且适用于 MySQL,而不使用存储过程。

I'd like to know a good solution for converting a LEFT JOIN into a table/matrix.

For example, given this schema:

objects
   id (integer)
   name (string)

attributes
   id (integer)
   object_id (integer)
   name (string)
   value (string)

And these values:

1,ball
2,box

1,1,colour,red
2,1,shape,sphere
3,1,material,rubber
4,2,colour,brown
5,2,shape,cube

I'd like to get this:

object_id | object_name | colour | shape  | material

1         | ball        | red    | sphere | rubber
2         | box         | brown  | cube   | NULL

Ideally this would be with an unknown number of attributes for each object, and be for MySQL without using stored procedures.

如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。

扫码二维码加入Web技术交流群

发布评论

需要 登录 才能够评论, 你可以免费 注册 一个本站的账号。

评论(4

野侃 2024-08-03 12:36:00

RDBMS 的特点是它们的目的是存储和呈现您已有的内容。 动态创建列从来都不是有意的。

这种情况应该在要呈现数据的客户端上处理。

不管 RDBMS 应该如何使用,您仍然找不到一种简单的方法来做到这一点。 尤其是如果您希望列列表是动态的。 您可以构建动态 SQL 查询并执行它,但无法编写标准 SQL 查询并获取结果(因为您总是显式指定所有列,禁止使用 *,但这不能在这种情况下可以利用你的优势。)

The thing with RDBMSs is that their purpose is to store and present what you already have. Creating columns dynamically was never intended.

This is a case that should be handled on the client that is to present the data.

With disregard to how RDBMSs are supposed to be used, you still won't find an easy way to do this. Especially not if you want the column list to be dynamic. You can build a dynamic SQL query and execute that, but there's no way to write a standard SQL query and get that result (since you always specify all columns explicitly, bar using *, but that can't be used to your advantage in this case.)

御守 2024-08-03 12:36:00

如果不使用 SQL 中的动态查询,则无法做到这一点。

SQL 处理的任何集合都假定具有固定数量的列和固定名称。

否则会导致许多 SQL 无法处理的事情。

就像,想象一下根据您要求的查询创建的视图(是否可能):

SELECT  object_id, colour
FROM    myquery

该视图是否编译将取决于查询的结果,这会给您带来运行时错误(很难处理),而不仅仅是空的结果集(很容易处理)。

您在这里基本上谈论的是结果呈现,而这些事情通常是在客户端完成的。

如果您有包含所有可能的属性名称的 attribute_names 表,您可以执行以下操作:

SELECT  o.name, an.name, a.value
FROM    objects o
CROSS JOIN
        attribute_names an
LEFT JOIN
        attributes a
ON      a.object = o.id
        AND a.name = an.name

,它将包含所有可能的 object_id / attribute_name 对以及相应的值,并且您可以在客户端使用它来更轻松地填充矩阵。

You cannot do it without using dynamical queries in SQL.

Any set SQL deals with assumes a fixed number of columns, with fixed names.

Otherwise it would lead to many things SQL is just not designed to deal with.

Like, imagine a view created on the query you asking for (were it possible):

SELECT  object_id, colour
FROM    myquery

Whether this view compiles or not would depend on the results of the query and this would give you runtime errors (which are hard to handle) instead of mere empty resultset (which is easy to handle).

You are basically talking about the results presentation here, and these things are usially done on the client side.

If you have attribute_names table which contains all possible names of your attributes, you can do something like this:

SELECT  o.name, an.name, a.value
FROM    objects o
CROSS JOIN
        attribute_names an
LEFT JOIN
        attributes a
ON      a.object = o.id
        AND a.name = an.name

, which will contain all possible object_id / attribute_name pairs with corresponding values, and you can use it on client side to fill a matrix more easily.

奢望 2024-08-03 12:36:00

假设结果表被称为合并,这将解决您的问题,尽管我认为您拥有的模式更好。

insert into merged (object_id, object_name) select id, name from objects;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'colour' set m.colour = a.value;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'material' set m.material = a.value;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'shape' set m.shape = a.value;

This would solve your problem assuming the result table is called merged, although the schema you have I'd say was better.

insert into merged (object_id, object_name) select id, name from objects;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'colour' set m.colour = a.value;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'material' set m.material = a.value;
update merged m inner join attributes a on a.object_id = m.object_id and a.name = 'shape' set m.shape = a.value;
末蓝 2024-08-03 12:36:00

嗯,这可能是您缺少 2 个逻辑步骤来实现这一点。

您需要的是 AttributeTypes 表和 AttributeValues 表。

objects
   id (integer)
   name (string)

attribute_types
   id (integer)
   name (string)
   description (string)
   enabled (bit)

attribute_values
   id (integer)
   attribute_type_id (integer)
   value (string)
   enabled (bit)

attributes
   id (integer)
   object_id (integer)
   attribute_type_id (integer)
   attribute_value_id (integer)

所以一个对象可以有属性,这些属性由attribute_type_id和attribute_value_id指定。

然后,这允许您拥有一个具有多个属性的对象。

例如

object 
-> 1, ball

attribute_types
-> 10, Colour, null, enabled
-> 20, Shape, null, enabled
-> 30, Material, null, enabled

attribute_values 
-> 100, 10, blue, enabled
-> 200, 10, red, enabled
-> 300, 10, green, enabled

-> 400, 20, round, enabled
-> 500, 20, square, enabled
-> 600, 20, triangle, enabled

,一个属性看起来像:

attributes
-> 1000, 1, 10, 100 // this item has a color and it is blue
-> 1001, 1, 20, 400 // this item has a shape and it is round
-> 1002, 1, 10, 200 // this item has a color and it is red

所以现在对象可以有多个属性,这些属性在不同的表中指定它们的值。 现在重要的问题是如何查询这个? 您可能必须将查询分成多个部分,具体取决于您的 SQL 的强度。

@attribute_type_id = select Id from attribute_types where name = 'Color' // 10

select * from objects 
inner join attributes on objects.id = attributes.object_id
inner join attribute_values on objects.attribute_value_id = attribute_values.id
where attribute_type_id = @attribute_type_id
and attribute_values.value= 'blue'

现在您应该带回具有 Color 的 attribute_type 和蓝色的 attribute_value 的每个对象。

我的 Sql 不是那么强大,但如果您想一次搜索多个属性,您应该能够执行多个子句。 在我看来,属性表中的 attribute_type_id 没有加入到 attribute_type 表中,但它可以促进在 1 次命中中执行查询,尽管我拥有它,但性能方面,它可以通过不必加入 a 来加速查询表,但差异可能可以忽略不计,具体取决于一切的大小。

注意:我通常使用 msssql 而不是 mysql,所以如果数据库类型不匹配,这就是原因。

Hmm this is possible your just missing 2 logical steps to make this happen.

What you need is an AttributeTypes table and an AttributeValues table.

objects
   id (integer)
   name (string)

attribute_types
   id (integer)
   name (string)
   description (string)
   enabled (bit)

attribute_values
   id (integer)
   attribute_type_id (integer)
   value (string)
   enabled (bit)

attributes
   id (integer)
   object_id (integer)
   attribute_type_id (integer)
   attribute_value_id (integer)

So an object can have attributes, these attributes are specified by an attribute_type_id and attribute_value_id.

This then allows you to have an object with multiple attributes..

eg

object 
-> 1, ball

attribute_types
-> 10, Colour, null, enabled
-> 20, Shape, null, enabled
-> 30, Material, null, enabled

attribute_values 
-> 100, 10, blue, enabled
-> 200, 10, red, enabled
-> 300, 10, green, enabled

-> 400, 20, round, enabled
-> 500, 20, square, enabled
-> 600, 20, triangle, enabled

So an attribute would look like:

attributes
-> 1000, 1, 10, 100 // this item has a color and it is blue
-> 1001, 1, 20, 400 // this item has a shape and it is round
-> 1002, 1, 10, 200 // this item has a color and it is red

So now objects can have multiple attributes which specify their values in different tables. Now the important question how do you query this? You may have to split the query into multiple parts depending how strong your sql is.

@attribute_type_id = select Id from attribute_types where name = 'Color' // 10

select * from objects 
inner join attributes on objects.id = attributes.object_id
inner join attribute_values on objects.attribute_value_id = attribute_values.id
where attribute_type_id = @attribute_type_id
and attribute_values.value= 'blue'

And there you have it you should bring back every object that has the attribute_type of Color and the attribute_value of blue.

My Sql isn't that strong but you should be able to do more than one clause if you want to search for multiple attributes at a time. In my mind attribute_type_id in attributes table isn't joined onto the attribute_type table, but it can be to facilitate doing the query in 1 hit, performance wise though I'de have it so it would speed up the queries by not having to join a table but the difference might be negligible depending on how big everything gets.

Note: I normally work with msssql and not mysql so if the database types don't match up that's why.

~没有更多了~
我们使用 Cookies 和其他技术来定制您的体验包括您的登录状态等。通过阅读我们的 隐私政策 了解更多相关信息。 单击 接受 或继续使用网站,即表示您同意使用 Cookies 和您的相关数据。
原文