get all nested children for a parent id

问题:

Started Fiddling

Work Table
ProductId, LabelName, CategoryId, ChildCategoryId
------------------------------------
1, Widget A, 1, null
null, Category A, 2, 1 
2, Widget B, 3, null

Categories Table
CategoryId, CategoryName
---------------------------
1, Category A
2, Category B
3, Category C

Given the information above, how would you get all categories for a product id?

For example, given a product id of 1, the following would be the desired results.

Desired Results
ProductId, LabelName, CategoryId, ChildCategoryId
------------------------------------
1, Widget A, 1, null
null, Category A, 2, 1 
null, Category B, null, 2

It is supposed to be hierarchical data and I apologize for not being able to explain very well. It is just boggling my mind. Widget A has a product id of 1 and category id of 1. This means all records that have a ChildCategoryId of 1 is included, which gives us Category A. CatA has a category id of 2, so like before, all records that have a ChildCategoryId of 2 is included in result, which is why Category B is included.

回答1:

This mess produces the sample result from the sample data. It still isn't clear what you think the algorithm ought to be.

declare @CategoryItems as Table (
  CategoryName NVarChar(255),
  Label NVarChar(255),
  ProductId Int,
  ChildCategoryId Int,
  CategoryId Int );

declare @Categories as Table (
  CategoryId Int,
  Name NVarChar(100) );

insert into @CategoryItems ( CategoryName, Label, ProductId, ChildCategoryId, CategoryId ) values
  ( 'CategoryA', 'Widget A', 1, 0, 1 ),
  ( 'CategoryB', 'CategoryA', 0, 1, 2 ),
  ( 'CategoryC', 'Widget B', 2, 0, 3 );
insert into @Categories ( CategoryId, Name ) values
  ( 1, 'CategoryA' ),
  ( 2, 'CategoryB' ),
  ( 3, 'CategoryC' );

select * from @Categories;
select * from @CategoryItems;

declare @TargetProductId as Int = 1;

with Leonard as (
  -- Start with the target product.
  select 1 as [Row], ProductId, Label, CategoryId, ChildCategoryId
    from @CategoryItems
    where ProductId = @TargetProductId
  union all
  -- Add each level of child category.
  select L.Row + 1, NULL, CI.Label, CI.CategoryId, CI.ChildCategoryId
    from @CategoryItems as CI inner join
      Leonard as L on L.CategoryId = CI.ChildCategoryId ),
  Gertrude as (
    -- Take everything that makes sense.
    select Row, ProductId, Label, CategoryId, ChildCategoryId
      from Leonard
    union
    -- Then tack on an extra row for good measure.
    select L.Row + 1, NULL, C.Name, NULL, C.CategoryId
      from Leonard as L inner join
        @Categories as C on C.CategoryId = L.CategoryId
      where L.Row = ( select Max( Row ) from Leonard ) )
  select Row, ProductId, Label, CategoryId, ChildCategoryId
    from Gertrude
    order by Row;

I suspect that the problem is that you have mixed your data in a lopsided manner. A hierarchy of categories is usally represented something like:

declare @Categories as Table (
  CategoryId Int Identity,
  Category NVarChar(128),
  ParentCategoryId Int Null );

The root of each hierarchy is indicated by ParentCategoryId is NULL. This allows any number of independent trees to coexist in a single table and does not depend on the existence of any products.

If products are assigned to a single (sub)category then just include the CategoryId in the Products table. If a product may be assigned to several (sub)categories, possibly in different hierarchies, then use a separate table to relate them:

declare @ProductCategories as Table (
  ProductId Int,
  CategoryId Int );