UNION操作符对您有用吗?每个部分有一个SELECT，然后将它们联合在一起。不过，我猜它只适用于固定数量的部分。

我知道这个帖子有点老了，但我刚刚遇到了一个类似的问题(从每个类别中选择最新的文章)，这是我想出的解决方案:

WITH [TopCategoryArticles] AS (
    SELECT 
        [ArticleID],
        ROW_NUMBER() OVER (
            PARTITION BY [ArticleCategoryID]
            ORDER BY [ArticleDate] DESC
        ) AS [Order]
    FROM [dbo].[Articles]
)
SELECT [Articles].* 
FROM 
    [TopCategoryArticles] LEFT JOIN 
    [dbo].[Articles] ON
        [TopCategoryArticles].[ArticleID] = [Articles].[ArticleID]
WHERE [TopCategoryArticles].[Order] = 1

这与Darrel的解决方案非常相似，但克服了可能返回比预期更多行的RANK问题。

我是这样做的:

SELECT a.* FROM articles AS a
  LEFT JOIN articles AS a2 
    ON a.section = a2.section AND a.article_date <= a2.article_date
GROUP BY a.article_id
HAVING COUNT(*) <= 10;

更新:这个GROUP BY的例子只适用于MySQL和SQLite，因为这些数据库在GROUP BY方面比标准SQL更允许。大多数SQL实现要求选择列表中不属于聚合表达式的所有列也在GROUP BY中。

虽然这个问题是关于SQL Server 2005的，但大多数人已经继续前进了，如果他们确实发现了这个问题，在其他情况下首选的答案可能是使用CROSS APPLY，就像这篇博客文章中所说明的那样。

SELECT *
FROM t
CROSS APPLY (
  SELECT TOP 10 u.*
  FROM u
  WHERE u.t_id = t.t_id
  ORDER BY u.something DESC
) u

该查询涉及两个表。OP的查询只涉及1个表，在这种情况下，基于窗口函数的解决方案可能更有效。

如果我们使用SQL Server >= 2005，那么我们可以只用一个选择来解决任务:

declare @t table (
    Id      int ,
    Section int,
    Moment  date
);

insert into @t values
(   1   ,   1   , '2014-01-01'),
(   2   ,   1   , '2014-01-02'),
(   3   ,   1   , '2014-01-03'),
(   4   ,   1   , '2014-01-04'),
(   5   ,   1   , '2014-01-05'),

(   6   ,   2   , '2014-02-06'),
(   7   ,   2   , '2014-02-07'),
(   8   ,   2   , '2014-02-08'),
(   9   ,   2   , '2014-02-09'),
(   10  ,   2   , '2014-02-10'),

(   11  ,   3   , '2014-03-11'),
(   12  ,   3   , '2014-03-12'),
(   13  ,   3   , '2014-03-13'),
(   14  ,   3   , '2014-03-14'),
(   15  ,   3   , '2014-03-15');


-- TWO earliest records in each Section

select top 1 with ties
    Id, Section, Moment 
from
    @t
order by 
    case 
        when row_number() over(partition by Section order by Moment) <= 2 
        then 0 
        else 1 
    end;


-- THREE earliest records in each Section

select top 1 with ties
    Id, Section, Moment 
from
    @t
order by 
    case 
        when row_number() over(partition by Section order by Moment) <= 3 
        then 0 
        else 1 
    end;


-- three LATEST records in each Section

select top 1 with ties
    Id, Section, Moment 
from
    @t
order by 
    case 
        when row_number() over(partition by Section order by Moment desc) <= 3 
        then 0 
        else 1 
    end;

SELECT r.*
FROM
(
    SELECT
        r.*,
        ROW_NUMBER() OVER(PARTITION BY r.[SectionID]
                          ORDER BY r.[DateEntered] DESC) rn
    FROM [Records] r
) r
WHERE r.rn <= 10
ORDER BY r.[DateEntered] DESC