Top-N By Decade 连续几十年（在 SQL 服务器中）

Question

我正在尝试获取 Top 5（即最常见的）文档标题的排名列表，按十年分组，对于最近 6 个十年中的每一个。

文档标题是non-unique。在任何给定的日历年中，可能会有数十个甚至数百个具有相同标题的文档。

以下查询是我所能做的。它给了我前 5 名的头衔，但仅限于 'all others' 十年。

我如何修改查询以获取其他每个十年的前 5 个标题？

SELECT
    Top 5 documentTitle AS 'Title',
    RANK() OVER (PARTITION BY calendarYear ORDER BY COUNT(documentTitle) DESC) AS Rank,
    COUNT(tblDocumentFact.inventionTitleEnID) AS 'Number of Occurrences',
    CASE
        WHEN calendarYear BETWEEN 2010 AND 2019 THEN '2010 - 2019'
        WHEN calendarYear BETWEEN 2000 AND 2009 THEN '2000 - 2009'
        WHEN calendarYear BETWEEN 1990 AND 1999 THEN '1990 - 1999'
        WHEN calendarYear BETWEEN 1980 AND 1989 THEN '1980 - 1989'
        WHEN calendarYear BETWEEN 1970 AND 1979 THEN '1970 - 1979'
        WHEN calendarYear BETWEEN 1960 AND 1969 THEN '1960 - 1969'
        ELSE 'all others'
    END AS Decade
FROM        tbldocumentTitleDimension
INNER JOIN  tblDocumentFact     ON  tbldocumentTitleDimension.documentTitleID   = tblDocumentFact.documentTitleID
INNER JOIN  tblDateDimension    ON  tblDocumentFact.publicationDateID           = tblDateDimension.dateID
GROUP BY    documentTitle,
            calendarYear
ORDER BY    [Number of Occurrences] DESC

Answer 1

如果我没听错，你想每十年排名前 5。如果是：

您需要 group by 十年而不是日历年才能获得正确的计数；在子查询中计算十年更容易，因此您不必重复 case 表达式
排名应该根据 decade 个分区而不是每年计算
然后您可以使用该列在外部查询中进行过滤

考虑：

select *
from (
    select
        dtd.documenttitle as title,
        rank() over (partition by dd.decade order by count(*) desc) as rnk,
        count(*) as number_of_occurrences,
        dd.decade
    from tbldocumentTitleDimension dtd
    inner join tblDocumentFact df on dtd.documenttitleid   = df.documenttitleid
    inner join  (
        select 
            dateid,
            case
                when calendarYear between 2010 and 2019 then '2010 - 2019'
                when calendarYear between 2000 and 2009 then '2000 - 2009'
                when calendarYear between 1990 and 1999 then '1990 - 1999'
                when calendarYear between 1980 and 1989 then '1980 - 1989'
                when calendarYear between 1970 and 1979 then '1970 - 1979'
                when calendarYear between 1960 and 1969 then '1960 - 1969'
                else 'all others'
            end AS decade
            from tblDateDimension
    ) dd on df.publicationdateid  = dd.dateid
    group by dtd.documenttitle, dd.decade
) t
where rnk <= 5
order by decade, number_of_occurrences desc

旁注：

不要对标识符使用单引号（虽然 SQL 服务器允许，单引号应该保留用于垃圾字符串，如 SQL 标准中所定义） -更好的是，您可以使用不需要引用的标识符
在多table查询中，总是用它们所属的table限定所有列名；我在这里做了一些假设
除非您不想将 documentTitle 列中的 null 值计入，否则您可以使用 count(*) 而不是 count(documentTitle) - 这很直接，而且效率更高

Top-N By Decade 连续几十年（在 SQL 服务器中）

Top-N By Decade for successive decades (in SQL Server)

sql

sql-server

group-by

greatest-n-per-group

window-functions