数据库设计是否规范化

时间:2012-07-07 15:20:24

标签: sql sql-server tsql database-design

我有以下表格

Groups
- Id
- Name

People
- Id
- GroupId
- EntryDate (DateTime)

现在我想要检索按组的人员入职日期排序的所有组,这是最新的。

//pseudo code

select * from groups order by (select top 1 from people order by entrydate desc)

将LatestEntryDate字段添加到Groups并在添加人员时更新它会更好吗?因为这将是一个常见的查询。

2 个答案:

答案 0 :(得分:2)

SELECT g.Id, g.Name, MAX(p.EntryDate) AS LatestPeopleEntryDate
FROM Groups g
    INNER JOIN People p
        ON p.GroupId = g.Id
GROUP BY g.Id, g.Name
ORDER BY MAX(p.EntryDate)

不,我不会在LatestEntryDate表中添加Groups字段。这很容易计算。

正如@danihp所指出的,你可能想在这里使用左连接。上述查询将要求People表中存在一个值。您选择使用的内容取决于您的具体要求。

SELECT g.Id, g.Name, MAX(p.EntryDate) AS LatestPeopleEntryDate
FROM Groups g
    LEFT JOIN People p
        ON p.GroupId = g.Id
GROUP BY g.Id, g.Name
ORDER BY MAX(p.EntryDate)

<强>更新

要回答关于在Linq中执行此操作的问题,我认为它会是这样的:

var result = groups.Select(g => 
    new {
            g.Id, 
            g.Name, 
            LatestPeopleEntryDate = g.People.Max(p => p.EntryDate)
        });

答案 1 :(得分:0)

这是一个测试示例,其中包含非常小的样本数据,显示了三种方法。

要获得最佳查询计划:

在People上放置一个非唯一索引(GroupID,EntryDate。)

一个。使用原始伪代码(按顺序执行,但不显示日期)或

B中。使用前1个子查询来获取并显示日期。

A和B的查询计划是相同的。

使用左边连接到People表和max()将导致扫描(在GroupID中,充其量)获得最大值,而不是探测索引中的单行。

set nocount on
if object_id('Groups') is not null drop table Groups
if object_id('People') is not null drop table People
go

-- set up tables
create table Groups
(
    ID int primary key,
    Name varchar(20)
)

create table People 
(
    ID int,
    GroupID int,
    EntryDate datetime
)
-- make an index that is ordered by Group, EntryDate
create index IE_GroupDate on People(GroupID, EntryDate)

-- Sample data
insert into Groups (ID, Name)
values
    (1, 'Group1'),
    (2, 'Group2'),
    (3, 'GroupC')

insert into People (ID, GroupID, EntryDate)
values 
    (1, 1, '2012-01-01'),
    (2, 1, '2012-02-01'),
    (1, 3, '2007-12-31')

-- Queries
-- Equivalent to the original poster's intent.  Note that it doesn't actually
-- show the entry date
select *
from Groups G
order by (
    select top 1 EntryDate
    from People P
    where P.GroupID = G.ID order by EntryDate desc)

-- Same query (by query plan) but includes the EntryDate in the result set
select
    G.ID,
    G.Name,
    LatestEntryDate = Latest.EntryDate
from Groups G
outer apply (
    select top 1 EntryDate
    from People P
    where P.GroupID = G.ID
    order by EntryDate desc
) Latest
order by LatestEntryDate

-- Query using left join.  This causes a scan of the left join table to
-- compute the max.  (The optimizer isn't smart enough to turn this into 
-- a TOP 1)
select
    G.ID,
    G.Name,
    LatestEntryDate = max(P.EntryDate)
from Groups G
left join People P on P.GroupID = G.ID
group by G.ID, G.Name
order by max(P.EntryDate)