Question

What would be the best way to model 1 table with multiple 1 to many relatiionships.

With the above schema if Report contains 1 row, Grant 2 rows and Donation 12. When I join the three together I end up with a Cartesian product and result set of 24. Report joins to Grant and creates 2 rows, then Donation joins on that to make 24 rows.

Is there a better way to model this to avoid the caresian product?

example code

DECLARE @Report
TABLE   (
        ReportID    INT,
        Name        VARCHAR(50)
        )

INSERT
INTO    @Report 
        (
        ReportID,
        Name 
        )

SELECT  1,'Report1'


DECLARE @Grant 
TABLE   (
        GrantID     INT IDENTITY(1,1) PRIMARY KEY(GrantID),
        GrantMaker  VARCHAR(50),
        Amount      DECIMAL(10,2),
        ReportID    INT
        )

INSERT
INTO    @Grant 
        (
        GrantMaker,
        Amount,
        ReportID
        )

SELECT  'Grantmaker1',10,1
UNION ALL
SELECT  'Grantmaker2',999,1


DECLARE @Donation
TABLE   (
        DonationID      INT IDENTITY(1,1) PRIMARY KEY(DonationID),
        DonationMaker   VARCHAR(50),
        Amount          DECIMAL(10,2),
        ReportID        INT
        )

INSERT
INTO    @Donation 
        (
        DonationMaker,
        Amount,
        ReportID
        )

SELECT  'Grantmaker1',10,1
UNION ALL
SELECT  'Grantmaker2',3434,1
UNION ALL
SELECT  'Grantmaker3',45645,1
UNION ALL
SELECT  'Grantmaker4',3,1
UNION ALL
SELECT  'Grantmaker5',34,1
UNION ALL
SELECT  'Grantmaker6',23,1
UNION ALL
SELECT  'Grantmaker7',67,1
UNION ALL
SELECT  'Grantmaker8',78,1
UNION ALL
SELECT  'Grantmaker9',98,1
UNION ALL
SELECT  'Grantmaker10',43,1
UNION ALL
SELECT  'Grantmaker11',107,1
UNION ALL
SELECT  'Grantmaker12',111,1

SELECT  *
FROM    @Report r
INNER JOIN
        @Grant g
        ON  r.ReportID = g.ReportID 
INNER JOIN
        @Donation d
        ON  r.ReportID = d.ReportID 

Update 1 2011-03-07 15:20

Cheers for the feedback so far, to add to this scenario there are also 15 other 1 to many relationships coming from the one report table. These tables can’t for various business reasons be grouped together.

Answer 1

If you’re going to join on ReportID, then no, you can’t avoid a lot of rows. When you omit the table “Report”, and just join “Donation” to “Grant” on ReportId, you still get 24 rows.

SELECT  *
FROM    Grant g
INNER JOIN
        Donation d
        ON  g.ReportID = d.ReportID 

But the essential point is that it doesn’t make sense in the real world to match up donations and grants. They’re completely independent things that essentially have nothing to do with each other.

In the database, the statement immediately above will join each row in Grants to every matching row in Donation. The resulting 24 rows really shouldn’t surprise you.

When you need to present independent things to the user, you should use a report writer or web application (for example) that selects the independent things, well, independently. Select donations and put them into one section of a report or web page, then select grants and put them into another section of the report or web page, and so on.

If the table “Report” is supposed to help you record which sections go into a particular report, then you need a structure more like this:

create table reports (
    reportid integer primary key,
    report_name varchar(35) not null unique
);

create table report_sections (
    reportid integer not null references reports (reportid),
    section_name varchar(35),  -- Might want to reference a table of section names
    section_order integer not null,
    primary key (reportid, section_name)
);