I’m having trouble joining two hold files together. The problem lies in that the unique identifier I want to join on is a composite of two fields, for example’s sake, let’s call them EmpNumber and Company. EmpNumber is unique within a Company, but multiple companies can have the same EmpNumber.
So, here’s a brief and oversimplified version of what I’ve got to do:

TABLE FILE TABLEA
PRINT
Customer
EmpNumber
DATAA
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDA FORMAT FOCUS INDEX EmpNumber Customer
END

TABLE FILE TABLEB
PRINT
Customer
EmpNumber
DATAB
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDA FORMAT FOCUS INDEX EmpNumber Customer
END

JOIN
LEFT_OUTER FILE HOLDA AT HOLDA.SEG01.EmpNumber TO UNIQUE FILE HOLDB
AT HOLDB.SEG01.EmpNumber TAG J0 AS J0
END

TABLE FILE EMP
PRINT
'HOLDA.SEG01.Customer'
'HOLDA.SEG01.EmpNumber'
'HOLDA.SEG01.DATAA'
'J0.SEG01.DATAB'

ON TABLE SET BYDISPLAY ON
ON TABLE PCHOLD FORMAT HTML
ON TABLE NOTOTAL
ON TABLE SET PAGE-NUM OFF
ON TABLE SET HTMLCSS ON
END

What I'm having a hard time doing is making the join be on both EmpNumber and Client. From reading the documentation, it suggests that one can AND join fields together. So, I've tried the equivalent of:

JOIN
LEFT_OUTER FILE HOLDA AT HOLDA.SEG01.EmpNumber AND HOLDA.SEG01.Customer TO UNIQUE FILE HOLDB
AT HOLDB.SEG01.EmpNumber AND HOLDB.SEG01.Customer TAG J0 AS J0
END

But it just errors out. Has anyone else out there been in a similar situation and come up with a good solution? Ideas would be appreciated... I have tried a "Where" based join, but it runs at least 20 times slower, so that's not really an option.

You are using a conditional join.

To do equality tests, you will have to add WHERE host field EQ join field ; before the END of the join.

You can have multiple WHERE clauses in conditiona join, each one on a separate lina dn ending ina ;.


What is the error message you are getting ?

When joining FOCUS files, the host does not need to have an INDEX, only the x-referenced file. Also a join to a FOCUS file can only be to ONE field, unlike SQL based data. The host file can use up to 4 fields (I think) to JOIN to the x-referenced file. The following is roughly what I think you are after.
(Being old school I prefer the old style join syntax...)

TABLE FILE TABLEA
PRINT
Customer
EmpNumber
DATAA
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDA FORMAT FOCUS 
END

TABLE FILE TABLEB
PRINT
DATAB
COMPUTE LINK/Ann = Customer | EmpNumber;
BY Customer
BY EmpNumber
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDB FORMAT FOCUS INDEX LINK
END

JOIN
        Customer and EmpNumber IN HOLDA TAG A 
TO      LINK                   IN HOLDB TAG B AS J0
END

TABLE FILE HOLDA
PRINT
A.Customer
A.EmpNumber
A.DATAA
B.DATAB
ON TABLE SET BYDISPLAY ON
ON TABLE PCHOLD FORMAT HTML
ON TABLE NOTOTAL
ON TABLE SET PAGE-NUM OFF
ON TABLE SET HTMLCSS ON
END

Another method, which I also sometimes use, is:

JOIN
        Customer IN HOLDA TAG A 
TO ALL  Customer IN HOLDB TAG B AS J0
END

TABLE FILE HOLDA
PRINT
A.Customer
A.EmpNumber
A.DATAA
B.DATAB
WHERE A.EmpNumber EQ B.EmpNumber;
END

In this case you have to have the ALL (1-n relation) in the join, the WHERE will filter out unequal records.

Thanks everyone for the suggestions. The error was related to the rule Alan mentioned that FOCUS won't allow joins on 2 fields. I've done a lot of SQL, so I guess I took that for granted... Anyways, taking that option off the table, I've used the input to craft a solution that works for me (all 3 suggestions do in fact work).
In the report(s) I've actually got, I've got a few more hold files involved. Also, Customer and EmpNumber are both whole numbers, and the max allowable value on Customer is 99999. Since most systems are most efficient at comparing numbers, I've tried this approach:


TABLE FILE TABLEA
PRINT
Customer
EmpNumber
DATAA
COMPUTE CliEmp/P26 = (EmpNumber * 100000) + Customer; NOPRINT
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDA FORMAT FOCUS INDEX CliEmp
END

TABLE FILE TABLEB
PRINT
Customer
EmpNumber
DATAB
COMPUTE CliEmp/P26 = (EmpNumber * 100000) + Customer; NOPRINT
ON TABLE SET PAGE-NUM OFF
ON TABLE NOTOTAL
ON TABLE HOLD AS HOLDA FORMAT FOCUS INDEX CliEmp
END

JOIN
LEFT_OUTER HOLDA.SEG01.CliEmp IN HOLDA TO UNIQUE HOLDB.SEG01.CliEmp IN HOLDB TAG J0 AS J0
END

TABLE FILE EMP
PRINT
'HOLDA.SEG01.Customer'
'HOLDA.SEG01.EmpNumber'
'HOLDA.SEG01.DATAA'
'J0.SEG01.DATAB'

ON TABLE SET BYDISPLAY ON
ON TABLE PCHOLD FORMAT HTML
ON TABLE NOTOTAL
ON TABLE SET PAGE-NUM OFF
ON TABLE SET HTMLCSS ON
END


I've still got to run it through the performance analyzer, and I'll compare the results with the other 3 approaches suggested to find the one that will scale best for me.

Again, thanks everyone for putting me on the path to solving this. Your input, as always, is invaluable.

This message has been edited. Last edited by: Byrnsy,