Re: Attempt to write duplicate record...

[John McKay <jmckay@xxxxxxxxxxxxxxxx>]

I am assuming an externally defined file here.

Put simply, I believe the format name is more precise and reliable than 
the using the file name.

On the READ / CHAIN / SETLL / SETGT statement, a logical file with more 
than one format may have different keys and/or select statements.  Thus, 
a READ to the format name ensures that I am attempting to read precisely 
the data record I require.  The use of the file name will get the 
next/previous record in the file, but not necessarily the expected format.

On the WRITE statement, "When name is a record format name, the current 
values in the program for all the fields in the record definition are 
used to construct the record." (ILE RPG Language Reference).

With the use of OVRDBF and EXTFILE which can change the file in use at 
run time, as opposed to that used at compile time, there is opportunity 
for a file with a different record format to be accessed by this 
program.  Here, the use of the format name instead of the file name will 
ensure data integrity.

I would examine the record format here.  The Primary Key Constraint is  
not the same as keys specified on the format, nor is the uniqueness.

If you were to capture the error code, you might find it is either of ...

1021 - duplicate record
1022 - referential constraint violation


Regards,
John McKay  mba

On 15/11/2011 17:24, Robert Rogerson wrote:
>     Thanks John, but I'm curious why you say this and why you think this would
>     make a difference?
>
>     Personally my standard if to Read a FILE and write a RECORD.
>
>     On 2011-11-15 11:44 AM, John McKay wrote:
>
>   I would use the record format name instead of the file name on the CHAIN
>   and on the WRITE ...
>
>   Regards,
>   John McKay  mba
>   [1]www.rpglanguage,com
>
>   On 15/11/2011 16:29, Robert Rogerson wrote:
>
>       Hi All,
>
>       I had a stored procedure fail with RNX1021 Attempt to write a duplicate
>       record to file SIGNTMP.
>
>       SIGNTMP has the following constraint:
>
>       Constraint
>       Description
>           Primary Key
>       Constraint
>             Constraint  . . . . . . . . . . . . . . : CST
>       Q_QS36F_SIGNTMP_WK
>               Type  . . . . . . . . . . . . . . . . : TYPE
>       *PRIMARY
>               Key . . . . . . . . . . . . . . . . . : KEY
>       WKCUS#
>
>       WKITM#
>
>       WKDOCT
>
>       WKDOC#
>               Number of fields in key . . . . . . . :
>       4
>       In the code I have...
>              // Write to the work file if not already there
>              Chain(n) (p_store: p_item: p_DocumentType: p_Document) signtmp;
>              If not %Found(signtmp);
>                wkcus# = p_store;
>                wkitm# = p_item;
>                wkdoct = p_DocumentType;
>                wkdoc# = p_Document;
>                Write signtm;
>              Endif;
>
>       I thought this was pretty solid code but the program still failed.  A new
>       record is only written if a record with the same key is not found.
>
>       So my only explanation is that another process wrote a record with the
>       same key in between the Chain and the Write.
>
>       Am I missing something of does some have another explanation.
>
>       BTW, I understand that I can add an error extender to the Write and
>       capture the error but my question still remains if there's another
>       explanation for the failure...
>
>       Thanks for all ideas,
>
>       Robert Rogerson
>
> References
>
>     Visible links
>     1. http://www.rpglanguage,com/