×
The internal search function is temporarily non-functional. The current search engine is no longer viable and we are researching alternatives.
As a stop gap measure, we are using Google's custom search engine service.
If you know of an easy to use, open source, search engine ... please contact support@midrange.com.
So occasionally we come across data in our files that SQL does not like to
play with.
RLA handles it fine. But as we move more towards SQL it is becoming a
problem.
Example
Embedded sql...
select * from table into :ds_table
when the fetch occurs to move the row into the host ds, we get the
following error if "There is data in a decimal field that is not valid.
Date sent . . . . . . : 02/20/19 Time sent . . . . . . :
15:26:00
Message . . . . : Data mapping error on member SRVCOLI.
Cause . . . . . : A data mapping error occurred on field
SRVCOLI_1.COLI_CHANGE_DATE in record number 0, record format *FIRST,
member
number 1, in member SRVCOLI file SRVCOLI in library LDATASV3, because of
error code 1. The error code meanings follow:
1 -- There is data in a decimal field that is not valid.
2 -- A significant digit was truncated.
3 -- A floating point value exceeded the maximum representable value.
4 -- A floating point value was less than the minimum representable
value.
5 -- A binary floating point value could not be used for the attempted
operation because it is not a number, or is not valid.
6 -- A floating point value could not be mapped to packed decimal,
zoned
More...
I know the very first plan of action should be figure out where this
invalid data is making its way into the file and my guess is a COBOL pgm,
but at risk of major judgement, we just don't have time to go find and fix
that. We have however looked into the issue and determined they are fields
that are simply initialized incorrectly and then written out to the file.
Our suspicions are that a COBOL pgm does it.
Now, we have a simple fix pgm we can run over the file when we encounter
this issue and it cleans the entire file. The fields essentially are
initialized correctly back to 0. The pgm is short and sweet and looks
like...
halwnull(*inputonly) fixnbr(*zoned)
fsrvcoli up e k disk
c update rcolst
That's all there is to it.
BUT I'm wanting to try and have SQL detect and handle the data during
processing so that processing is not interrupted.
If anyone has any experience with this issue and wish to share the
resolution I'd appreciate it.
Jay
As an Amazon Associate we earn from qualifying purchases.
This mailing list archive is Copyright 1997-2024 by midrange.com and David Gibbs as a compilation work. Use of the archive is restricted to research of a business or technical nature. Any other uses are prohibited. Full details are available on our policy page. If you have questions about this, please contact
[javascript protected email address].
Operating expenses for this site are earned using the Amazon Associate program and Google Adsense.