Author |
Message |
dwh_user

Group memberships: Premium Members
Joined: 18 Oct 2018
Posts: 14
Location: Sydney
Points: 168
|
|
DataStage® Release: 11x |
Job Type: Parallel |
OS: Windows |
|
Hi Experts,
I have encountered an issue while profiling flat files in information Analyzer v11.5. A field which is more than 1000 characters is being Inferred as STRING 255.
The analysis setting to “Include columns with length over 512 characters” is enabled. Also, I couldn’t edit the defined metadata for this field as part of the import process to make the length longer than 255.
I have found some IBM documentation related to this:
https://www-01.ibm.com/support/docview.wss?uid=swg21698836
https://www-01.ibm.com/support/docview.wss?uid=swg1JR56739
Is this the default behaviour of IA v11.5, where a STRING longer than 255 characters is Inferred as STRING 255? Or is there any setting that has to be changed where the inferred results are accurate for strings longer than 255 characters?
Thanks in advance
|
|
|
|
 |
chulett
 since January 2006
Group memberships: Premium Members, Inner Circle, Server to Parallel Transition Group
Joined: 12 Nov 2002
Posts: 43085
Location: Denver, CO
Points: 222463
|
|
|
|
|
|
FYI - site has been a bit cranky lately, finally able to get in and reply. Also fixed your links so they work now. Neither seems to apply here and the second one notes "Issue is fixed with APAR JR56739 and is delivered with Governance Rollup5" in 2016. If you haven't already, open a case with your official support provider to check.
Never used this tool but going to make an educated guess based on general sequential file metadata handling across the suite. From what I recall - if what it sees during the import looks like a small string, it will treat it as a CHAR rather than a VARCHAR and you'll get the STRING(255) you are seeing... and it only looks at a handful of records to decide what to do. Are your longer values in ALL records or only some records? If the latter, I'd import one where they were right up front and if you don't have one like that, engineer one.
As a general statement, I'm not sure why you can't "edit the defined metadata" if it's not correct. We had to do that all the time when the import didn't get it
quite
right, be it flat files or Oracle tables or whatever. A people will always be better at inferring metadata than some silly program.
|
_________________ -craig
"You can never have too many knives" -- Logan Nine Fingers
|
|
|
 |
stuartjvnorton
Participant
Joined: 19 Apr 2007
Posts: 528
Location: Melbourne
Points: 3927
|
|
|
|
|
|
Out of curiosity, what is the maximum length of the actual data in your field?
|
|
|
|
 |
dwh_user

Group memberships: Premium Members
Joined: 18 Oct 2018
Posts: 14
Location: Sydney
Points: 168
|
|
|
|
|
|
The maximum length of the string in the field was around 3000.
|
|
|
|
 |
ray.wurlod
Participant
Group memberships: Premium Members, Inner Circle, Australia Usergroup, Server to Parallel Transition Group
Joined: 23 Oct 2002
Posts: 54601
Location: Sydney, Australia
Points: 296091
|
|
|
|
|
|
All strings in delimited text files are reported as length 255, unless you change that fact in QETXT.INI. And that, curiously, limits what can be inferred. I agree that it shouldn't.
|
_________________ IBM Software Services Group
Any contribution to this forum is my own opinion and does not necessarily reflect any position that IBM may hold. |
|
|
 |
|