Zero Byte file Input problem

muruganr117 · Post by **muruganr117** » Mon Apr 30, 2007 8:59 pm

Hi ,

I wanna know, how is it able to write one record to output file, if the input file is a zero byte file, means there is no data in input file. I need one record to be written to output file. Both are sequential flat files. Please assist me in this regard

Thank you in advance

regards
Murugan

chulett · Post by **chulett** » Mon Apr 30, 2007 9:45 pm

You're right, an empty source file will process zero records in the job and then nothing will be written to the target file. What is special about this 'one record' that needs to be written when you have no input? Describe it, please.

muruganr117 · Post by **muruganr117** » Mon Apr 30, 2007 10:13 pm

chulett wrote:You're right, an empty source file will process zero records in the job and then nothing will be written to the target file. What is special about this 'one record' that needs to be written when you have no input? Describe it, please.

Hi,

Thank you for your reply, The output file provides information related to count and time stamp related stuff. Hence if the input file is a zero byte file, the output should have zero in the count and then time stamp, This will be the record written in to the output file.

regards
Murugan R

chulett · Post by **chulett** » Mon Apr 30, 2007 10:38 pm

So do you always write only one record to the output file? Or do you process all of the incoming data, write some/all to the output and then always append one more record to the file? More of a curiosity than anything.

I'd suggest you write an 'after job' routine. Pass it in the Stage Name of the Transformer in the job, the Output Link from the transformer that writes to the output file and the full path to the output file. You can use that information in a call to DSGetLinkInfo with LINKROWCOUNT to get your count of how many records were written out, including zero. And the DSJobStartTimestamp macro can get you a good timestamp for the activity unless you already have one. Then simply build the record and echo/cat/type it onto the end of your file with '>>' style redirection. That or get fancy with the OPENSEQ, SEEK, WRITESEQ (etc) BASIC functions. Or build a second DataStage job run after the first (via a Sequence job) to add this 'trailer' record to the file.

Food for thought.

JoshGeorge · Post by **JoshGeorge** » Sat May 05, 2007 3:48 am

A small routine - Check for file size and if zero byte write '0, current timestamp' into the file.

chulett · Post by **chulett** » Sat May 05, 2007 7:16 am

And if file has a header record for the column names?

A generic link row count check for zero always works.

JoshGeorge · Post by **JoshGeorge** » Sat May 05, 2007 7:36 pm

ONE record is the requirement

chulett · Post by **chulett** » Sat May 05, 2007 8:54 pm

You missed my point. Completely.

JoshGeorge · Post by **JoshGeorge** » Sun May 06, 2007 2:41 am

If the incoming file is zero byte it implies it won't have a header.
So if we check and find it is zero byte then we just have to write '0, current timestamp' to the output file (Append).

chulett · Post by **chulett** » Sun May 06, 2007 6:59 am

An 'empty' file can also have a single header record, in which case it will not be zero bytes in size but still process zero rows. My advise was more for the masses, general advice for determining emptiness and not specific to this issue where the OP 'knows' empty does in fact = zero bytes.

JoshGeorge · Post by **JoshGeorge** » Sun May 06, 2007 7:42 am

Was just making my point. Yep for 'masses'

Didn't say that it is not a good idea.

muruganr117 · Post by **muruganr117** » Wed May 09, 2007 9:33 am

chulett wrote:So do you always write only one record to the output file? Or do you process all of the incoming data, write some/all to the output and then always append one more record to the file? More of a curiosity than anything.

I'd suggest you write an 'after job' routine. Pass it in the Stage Name of the Transformer in the job, the Output Link from the transformer that writes to the output file and the full path to the output file. You can use that information in a call to DSGetLinkInfo with LINKROWCOUNT to get your count of how many records were written out, including zero. And the DSJobStartTimestamp macro can get you a good timestamp for the activity unless you already have one. Then simply build the record and echo/cat/type it onto the end of your file with '>>' style redirection. That or get fancy with the OPENSEQ, SEEK, WRITESEQ (etc) BASIC functions. Or build a second DataStage job run after the first (via a Sequence job) to add this 'trailer' record to the file.

Food for thought.

Hi Craig..

Thank you for your suggestions. I need to get a understanding of a point stated. Do you mean to say that code a routine to write a record into a file as in UNIX style?

DSXchange

Zero Byte file Input problem

Zero Byte file Input problem

Re: Zero Byte file Input problem