Clearing Hash files

[swwilbert]

I have a process that has multiple inputs from a single transformer into a hash file. I have set the transformer link ordering. On the first input link into the hash file, I have the hash properties checked: Create File & in options, Delete file. The 2nd & 3rd input, nothing is checked.

I am trying to ensure that the hash file will be cleared by the first link and the 2nd & 3rd will append data to the hash file.

Is this the proper way to go about this? Any one know of a better way?

Thanks in advance

Steve

[kcbland]

It will work fine as you've described, as long as you don't have any IPC stages between the Transformer and hashed files or inter-process buffering turned on.

[ray.wurlod]

... and provided that row #1 is directed to output link #1; not prevented by the constraint expression.

[Abu@0403]

I think that this is not the way actually the job would work.

Even though link ordering is mentioned in the transformer. Its not like, first all the records of the first link would be loaded into the hash file followed by the second and the third.

Based on the input the loading process would differ.

I think that the link ordering will make sure that the first record from input goes to the first link and then to the next based on link ordering. Similarly the second input would do so in the same manner based on the constraints.

[ray.wurlod]

But that wasn't the question.

To be absolutely sure that the hashed file is cleared use a before-stage or before-job subroutine ExecTCL to clear the hashed file. The command is CLEAR.FILE hashedfilename (which requires a VOC entry for hashedfilename).

[chulett]

Or link a standalone Transformer to it to perform the Clear. Create a stage variable so it will compile, use the proper metadata (doesn't matter what's in the derivation, I use @NULL) set the constraint to @FALSE and enable the 'Clear' option. It will run first and clear the hashed file.