MVSFORUMS.com

[misi01]

I don't actually have a need for a sort solution as yet, but the final results will (I'm sure) need it at a later date.

My thought was to explain the problem we have and hopefully, with your help come up with a solution that will work easily using DFSORT.

Okay, background.

We have a program that creates a file (the yearly bank details for all our customers).
The first 50 bytes in this file contains details that are used to sort it and then split it into smaller files.
These details contain information such as zip code, how many pages are needed for each letter etc.
All of this, so that we save money vis-a-vis the post office if the letters are sorted in zip-code order.
In addition, how many pages there are in each letter, determines the post-processing in the automated "kuvertering" (don't know the word in English, but it's about automatically putting the letters into envelopes).

Present day.

The program that creates the first file above has been converted to create an XML file instead (on the mainframe).

Problem.

The XML file (at the moment) obviously doesn't have these 50 bytes in it.
We need a way of facilitating the sort and split of the file.

Obviously, we could pre-pend those 50 bytes to each record in the XML file, but this has the following problems:-

1 The XML file is no longer readable using standard XML s/w (such as XML notepad)
2 The XML file contains x leading records with namespace tags and any split of the file would need to include those first records in each output file.
3 The XML file also contains the closing tag for the "major" tag - this record would need to be appended to each output file (though I imagine any DFSORT solution could include some hard-coded value for this closing tag)

Another option might be to create a second file that only contains the equivalent of the leading 50 bytes for each record and then use DFSORT to merge and split the files based on this second file (somehow or other).

Another option would be to create the XML file as it is at the moment and also create the same file with the 50 bytes pre-pended for each record (the first file is then readable as XML, the second file is used for the sort and split)

Another option would be .......

Any thoughts/suggestions would be gratefully appreciated.
_________________
Michael

[William Collins]

Is your XML "horizontal" (one long record) or "vertical" (lots of physical records for one logical record)?

How about just adding the 50 bytes as a new XML element? A bit like using the FILLER at the end of a record (except you'd prefer this one to be first, so that it is in a fixed location) it should/may not affect other processing, as everything should be extracted by name, rather than fixed or relative position.

[misi01]

When creating the file, we set a flag to indicate whether the file should be split, one tag per record or created with as many tags as will fit into 256 bytes.

Your idea is interesting. Basically (?) create a dummy tag containing the 50 bye sort information.

DFSORT would then have to a sort the file based on this tag (as well as all tags following it up to the next 50-byte tag). Although writing the sort parms is outside my competence, I would be prepared to Google and experiment on how to do it.

The advantage of your idea is that the XML file is still readable (albeit with an extra "weird" tag).

I'll certainly look into it. Thanx for the idea.
_________________
Michael

[misi01]

BTW. Can you please give me the DFSORT keyword I need to look at so as to split the file based on this tag

(Note, I'm not asking for a solution, only a pointer as to where to start looking)
_________________
Michael

[William Collins]

To create multiple files, you use multiple OUTFIL statements. There you can use INCLUDE= or OMIT=, and also for one, SAVE. The INCLUDE= and OMIT= are similar to the INCLUDE/OMIT COND=, except they work on the final output data.

[misi01]

I should have asked for the DFSORT keywords to sort the actual file

Googling, I'm guessing I need some sort of combination of

.. IFTHEN=(WHEN=GROUP

or similar
_________________
Michael

[William Collins]

Yes, if you have multiple physical records, you're going to need WHEN=GROUP to get the records to sort together by PUSHing the key to a temporary extension. After the SORT, in OUTREC or OUTFIL, you can cut the records back down to the original data with BUILD or with IFOUTLEN if you have further handy IFTHENs.

[misi01]

I'll try looking at it this coming week. For those who "speak" fluent DFSORT, the solution will probably be pretty easy, but if I get it working, I'll post the solution here anyway
_________________
Michael

[misi01]

Here's my file (the first few records)

[misi01]

Okay, I got the group ID indicators in the file using

[misi01]

After a lot of experimenting, I arrived at the following solution (this doesn't include the BUILD, but that's the last, simple part).

[William Collins]

If your input is variable-length, then you should extend at the beginning of the record.

[kolusu]

misi01,

As william pointed out you just need to copy the key you want to sort after the RDW.

Assuming your sort key is at position 18 for length of 16 which is 58221 0007056702

All you need is to push that key using WHEN=GROUP and then remove it after sorting.

[misi01]

William, Kolusu. I think I've understood what you meant. Here's my code (I've added comments so there should be no doubt as to what I think I'm doing)

[misi01]

Okay, got it !!!!