DSXchange

Thanks Craig.

Remember:

If you compare "10000 12345" against "2 12345" then you are doing a STRING compare. However, day 2 is January 2, 1968 and day 10000 is May 17th, 1995. So, you will not sort correctly, nor will you have a valid > or < comparison.

What are you doing with the duplicates, are you just taking the last one? Or are you "rolling" them into a final row (meaning insert then update, update, update)? What is your volume relative to your environment (small, medium, large, extreme)?

HASHa --> XFMa --> HASHb --> XFMb --> LINKCOLLECTOR --> AGG --> SEQ HASHc --> XFMc --> Use your target file metadata in the XFM? stages. Map each hash file into that metadata, putting @NULL as the derivations for the columns that don't come from that hash file. In the AGGregator stage, group by the...

No, no, no. You have been steered wrong and must not be watching the previous posts. ICONV will turn a date in to the integer number of days since Jan 1, 1968. It turns time into the number of seconds since midnight. Concatenating those two values together will NOT give you a number you can do equiv...

Yes, but is the data in a text file, a hash file, or a table? You state duplicate composite primary keys, which by definition can't be duplicated in a table. In a hash file, you would only have one occurrence, the last occurrence written to the hash file under that primary key. So, I'm guessing that...

chulett wrote:

kcbland wrote:The fact that the entire world has failed to comprehend this elegance mystifies me.

Well... not quite the entire world.

By entire world I mean ALL SOFTWARE MAKERS!!! Arrrggghhh, standards, please. Meaningful and consistently applied, please please please.

Describe what you are trying to do, then maybe we can suggest something. Duplicate rows, duplicate primary keys, what is your problem?

It doesn't keep track of date and time components. Everything to DS Server is a STRING. The point is that the ISO format will allow you to STRING sorts in any situation. Experiment with it in a text file and use the sort command from Unix. You'll see that it works! The same thing applies in SQL and ...

Yep. One SEQ stage to output the contents of the file to a XFM stage to map it to the input of another SEQ stage. The XFM stage applies whatever changes you need to the data content.

Nahh, use ISO formatted timestamps and you have a natural lexicographic sort capability. YYYY-MM-DD HH24:MI:SS.xxxxxx will always do a left-justified sort and give you the correct answer. The fact that the entire world has failed to comprehend this elegance mystifies me. You can sort a timestamp fie...

Dynamic hash files are binary data structures, you cannot simply do text file commands against them. If you want to run SQL against a hash file, then search the forum for the ways to do that. If your hash file exists within a project (not fully qualified path), then use the DS Admin client and you c...

To "update" a file you must read the existing file and write a new one. Do not attempt to read and write to the same sequential file, as it won't work. Simply design a job of SEQ --> XFM --> SEQ.

If your data is presented to you thru a database, why not use the SQL ORDER BY clause? If your data is presented to you as a sequential file, why not pre-sort it using the unix sort command with a -k option to specify the sort key and redirect the results to a sorted file? You could always use the S...

Maybe I'm just dense this morning, but I'm still not following. If you have 5 hash files with a single column, then that column MUST be the primary key. If you're "trying to merge all these files into a single file", then you need to in effect concatenate and then unique the results. Now you talk ab...

Can you please post an example, as I cannot follow what you are trying to do.