DSXchange

Reverse your sort order. Last one in wins, period - so if you want the 'first' one you need to make sure it arrives last.

Thanks for your quick reply, but sorting is not possible because I have only 2 columns in hash file (member_key,sese_id) and member_key is key column.

chulett wrote:Reverse your sort order. Last one in wins, period - so if you want the 'first' one you need to make sure it arrives last.

Why isnt reverse sorting possible? Any who,
Another approach is , that you can write to a hashed file and reference the same hashed file. If the key is found dont write it again. Hence your constraint will be reflink.NOTFOUND.
Make sure your hashed file is locked for updates.

You don't - can't - have multiples of the key value in a hashed file. They way they work relies absolutely upon having unique keys.

Therefore, you will need a different approach - perhaps a secondary key lookup via a UV stage from a hashed file with an artificial key. If you do that, you can readily constrain for the first matching record (you will need to sort the result set, of course, using an ORDER BY clause - you can use FIRST 1 as the additional constraint).

rajkraj wrote:Thanks for your quick reply, but sorting is not possible because I have only 2 columns in hash file (member_key,sese_id) and member_key is key column.

Not sure what having two columns in the hashed file has to do with the ability to sort. You stated that when you build a hashed file and send more than one row for a particular key, the last one in is what is retained. That is correct due to the 'destructive overwrite' nature of hashed files.

You then stated that you wanted the first record for any particular key to be the one 'saved'. You do that a couple of different ways. One way is to ensure that the first record per key arrives last, hence the 'reverse your sort order' bit of advice. There's no reason that can't happen.

Or you can arrange for the first record per key to be the only record per key written. This requires the use of stage variables and the ability to detect changing 'groups' - in your case, only allow the first occurance of member_key to be written to the hashed file.

Either approach will solve your problem as I've understood it.

Thanks for your reply...but how would we do 'reverse sort' on a single key.

Ex:Member_key Id_key
1234 23
1234 24
1234 25

When we store this data into a Hashed file it will keep the last record but I want the first record.

I used order by while extracting the data from table but it's not changing the order

Please advise me

chulett wrote:

rajkraj wrote:Thanks for your quick reply, but sorting is not possible because I have only 2 columns in hash file (member_key,sese_id) and member_key is key column.

Not sure what having two columns in the hashed file has to do with the ability to sort. You stated that when you build a hashed file and send more than one row for a particular key, the last one in is what is retained. That is correct due to the 'destructive overwrite' nature of hashed files.

You then stated that you wanted the first record for any particular key to be the one 'saved'. You do that a couple of different ways. One way is to ensure that the first record per key arrives last, hence the 'reverse your sort order' bit of advice. There's no reason that can't happen.

Or you can arrange for the first record per key to be the only record per key written. This requires the use of stage variables and the ability to detect changing 'groups' - in your case, only allow the first occurance of member_key to be written to the hashed file.

Either approach will solve your problem as I've understood it.

So, from your eample:
You would like the Id_key of 23 to be stored for Member_key 1234? You've said you've done an order by, now you just need to make the secondary sort descending. Something like:
That should make them arrive like this:
And solve your problem. Yes?

Thanks for quick reply ....but my issue is still ther ...

Actually my data looks like this....

Member_key Id_key
1234 75
1234 abc
1234 %Agh
1234 =AVX

here order by is not useful...right....

chulett wrote:So, from your eample:

Code: Select all

Member_key Id_key 
1234        23 
1234        24 
1234        25 

You would like the Id_key of 23 to be stored for Member_key 1234? You've said you've done an order by, now you just need to make the secondary sort descending. Something like:

Code: Select all

ORDER BY Member_key, Id_key desc

That should make them arrive like this:

Code: Select all

Member_key Id_key 
1234        25
1234        24 
1234        23 

And solve your problem. Yes?

Did you try giving method a shot. Its easy and painless with no sorting involved. It will retain the first record of a key in the order it came in.

Would have been nice to have sample data that matched reality a little better.

You can still sort, yes. Just to beat this horsey a little longer...

Assign a record number to each row in a transformer, @INROWNUM will work nicely. Then use a Sort stage to sort ascending on your key field and descending on the record number. For example:
What will come out of the Sort stage will be:
It *can* be done.

"THANKS A LOT" It would work.


quote="chulett"]Would have been nice to have sample data that matched reality a little better.


You can still sort, yes. Just to beat this horsey a little longer...

Assign a record number to each row in a transformer, @INROWNUM will work nicely. Then use a Sort stage to sort ascending on your key field and descending on the record number. For example:
What will come out of the Sort stage will be:
It *can* be done.[/quote]