Hash Output to update the Initial Dataset

[jpm2478]

Hi,

I'm using Hash Object and I'm trying to get the Output to overwrite the initial dataset (Y). I'm using  if last then h.output(dataset:'work.y');

where work.Y is an initial dataset.

When I run this code the Output is not shown in Work.Y but when I use a newdataset name like  if last then h.output(dataset:'AA'); the output is shown in new dataset AA.

How do I get Hash Output to be shown on the Initial Dataset.

[SuryaKiran]

Are you trying to update a dataset using the values from other dataset? 

Try this if so:

data have;  
  if _n_=1 then do;
    declare hash ud(dataset:'update');
    ud.defineKey('id');
    ud.defineData('val1', 'val2');
    ud.defineDone();
  end;
  modify have;
  rcUpdate = ud.find();
  if rcUpdate=0 then replace;
run;

[novinosrin]

Hi @jpm2478 If your question relates to the thread https://communities.sas.com/t5/SAS-Procedures/Compare-two-datasets-using-Hash-and-update-values/m-p/...

You can try options 1. modify statement, 2. @ballardw 's non hash solution among others etc

Of course if you are comfortable with hashes, by all means do explore. Hash understanding is tricky for the reason understanding 

PDV::Memory table   key value pair matches can be confusing and pulling from Memory table ::  PDV along with it can be daunting. But  totally up to you

[mkeintz]

Are you trying to use hash output method to overwrite a data set that your are reading from in the same data step?   I don't believe you can do that.

[novinosrin]

@mkeintz Sir, I believe that's exactly OP's req. I wish I was brilliant using modify statement and automatic vars associated with it like you demonstrated in one your posts which I regret I failed to add to my notes. If i had it, I would have plagiarized it and demonstrated here3

[mkeintz]

@novinosrin

You may be referring to Update and replace data in a table, but that problem was only to decide whether to replace the last record in the master, or to append the record.  I think this problem has to do with multiple transaction/update records.  In which case @SuryaKiran's solution seems apt - as long as the UPDATE dataset has only one record per id.

edited revision:  ... and also assuming that there are no new id's in UPDATE.