topic Re: add column for output method Hash Object in SAS Programming

add column for output method Hash Object

mariopellegrini — Wed, 22 Feb 2023 10:17:20 GMT

Good morning. I would like to use the "output" method of hash sas, but I need to add the "key" column and display it on the "class_hash" output dataset. I can't figure out how to do it. I give the example:

data large;
   if 0 then set sashelp.class;
   declare hash obj(dataset: 'sashelp.class', multidata:'y');
   obj.definekey ("name" );
   obj.defineData(all:'yes');
   obj.definedone();
   length key $ 200;
   key = catt(name,sex);
   obj.output(dataset:'class_hash');
   stop;
run;

Mario

Re: add column for output method Hash Object

Patrick — Wed, 22 Feb 2023 10:25:18 GMT

The output method can only output what's in the hash defined as data.

The hash key column name is with all:"yes" also part of the data and will be included.

Column key which you create in your SAS data step is not in the hash table and though not available when writing the hash table in memory to disk.

Not sure if that's going to be of any use for your real use case but below one option how to add another variable to the hash.

data v_demo / view=v_demo;
  set sashelp.class;
   length key $ 200;
   key = catt(name,sex);
run;

data large;
   if 0 then set v_demo;
   declare hash obj(dataset: 'v_demo', multidata:'y');
   obj.definekey ("name" );
   obj.defineData(all:'yes');
   obj.definedone();
   length key $ 200;
   key = catt(name,sex);
   obj.output(dataset:'class_hash');
   stop;
run;

proc print data=class_hash;
run;

Re: add column for output method Hash Object

mariopellegrini — Wed, 22 Feb 2023 11:36:08 GMT

Thank you, but this way I don't get any advantage in terms of processing time. I was hoping you could save processing time compared to a simple data or proc sql step, using hash "output" method

Re: add column for output method Hash Object

Patrick — Wed, 22 Feb 2023 11:43:43 GMT

Loading and then writing a hash table is still a read/write operation so for your use case wouldn't offer any advantage over a simple data step.

Using the hash object offers often real advantages for lookups as it helps to avoid sorts.