Splitting a SAS table into multiple tables bases on number of rows

[SasStatistics]

In reality I have a SAS table having very many rows, think milions, and I want to split it into sub-tables. The reason for this is that I will export the data and those who will process it use a system that cannot process so many rows at once, so it has to be split. 

For clarity I will provide an demonstrative example that would solve my problems. 

Assume I have a SAS table consisting of 334 rows. 

Now I want to split this table into 3 tables, since 334/3 is 111.3333 the function would return: 
- table1 consisting of 111 rows. 

- table2 consisting of 111 rows. 

- table3 consisting of 112 rows. 

So I just say that I want 3 tables and then it puts the suitable amount of observations in each new table (111, 111 and 112 rows in this case). 

Any ideas on how to achieve this? 

[tarheel13]

just make a counter variable and use that. you could do if 1 <= count <= 111 then output table1. else if count 111 < count <= 223 then output table2; else if count > 233 then output table3;

[Ksharp · Accepted solution]

proc surveyselect data=sashelp.heart groups=3 out=temp;
run;

data one two three;
 set temp;
 if GroupID=1 then output one;
 if GroupID=2 then output two;
 if GroupID=3 then output three;
run;

[Holmes]

---

@Ksharp  escreveu:

proc surveyselect data=sashelp.heart groups=3 out=temp;
run;

data one two three;
 set temp;
 if GroupID=1 then output one;
 if GroupID=2 then output two;
 if GroupID=3 then output three;
run;

---

Can I use this to make a PROC SQL split a query?

I have a long-lasting query and want to split the results in one million rows per time. 

How can I do this?

[Ksharp]

Please start a brand new session, this question is almost three years ago.
" want to split the results in one million rows per time. “
You need make id variable.Like:
data have;
set have;
id+1;
run;

proc sql;
create table part1 as
select * from have where id in (1:1000000);
............
quit;

[LinusH]

First check the no of records in the table. This can be achieved by PROC DATASETS/CONTENTS, data file functions and PROC SQL DICTIONARAY.TABLES.

Calculate the intervals, and put them into macro variables.

Use a data step, and explicitly output to your target tables depending on the value of the automatic _n_ variable compared with the macro variables above.

[Patrick]

IF you can fit a single slice of your table (&max_rows_per_table) into memory then below coding approach should work.

%let max_rows_per_table=5;

data have;
  set sashelp.class;
run;

data _null_;
  if _n_=1 then
    do;
      dcl hash h1(dataset:'have(obs=0)',multidata:'y');
      h1.defineKey('sex');
      h1.defineData(all:'y');
      h1.defineDone();
    end;
  set have end=last;
  h1.add();
  if mod(_n_,&max_rows_per_table)=0 or last then
    do;
      h1.output(dataset:cats('work.want_',put(_n_,z10.)));
      h1.clear();
    end;
run;

proc sql;
  select *
  from dictionary.tables
  where libname='WORK' and memname like 'WANT%'
  ;
quit;

The logic for naming the table slices could of course be different - like 1, 2, 3...

In the hash definition for h1.defineKey('<some variable>'); you can select any column that exists in your source table - but ideally choose the one with the lowest length as this will save some memory. 

Using hash tables is the only way I know of that allows you to create output tables dynamically during execution time. Any other approach will require some pre-processing to define the output tables (not that hard to do but will require some additional coding to generate the required syntax dynamically).

[ballardw]

No need to create data sets, just select the records desired from the source data set.

Pseudo code example to export records to different files in groups of 100:

proc export data=have (firstobs=1 obs=100)
    outfile=<your first file>
    <other options>
;
run;

proc export data=have (firstobs=101 obs=200)
    outfile=<your 2nd file>
    <other options>
;
run;

So you can use something like this to build the observation selection bit and a filename to plug into CALL EXECUTE code with boiler plate for the proc export options. Or create strings and write to a text program file to have documentation of what was created and run.

data _null_;
   do countobs = 1 to 3025 by 300;
      groupcount+1;
      string= catx(' ',"(firstobs=",countobs," obs=",groupcount*300,')');
      put string=;
      filename = cats("<path>\filebasename",put(groupcount,z3.),'.extension');
      put filename=;
   end;
run;

[Michael_Harper]

data one two three;
set SASHELP.baseball;

which= mod(_n_,3);
if which= 0 then output one;
else do;
if which < 2 then output two;
else output three;
end;
drop which;
run;

[PaigeMiller]

A solution using CALL EXECUTE for the data set SASHELP.CARS 

proc sql; /* Find number of observations in data set */
    select nobs into :nobs from sashelp.vtable where libname='SASHELP' and memname="CARS";
quit;
%let n_splits=5; /* User specifies number of splits desired */

data _null_;
    nobs_per_split=&nobs/&n_splits;
    do i=1 to &n_splits;
        start=round((i-1)*nobs_per_split+1);
        end=round(i*nobs_per_split);
        call execute(cat('proc export data=sashelp.cars(firstobs=',start,' obs=',end,') outfile="table',i,'.csv" dbms=csv; run;'));
    end;
run;

[Kurt_Bremser]

No need to split into datasets. Just use the FILEVAR= option of the FILE statement in the DATA step that does the export.

In which file format do you export your data?