So that is different than the technical question of what %SYSRC() does.
Look at the whole data step.
First thing you will notice it will end the way most data steps end, when it reads past the end of the input.
data sdata.latest_customer:
set curr_customer;
In this case when the SET statement runs out of data to read from curr_customer.
Now let's look at the DO loop.
do until(_IORC_=%SYSRC(_DSENOM));
Since it is a DO UNTIL() loop the DO loop will run at least once for every observation read from curr_customer.
It then tries to use the MODIFY statement with the KEY= option.
modify sdata.latest_customer key=i_customer_no;
So that is going to do a few things. One is lets the data step know that the dataset listed in the DATA statement is not being built brand new, but it is just going to be modified in-place. Second is that it will search for the observation to point to based on the index named in the KEY= option. So whatever value for the variables that are used to make the index were read from CURR_CUSTOMER is going to be used to find the record in SDATA.LATEST_CUSTOMER. And finally it is going to set the _IORC_ variable (and also the _ERROR_ variable) to indicate if the lookup in the index succeeded or not.
Now the SELECT block is going to decide what to do based on the _IORC_ value set by the MODIFY KEY= statement.
When the record is FOUND ( the value assigned to the _SOK pneumonic ) then it potentially updates the eff_to_date variable based on the value or the t_eff_to_date variable read from CURR_CUSTOMER.
And when the value is not found ( the value assigned to the _DSENOM pneumonic you asked about before) then instead it just resets the _ERROR_ flag to FALSE. That will prevent the data step from crashing as soon it one of the values read from CURR_COSTEMER is not found.
Now the UNTIL() clause is tested. So it didn't find a record the DO loop ends. If it did (or it got some other error) then the loop continues.
The purpose of the DO UNTIL() loop is to let you update multiple observations in sdata.latest_customer from a single observation read from CURR_CUSTOMER. For that to actually happen the INDEX needs to be defined without the UNIQUE option.
Here is an example that uses the SASHELP.CLASS dataset and AGE as the key variable for the INDEX.
data class (index=(age));
set sashelp.class;
run;
data class ;
input age new_name $ ;
put _n_= age= new_name= _iorc_= ;
do loop=1 by 1 until (_iorc_=%sysrc(_dsenom)) ;
modify class key=age ;
put _n_= loop= _iorc_= name= sex=;
select(_IORC_);
when (%SYSRC(_SOK)) do;
name=new_name;
replace class;
end;
when (%SYSRC(_DSENOM)) DO;
_error_=0;
end;
otherwise;
end;
end;
cards;
14 Zebra
99 Giraffe
;
proc print data=class;
run;
So you can see that the one transaction record with AGE=14 changed the name on multiple observations in the WORK.CLASS dataset since the DO loop ran multiple times. But the AGE=99 transaction is ignored since it did not match anything in the WORK.CLASS dataset.
