Multiple set statements

[Babloo]

I ran the code below in EG 7.1 to understand how multiple set statements works. But I could not understand why data step (whilst creating data set c) read only 2 observations from dataset a although it has 3 observations. See the log which is highlighted in red color.

Also can I assume that we will get the output in final datastep which holds records with least number of observations in set statements? e.g. A table has 1 obs, B table has 2 obs, C table has 3 obs then D table (data d;set a;

set b;set c;run;) should produce 1 obs which is similar to A table. (?)

data a;

b="a";output;

b="b";output;

b="f";output;

run;

data b;

b="c";output;

b="d";output;

run;

data d;

b="e";output;

run;

data c;

set b;

set a;

set d;

run;

Log:

23         data a;

24         b="a";output;

25         b="b";output;

26         b="f";output;

27         run;

NOTE: Compression was disabled for data set WORK.A because compression overhead would increase the size of the data set.

NOTE: The data set WORK.A has 3 observations and 1 variables.

NOTE: DATA statement used (Total process time):

      real time           0.00 seconds

      cpu time            0.00 seconds

28        

29        

30         data b;

31         b="c";output;

32         b="d";output;

33         run;

NOTE: Compression was disabled for data set WORK.B because compression overhead would increase the size of the data set.

NOTE: The data set WORK.B has 2 observations and 1 variables.

NOTE: DATA statement used (Total process time):

      real time           0.00 seconds

      cpu time            0.00 seconds

34        

35         data d;

36         b="e";output;

37        

38         run;

NOTE: Compression was disabled for data set WORK.D because compression overhead would increase the size of the data set.

NOTE: The data set WORK.D has 1 observations and 1 variables.

NOTE: DATA statement used (Total process time):

      real time           0.00 seconds

      cpu time            0.00 seconds

39        

40         data c;

41         set b;

42         set a;

43         set d;

44         run;

NOTE: Compression was disabled for data set WORK.C because compression overhead would increase the size of the data set.

NOTE: There were 2 observations read from the data set WORK.B.

NOTE: There were 2 observations read from the data set WORK.A.

NOTE: There were 1 observations read from the data set WORK.D.

NOTE: The data set WORK.C has 1 observations and 1 variables.

[data_null__]

Perhaps you should first consider why WORK.C has only 1 observation. 

[Babloo]

That's also one off my question.

Also can I assume that we will get the output in final datastep which holds records with least number of observations in set statements? e.g. A table has 1 obs, B table has 2 obs, C table has 3 obs then D table (data d;set a;

set b;set c;run;) should produce 1 obs which is similar to A table. (?)

[data_null__]

Test.  Consider you first example with 1 obs output.  Which data set did the value of B come from. 

Your example is somewhat confounded by the fact that the variable(s) are the same in all.  The values are being overwritten before anything is output.  If you could run the data step debugger you would have good way to study exactly what is happening but you must have DMS SAS no EG/Batch.

[Babloo]

Dataset c returns the value for b as e. It is coming from the dataset d.

[gergely_batho]

Yes. At the first data step iteration:

1. SAS reads table b,  b="c" - but no output to table c

2. SAS reads table a,  b="a" - but no output to table c

1. SAS reads table b,  b="e" - now there is an output to table c

[data_null__]

Yes from data D.  Explain why that is.

To better understand add a variable to each data set with a unique name. 

[gergely_batho]

A data step exits immediately if it executes a set statement, but there are no more observations in the data set.

In the second data step iteration (_N_=2) SAS reads table b, than table a, and then tries to read table d, which has no more rows. ->  exit (stop; ) , no implicit output.

[gergely_batho]

By the way, this is the usual way, how a data step terminates. Run this code:

data class2;

     putlog 'BEFORE ' _all_;   /*_N_=20 is printed here*/

     set sashelp.class;           /*exit here, when _N_=20 - no more obs*/

     putlog 'AFTER' _all_;

run;

[Babloo]

I'm sorry, still I couldn't understand.

[Babloo]

Are you saying that output will be from the first set statement?

May I request you to explain it with small example.

[gergely_batho]

I augmented your code, so you can see in the log, what is happening.

I also put an output statement at the end of the data step (this changes nothing in this case, just instead of implicit output, we use explicit output)

data c;

putlog '1:' _all_;

set b;

putlog '2:' _all_;

set a;

putlog '3:' _all_;

set d;

putlog '4:' _all_;

putlog 'OUTPUT will be here';

output;

run;

[Babloo]

Thanks.

So the table which holds minimum number of observations from the list of set statements will be returned as output.

e.g. A table has 1 obs, B table has 2 obs, C table has 3 obs then D table (data d;set a;set b;set c;run;) should produce 1 obs which is similar to A table. (?)

Re: Multiple set statements

[data_null__]

Babloo wrote:

Thanks.

So the table which holds minimum number of observations from the list of set statements will be returned as output.

NO, that is not the correct way to understand what is happening.

[Kurt_Bremser]

No. The input table with the least number of observations determines after how many iterations the data step stops (and how many observations are written).

The output is determined by the last statement that filled the output data vector before the implicit write at the end of the data step iteration is reached.

So, in your first example, your variable b is ALWAYS filled with values from dataset d (because set d; is the last statement that alters the value of b prior to the write). Your output table will NEVER have values of tables a or b.

In your second example (e.g. A table has 1 obs, B table has 2 obs, C table has 3 obs then D table (data d;set a;set b;set c;run;) should produce 1 obs which is similar to A table. (?)), the values always come from table c, but table a determines the number of iterations (1).