Reset observation pointer for the set statement

[gamotte]

Hello,

Consider the following program;

data have;
    input a;
    cards;
1
2
;
run;

proc sort data=sashelp.class out=class;
    by sex;
run;

data test;
    set have;

    do i=1 to nobs;
        set class nobs=nobs;
        by sex;
        if first.sex then total_weight=0;
        total_weight+weight;
        output;
    end;
run;

When execution reaches observation 2 of the have dataset, the whole sashelp.class has already been entirely read.

Thus the consecutive "set sashelp.class" will stop the data step.

I can not use the point= option because of by group processing.

Thanks for any advice to overcome this difficulty.

[data_null__]

What are you trying to do? Show data you want.

[gamotte]

Sorry for not being clear enough. Here is what the resulting dataset should look like :

a	i	Name	Sex	Age	Height	Weight	total_weight
1	1	Alice	F	13	56.5	84.0	84.0
1	2	Barbara	F	13	65.3	98.0	182.0
1	3	Carol	F	14	62.8	102.5	284.5
1	4	Jane	F	12	59.8	84.5	369.0
1	5	Janet	F	15	62.5	112.5	481.5
1	6	Joyce	F	11	51.3	50.5	532.0
1	7	Judy	F	14	64.3	90.0	622.0
1	8	Louise	F	12	56.3	77.0	699.0
1	9	Mary	F	15	66.5	112.0	811.0
1	10	Alfred	M	14	69.0	112.5	112.5
1	11	Henry	M	14	63.5	102.5	215.0
1	12	James	M	12	57.3	83.0	298.0
1	13	Jeffrey	M	13	62.5	84.0	382.0
1	14	John	M	12	59.0	99.5	481.5
1	15	Philip	M	16	72.0	150.0	631.5
1	16	Robert	M	12	64.8	128.0	759.5
1	17	Ronald	M	15	67.0	133.0	892.5
1	18	Thomas	M	11	57.5	85.0	977.5
1	19	William	M	15	66.5	112.0	1089.5
2	1	Alice	F	13	56.5	84.0	84.0
2	2	Barbara	F	13	65.3	98.0	182.0
2	3	Carol	F	14	62.8	102.5	284.5
2	4	Jane	F	12	59.8	84.5	369.0
2	5	Janet	F	15	62.5	112.5	481.5
2	6	Joyce	F	11	51.3	50.5	532.0
2	7	Judy	F	14	64.3	90.0	622.0
2	8	Louise	F	12	56.3	77.0	699.0
2	9	Mary	F	15	66.5	112.0	811.0
2	10	Alfred	M	14	69.0	112.5	112.5
2	11	Henry	M	14	63.5	102.5	215.0
2	12	James	M	12	57.3	83.0	298.0
2	13	Jeffrey	M	13	62.5	84.0	382.0
2	14	John	M	12	59.0	99.5	481.5
2	15	Philip	M	16	72.0	150.0	631.5
2	16	Robert	M	12	64.8	128.0	759.5
2	17	Ronald	M	15	67.0	133.0	892.5
2	18	Thomas	M	11	57.5	85.0	977.5
2	19	William	M	15	66.5	112.0	1089.5

[data_null__]

proc sort data=sashelp.class out=class;
   by sex;
   run;
data class;
   i = _n_;
   set class nobs=nobs;
   by sex;
   if first.sex then total_weight=0;
   total_weight+weight;
   run;
proc surveyselect noprint rep=2 rate=1 out=stacked(rename=(replicate=a));
   run;

[gamotte]

Thanks @data_null__ and @Astounding for your answers. I tried to simplify the problem in order to avoid distracting readers with unnecessary details  but i realize that it can give a misleading view of what i am trying to do.

The computation of sums only served as an example to justify a by statement. Actually, what is done in

the loop is dependant on the data in the currently processed row of the have dataset. That is,

for each row in have, i want to read the whole other dataset and create new variables that depend

on both datasets columns.

I just don't figure out how to restart reading the second dataset from observation 1 for each

new observation of the have dataset.

If you want the full context, i was trying to generalize my last answer in the following thread :

https://communities.sas.com/t5/SAS-Programming/Pairwise-comparisons-in-loops/m-p/497468#M131862

in order to take different sectors/quarters into account and stumbled upon this difficulty.

[Astounding]

It should be easy enough to summarize SASHELP.CLASS down to 2 observations (M and F) before bringing in HAVE.  Then you could use POINT=  (or possibly a sort and merge if that is a better choice).

[Astounding · Accepted solution]

For the more complex problem, I confess to not reading it through sufficiently.  But it should still be possible to use POINT= to read every observation without a BY statement.  You would need to set up multiple buckets for the calculations (analogous to one set of calculations for F and one for M) when looping through with POINT=.

[gamotte]

Thanks, i finally managed to overcome the difficulty. I needed in particular first.var and last.var created by the by statement so i used a prior data step to save them as new columns of the dataset. The by statement thus became useless and i could use the POINT= option.