let me try:

data test;
  input serial var $;
  datalines;
  1 v1
  2 v2
  3 v3
  ;
  run;

proc sql;
    create table tmp as(
    select var as newvar from test union all
    select cat(a.var,b.var) as newvar from test a,test b where  a.serial<b.serial union all
    select cat(a.var,b.var,c.var) as newvar from test a,test b,test c where  a.serial<b.serial and b.serial<c.serial);
quit;

data fstr;
     set tmp;
     serial +1;
run;

proc sql;
   select newvar into :str1 from fstr where serial=1;
   select newvar into :str2 from fstr where serial=2;
   select newvar into :str3 from fstr where serial=3;
   select newvar into :str4 from fstr where serial=4;
   select newvar into :str5 from fstr where serial=5;
   select newvar into :str6 from fstr where serial=6;
quit;

%put &str1;
%put &str2;
%put &str3;
%put &str4;
%put &str5;
%put &str6;

1. Functions allcomb, allperm
http://support.sas.com/documenta ... .htm#a003112302.htm
2. PROC PLAN

嗯，我后来用allcomb, 好像实现起来还比较快捷，你的方法也不错：

data _null_;
   array x[5] $3 ('ant' 'bee' 'cat' 'dog' 'ewe');
   n=dim(x);
   k=3;
   ncomb=comb(n,k);
   do j=1 to ncomb+1;
      call allcomb(j, k, of x

);
      put j 5. +3 x1-x3;
   end;
run;
SAS writes the following output to the log:
    1   ant bee cat
    2   ant bee ewe
    3   ant bee dog
    4   ant cat dog
    5   ant cat ewe
    6   ant dog ewe
    7   bee dog ewe
    8   bee dog cat
    9   bee ewe cat
   10   dog ewe cat
   11   dog ewe cat

嗯，谢谢，我后来自己找到了这种方法，挺不错的

使用allcomb()方程是非常美妙的程序.在以前我也有此需要,但是没有静下来考虑是否可以用宏来编写.今天决心试一试.

我的程序此处可见
但是,在运行超过计13个字符串的时候,有BUG.也许大家有什么意见.
京剧