我也想问关于观测的一些特殊属性怎样对变量进行修改

data a;
input x y z h;
cards;
1 2 0 0
2 4 0 0
4 5 0 0
;
proc sql noprint;
select name into:vname separated by ' '
from dictionary.columns
where libname='WORK' and memname='A' and type='num';
quit;
%put &vname;
%macro zero;
%let i=1;
%do %while(%length(%scan(&vname,&i))>0);
%let name=%scan(&vname,&i);
proc sql noprint;         
select sum(&name) into :num
from a;
quit;
data a;
set a;
%if &num=0 %then %do;
drop &name;
%end;
run;
%let i=%eval(&i+1);
%end;
%mend;
%zero;

全部用DATA步完成，不需要调用复杂的宏程序以及SQL，效率要高些。
data crackman;
fen="woody";
wodu="sinacom";
do i=1 to 100;
x=ranuni(90);
y=x*1.567;
z=x+y;
m=0;
output;
end;
run;
data a;
set crackman;
keep _numeric_;
call symput('obs',compress(attrn(open('work.a','i'),'nobs')));
run;
data b;
set crackman;
keep _character_;
run;
proc transpose data=a out=c;
var _all_;
run;
data c;
set c;
array s col1-col&obs.;
if col1-col&obs.^=0 then output;
run;
proc transpose data=c out=c;
ID _NAME_;
run;
data c;
set c;
drop _name_;
run;
data result;
merge  c b;
run;

如何保证结果变量的顺序和原表一样呢？这个把字符型提前了

因为我的字符型变量值是一个值
所以转置的时候出现问题
如果不是一个值
也许可以直接转就可以了
不需要分字符型和数值型了