Datatype conversion after proc import

[arunrami]

HI All, 

Here is my scenario - I am using proc import , which will set the datatype based on its content , hence I have taken a copy of dataset using 'set' and tried to convert the datatype using only if it satisfies my 'if logic', but it seems still goes in to the if condition even though it doesn't satisfy my condition .

Below sample code , please help out how can I fix this . 

Note: variable inside if not supposed to be created as it doesn't pass the criteria ..

data A;
infile cards dlm=',' missover;
input name : $32. age $ ;
cards;
hawking ,20
einstein , 40
carl sagan , 30
;
run;

data dummy;
set A;
if vtype(age) = 'N' then do ;
age_1 =  input(age,best.) ;
end;
run;

[ed_sas_member]

Hi @arunrami 

I suggest that you use conditional processing with %if %then %end because the type is not a function at the "row" level but a metadata

proc sql noprint;
	select upcase(type) into: type_age trimmed from dictionary.columns
	where libname='WORK' and memname = 'A' and name='age';
quit;

%if "&type_age"="NUM" %then %do ;
	data dummy;
		set A;
		age_1 =  input(age,best.) ;
	run;
%end;

[Kurt_Bremser]

From which type of file are you importing?

And why do you convert to numeric when age is already numeric (vtype() = 'N')?

[arunrami]

It was just for an example , age is already a char variable , hence my point is , it should not enter in to if logic as it is not satisfying the logic if vtype() = 'N'.

But still its creating variable declared in side if - do..?I wanna know why is it and how PDV works in this case?

*I am using csv file as input

[ballardw]

---

@arunrami wrote:

HI All, 

Here is my scenario - I am using proc import , which will set the datatype based on its content , hence I have taken a copy of dataset using 'set' and tried to convert the datatype using only if it satisfies my 'if logic', but it seems still goes in to the if condition even though it doesn't satisfy my condition .

---

Reliance on Proc import, especially if reading data sources that are text such as CSV, is a poor practice. If you know you are going  be reading same structured data, which is sort of implied by fixing specific named variables, then better is to read them correctly in the first place using a data step. Then you control the value types at the beginning.