Variable length statement not working

[claremc]

Hello, 

I am trying to cut down the length of my variables to save disc space and the usual length statement I use does not seem to be working. 

Here is my code: 

data new_union;
length ss_other_specify $255. Manufacturer1 $50. Manufacturer2 $50. Vaccination_Code_Id1 $50. Vaccination_Code_Id2 $50.;
set union_combine;

/*Vax code format*/
if Vaccination_Code_Id1 in ("No vaccine administered", "Other", "Unknown vaccine or immune globulin") 
then Vaccination_Code_Id1 = ""; 
if Vaccination_Code_Id2 in ("No vaccine administered", "Other", "Unknown vaccine or immune globulin") 
then Vaccination_Code_Id2 = ""; 

/*Vax date format*/
Vaccination_Date1_d= mdy(substr(Vaccination_Date1,6,2), substr(Vaccination_Date1,9,2), substr(Vaccination_Date1,1,4)); 
Vaccination_Date2_d= mdy(substr(Vaccination_Date2,6,2), substr(Vaccination_Date2,9,2), substr(Vaccination_Date2,1,4)); 
drop Vaccination_Date1; 
drop Vaccination_Date2; 

format Vaccination_Date1_d mmddyy10. Vaccination_Date2_d mmddyy10.; 
run; 

data new_union; 
set new_union; 

rename Vaccination_Date1_d = Vaccination_Date1; 
rename Vaccination_Date2_d = Vaccination_Date2; 
format Vaccination_Date1 mmddyy10. Vaccination_Date2 mmddyy10.; 

/*Manufacturer reformat*/
Manufacturer1 = upcase(Manufacturer1); 
Manufacturer2 = upcase(Manufacturer2); 
if Manufacturer1 = "JOHNSON AND JOHNSON" then Manufacturer1 = "JANSSEN (JOHNSON AND JOHNSON)";
if Manufacturer1 = "PFIZER" then Manufacturer1 = "PFIZER/BIONTECH";
if Manufacturer2 = "JOHNSON AND JOHNSON" then Manufacturer2 = "JANSSEN (JOHNSON AND JOHNSON)";
if Manufacturer2 = "PFIZER" then Manufacturer2 = "PFIZER/BIONTECH";

/* if case_id in ("100734763","100748803") then Manufacturer2 = "" and Vaccination_Date2 = . and Vaccination_Code_Id2 = "";  */
/* if Manufacturer2 = "0" then Manufacturer2 = "";  */
run; 

After I run proc contents on the "new_union" dataset, the variable length is still $350, not $50 like I would like. I am running 2 separate data steps because I've had trouble with keeping them all in one when I use drop and rename. 

Thanks for any help! 

[Rydhm · Accepted solution]

When you create a new dataset from existing dataset, then you should create new variable names. Length doesn't change on existing variables. 

In your case, Vaccination_Code_Id1 or 2 both are in union_combine dataset. So try creating somewhat similar but different names in length statement as well as using them in code.

[claremc]

Thank you! That worked

[Astounding]

Your program should have been effective at cutting down the length.  Are you sure you ran the PROC CONTENTS on the new data set, not the old?

[SASJedi]

I made some fake data like this:

data union_combine;
   length id 8 ss_other_specify $255. Vaccination_Date1 $10
          Manufacturer1 $350. Manufacturer2 $350. Vaccination_Code_Id1 $350. Vaccination_Code_Id2 $350.;
   retain Manufacturer1 "Big Pharma 1" Manufacturer2 "Big Pharma 2";
   do id=1 to 4;
      Vaccination_Code_Id1 =choosec(ID,"No vaccine administered", "Other", "Unknown vaccine or immune globulin"," "); 
      Vaccination_Code_Id2 =choosec(abs(5-ID),"No vaccine administered", "Other", "Unknown vaccine or immune globulin"," "); 
      Vaccination_Date1='----------';
      Vaccination_Date2='----------';
      substr(Vaccination_Date1,6,2)='07';
      substr(Vaccination_Date1,9,2)=put(ID,z2.);
      substr(Vaccination_Date1,1,4)='2021'; 
      substr(Vaccination_Date2,6,2)='07';
      substr(Vaccination_Date2,9,2)=put(ID+14,z2.);
      substr(Vaccination_Date2,1,4)='2021'; 
      Manufacturer1 = choosec(ID,"JOHNSON AND JOHNSON","PFIZER","JOHNSON AND JOHNSON","PFIZER");
      Manufacturer2 = choosec(ID,"PFIZER","JOHNSON AND JOHNSON","PFIZER","JOHNSON AND JOHNSON");
      output;
   end;
run;

Then I ran this single data step that did all the conversions you were doing:

data new_union;
   length ss_other_specify $255. Manufacturer1 $50. Manufacturer2 $50. Vaccination_Code_Id1 $50. Vaccination_Code_Id2 $50.;
   set union_combine (rename=(Vaccination_Date1 = _dt1 Vaccination_Date2 = _dt2)); 
   /*Vax code format*/
   if Vaccination_Code_Id1 in ("No vaccine administered", "Other", "Unknown vaccine or immune globulin") 
   then Vaccination_Code_Id1 = ""; 
   if Vaccination_Code_Id2 in ("No vaccine administered", "Other", "Unknown vaccine or immune globulin") 
   then Vaccination_Code_Id2 = ""; 

   /*Vax date format*/
   Vaccination_Date1= mdy(substr(_dt1,6,2), substr(_dt1,9,2), substr(_dt1,1,4)); 
   Vaccination_Date2= mdy(substr(_dt2,6,2), substr(_dt2,9,2), substr(_dt2,1,4)); 
   Manufacturer1 = upcase(Manufacturer1); 
   Manufacturer2 = upcase(Manufacturer2); 
   if Manufacturer1 = "JOHNSON AND JOHNSON" then Manufacturer1 = "JANSSEN (JOHNSON AND JOHNSON)";
   if Manufacturer1 = "PFIZER" then Manufacturer1 = "PFIZER/BIONTECH";
   if Manufacturer2 = "JOHNSON AND JOHNSON" then Manufacturer2 = "JANSSEN (JOHNSON AND JOHNSON)";
   if Manufacturer2 = "PFIZER" then Manufacturer2 = "PFIZER/BIONTECH";
   drop _:; 
   format Vaccination_Date1 mmddyy10. Vaccination_Date2 mmddyy10.; 
run; 

All the variable sizes I changed to $50 worked just fine:

proc contents data=new_union; 
run;

Alphabetic List of Variables and Attributes
#	Variable	Type	Len	Format
2	Manufacturer1	Char	50
3	Manufacturer2	Char	50
4	Vaccination_Code_Id1	Char	50
5	Vaccination_Code_Id2	Char	50
7	Vaccination_Date1	Num	8	MMDDYY10.
8	Vaccination_Date2	Num	8	MMDDYY10.
6	id	Num	8
1	ss_other_specify	Char	255

Can you share your results?

[ballardw]

For which specific variable? You have 4 variables where you might be setting the length to 50.

You can avoid all of the character to numeric conversion messages you get from using this code:

/*Vax date format*/
Vaccination_Date1_d= mdy(substr(Vaccination_Date1,6,2), substr(Vaccination_Date1,9,2), substr(Vaccination_Date1,1,4)); 
Vaccination_Date2_d= mdy(substr(Vaccination_Date2,6,2), substr(Vaccination_Date2,9,2), substr(Vaccination_Date2,1,4)); 

by using the proper INFORMAT on the entire string.

Data example;
   x='21020715';
   y=input(x,yymmdd8.);
   format y mmddyy10.;
run;

Or if your Vaccination_date variables are simple numbers convert to string and use informat

Data example2;
   x=21020715;
   y=input(put(x,f8.),yymmdd8.);
   format y mmddyy10.;
run;

[Ksharp]

data have;
set sashelp.class;
run;

proc sql;
alter table have
modify sex char(20),name char(10);
quit;