How to prevent automatic conversion of date informat when facing partial date

[sunshinezb1990]

I recently found that SAS has its own "black box" when it detects partial date. The reason I call it 'black box' is that we don't know what it does when it reads in partial date.

Example 1:

1)  

data test;
a= "201201";
run;

data test2;
set test;
b= input(a, yymmdd8.);
run;

Result: b=22250 (This is equivalent to 2020Dec01), which means SAS automatically reads the date as yymmdd6. when it can't find the last 2 digits.

Example 2:

data test;
a= "201301";
run;

data test2;
set test;
b= input(a, yymmdd8.);
run;

Result: b= . (missing). Clearly, "201301" can't be read in as 'yymmdd6.' because there is no 13th month in a year. Then SAS finally gives it up and assigns missing value.

How do we make sure SAS doesn't do this auto conversion? Also who knows what this 'black box' actually does besides this situation?

Many thanks!

[art297]

I think the best answer to your question is: Don't make it guess!

If your examples were attempts to get sas to read dates that had a 4 digit year and a two digit month:

/* Example 1: */
data test;
  a= "201201";
run;
 
data test1;
  set test;
  format b date9.;
  b= input(a, yymmn6.);
run;
 
/* Example 2: */
data test;
  a= "201301";
run;
 
data test2;
set test;
  format b date9.;
  b= input(a, yymmn6.);
run;

Art, CEO, AnalystFinder.com

[mkeintz]

This is documented (i.e. it is demonstrated in the example) in the YYMMDDw. informat documentation.  The box is not as black as you might think. 

[sunshinezb1990]

Yeah, the example does answer my question. Thanks!