SAS 9.3 bug?

[TD]

Hi

I'm experiencing weird SAS behavior:

data a;
/*    uncomment for input to work in the next data step*/
/*    length date1 $20; */
    date1 = '2015-10-26T23:59';
    date2 = '2015-10-26';
run;

data b;
    set a;
    not_related = trim(left(strip(date2)))||'T23:59';
    dt = input(date1, e8601dt.);
run;

the second datastep produces a note that input has invalid argument. However, if in first datastep I set date1 length to $20, then the second step has no problem to run and input date/time.

Tested on SAS 9.3 TS Level 1M2 X64_ES08R2.

Anyone else seen this before?

Thanks

[RW9]

I don't get this, using 9.4:

299 data a;
300 /* uncomment for input to work in the next data step*/
301 /* length date1 $20; */
302 date1 = '2015-10-26T23:59';
303 date2 = '2015-10-26';
304 run;

NOTE: The data set WORK.A has 1 observations and 2 variables.
NOTE: DATA statement used (Total process time):
real time 0.01 seconds
cpu time 0.01 seconds

305
306 data b;
307 set a;
308 not_related = trim(left(strip(date2)))||'T23:59';
309 dt = input(date1, e8601dt.);
310 run;

NOTE: There were 1 observations read from the data set WORK.A.
NOTE: The data set WORK.B has 1 observations and 4 variables.
NOTE: DATA statement used (Total process time):
real time 0.04 seconds
cpu time 0.00 seconds

What does your log look like?

[TD]

1    data a;
2    /*    uncomment for input to work in the next data step*/
3    /*    length date1 $20; */
4        date1 = '2015-10-26T23:59';
5        date2 = '2015-10-26';
6    run;

NOTE: The data set WORK.A has 1 observations and 2 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      user cpu time       0.00 seconds
      system cpu time     0.01 seconds
7
8    data b;
9        set a;
10       dt = input(date1, e8601dt.);
11   run;

NOTE: Invalid argument to function INPUT at line 10 column 10.
date1=2015-10-26T23:59 date2=2015-10-26 dt=. _ERROR_=1 _N_=1
NOTE: Mathematical operations could not be performed at the following places. The results of the
      operations have been set to missing values.
      Each place is given by: (Number of times) at (Line):(Column).
      1 at 10:10
NOTE: There were 1 observations read from the data set WORK.A.
NOTE: The data set WORK.B has 1 observations and 3 variables.
NOTE: DATA statement used (Total process time):
      real time           0.01 seconds
      user cpu time       0.00 seconds
      system cpu time     0.01 seconds

[RW9]

Well, I don't have 9.3 to hand and can't find a page on changes, but it may be that in 9.4 they relaxed the requirements on the string.  Does adding the seconds part change this, I remember I always used to have to add ":00" to the end to get the format to work for me.

[TD]

I think it's a SAS 9.3 bug. Adding seconds ':00' works the same way as increasing date1 length to $19 - input is performed correctly (as expected). Must be some kind of memory optimisation bug - because changing value of date2 to '2' also lets to input date1 without an error.

Just noticed, that if I initialize date2 before date1, the input works as expected.

Thanks

[ballardw]

When I read the documentation for E8601dt informat is says default length is 19 and minimum length is 19, ie. E8601dt19. max length 26.

When you use

date1 = '2015-10-26T23:59'

the length of date1 is 16, which is too short generating the error.

This doesn't generate the error.

data a;
/*    uncomment for input to work in the next data step*/
/*    length date1 $20; */
    date1 = '2015-10-26T23:59:00';
    date2 = '2015-10-26';
run;

data b;
    set a;
    not_related = trim(left(strip(date2)))||'T23:59';
    dt = input(date1, e8601dt.);
run;

Neither does:

data a;
/*    uncomment for input to work in the next data step*/
/*    length date1 $20; */
    date1 = '2015-10-26T23:59   ';
    date2 = '2015-10-26';
run;

data b;
    set a;
    not_related = trim(left(strip(date2)))||'T23:59';
    dt = input(date1, e8601dt.);
run;

The Input and informat are apparently checking on the length propery of the variable not the number of assigned characters.

[TD]

does not seem a consistent check: there's no problem to input a value from variable with length of $16 when the date2 is initialized first in the first data step.

[ballardw]

This is definitely odd:

data a;
    length date1 date2 $ 16 ; 
    date1 = '2015-10-26T23:59:00';
    date2 = '2015-10-26T23:59';
run;

data b;
    set a;
    dt1 = input(date1, e8601dt.);
    dt2 = input(date2, e8601dt.);
run;

It may be time to contact Tech support as this particular bit might be an actual bug.

But I would suggest if possible sticking to declaring any variable like this with a length at least 19 and not relying on implied length assignments. Explicit lengths are better programming practice.

[SuryaKiran · Accepted solution]

Its a known bug since SAS 9.2. What happening here is width of E8601DT must be 19 but the width
of value is only 16, so extra 3 characters are taken from the followed values.

Here in the example it is trying to take the values '201' as seconds i.e attempting to takes
values as ":ss". As 201 is not in the format as ":ss" it throwed an error as invalid argument.

This issue has been fixed in SAS 9.4

FYI: http://support.sas.com/kb/43/743.html