The CALL COMPCOST routine sets the costs of operations for later use by the COMPGED function.

When the CALL COMPCOST routine uses the PUNCTUATION operation and a value of zero, the COMPGED function ignores case sensitivity and returns the wrong value for the generalized edit distance.

This was discovered in SAS 9.2, but additional testing found that it also existed back in SAS 9.1.3.

At this time there is no fix for the behavior.

Click the Full Code tab and the Results tab to see this behavior.

Operating System and Release Information

Data _null_ ;
   STRING1 = "baboon" ;
   STRING2 = "BABOON" ;

   /* 1st comparison => OK */
   GED1 = compged(STRING1,STRING2) ; /* take in care case */
   GED2 = compged(STRING1,STRING2,20,"i") ; /*  not take in care case */
   put "BEFORE COMPCOST " STRING1= STRING2= GED1= GED2= ; /* GED1 = 700 , GED2 = 0 */


   /* 2nd comparison => KO
   COMPCOST is used to ignore punctuation, but in fact it ignores case too.
   */
   call compcost("PUNCTUATION=",0) ;
   GED3 = compged(STRING1,STRING2) ; 
   GED4 = compged(STRING1,STRING2,"i") ;

   put "AFTER COMPCOST " STRING1= STRING2= GED3= GED4= ; /* GED3 = 0 whereas it should be 700 */
run ;

Log :
BEFORE COMPCOST STRING1=baboon STRING2=BABOON GED1=700 GED2=0
AFTER COMPCOST STRING1=baboon STRING2=BABOON GED3=0 GED4=0