[Info-vax] Is this a Bug ?

[David Froble]

StGallen wrote:
> Dear All
> 
> VMS 7.3-1
> DEC Basic 1.4
> 
> I have a situation where I need to check if the integer part of a decimal number is > 0. The number could be upto 999999999999.999999. Thus, I need to check if the integer part 999999999999 is greater than 0.
> 
> INTEGER(sNum,LONG) works but throws 52 - Illegal Number since the LONG cannot handle a number this big. My alternative was to use QUAD. 
> This gives me BAS-F-MEMMANVIO, Memory management violation
> 
> Refering to the BASIC manual, I have added: option handle = severe
> Also I have compiled with the /CHECK qualifier.
> 
> Any ideas are most gratefully welcome.
> 
> 
> 
> 10
>         option handle = severe
> 
> %include "STR$ROUTINES" %from %library "sys$library:basic$starlet"
> %include "$lnmdef"      %from %library "sys$library:basic$starlet"
> %include "$libclidef"   %from %library "sys$library:basic$starlet"
> 
> 
> DECLARE STRING sNUM
> DECLARE STRING sPI
> DECLARE STRING sPF
> DECLARE LONG lSTAT
> sPI = ""
> sPF = ""
> 
> LINPUT "Enter a number  ";sNUM
> lStatus =  STR$ELEMENT(sPI,0,".",sNUM) ! Integer part
> lStatus =  STR$ELEMENT(sPF,1,".",sNUM) !fractional part
> 
>  IF INTEGER(sPI,QUAD) = 0 !<---- program aborts here
>  THEN
>     PRINT "INT part is 0"
>  ELSE
>         PRINT "INT part is ";INTEGER(sPI,QUAD)
>  END IF
> 32767 END

10      Declare Double Z1, Z2

         Z1 = 999999999999.999999
         Z2 = Fix( Z1 )
         Print Format$( Z1 , "#,###,###,###,###.######" )
         Print Format$( Z2 , "#,###,###,###,###.######" )

         If      Z2 > 0.
         Then    Print "True"
         Else    Print "False"
         End If

         End

Ready

run
NONAME    17-APR-2012 11:24

1,000,000,000,000.000000
1,000,000,000,000.000000
True
Ready

Ok, what's wrong with the above?  Double FP has if I remember correctly 15 digits of 
precision, but can handle larger numbers.  Since FP isn't precise (past the first 15 
digits) Format$() is rounding the output.

If all you want to do is get the "whole" part of a number, "Fix()" is what you want.

Warning, if you use the above code to test, and put more than 6 digits to the right of the 
DP, Format$() will round.  But, you stated your needs for the fractional part was only 6 
digits, so I coded the Format$() mask accordingly.

Warning #2, FP is NOT absolutely precise.  I remember some years ago Randy Parks ranting 
about some FP values that would give some rather interesting results.  It's just the way 
the FP works.

Hmmm......

Re-reading your code, you're input is a string, not a DECIMAL data type, (dumb Dave, you 
assumed again), you can use Val() to convert it to a floating point number.  As long as 
you have less than 14-15 digits to the left of the DP, then Fix() will give you a valid 
whole part of the number.