RE: reporting bug displaying real numbers

[McKee, Paul]

If you look at machine representation of real (floating point) numbers, you
will see that they cannot in many cases be represented precisely. This
applies to any language and is a limitation of the way that we have to store
numbers. Hence, those of us who have been working with these things for a
while, never check for equality of a real but user greater or less than or
check within a range.

Hence I would not call this a bug but a DSP error.

Paul McKee
IS EMEA CPC DBA :
Merck Sharpe & Dohme Europe Inc. 
Tel : (+322)  776.62.61 - Fax : (+322) 776.63.04
Clos du lynx 5, 1200 Brussels


-----Original Message-----
From: Erick Alphonse [mailto:address@hidden
Sent: Thursday, 14 June, 2001 20:41
Cc: address@hidden
Subject: reporting bug displaying real numbers


Hello,

running gprolog-1.2.6 on an athlon-linux, I have the following outputs:

| ?- write(1.0).
1.0

yes
| ?- write(1.1).
1.1000000000000001

yes
| ?- write(1.2).
1.2

yes
| ?- write(1.3).
1.3

yes
| ?- write(1.4).
1.3999999999999999

and

| ?- 1.4 == 1.3999999999999999.

yes

and it will be the same with write/1, writeq/1, write_canonical/1,
display/1 and print/1.

That can be the source of big problems.
Erick.

_______________________________________________
Users-prolog mailing list
address@hidden
http://mail.gnu.org/mailman/listinfo/users-prolog