Bigint Problem [FIXED IN v.1.3.10]

Bugs in ND1 (report'em, get'em fixed)

Bigint Problem [FIXED IN v.1.3.10]

Postby OrderEng » Fri Jun 17, 2011 12:30 am

If you multiple 123456791 by 123456791
you got the result 15241579244017680
OrderEng
 
Posts: 2
Joined: Fri Jun 17, 2011 12:18 am

Re: Bigint Problem

Postby oliver » Fri Jun 17, 2011 7:39 am

Hi OrderEng,

It's actually not a problem with using BigInts, but a problem of *not* using BigInts: The wrong result is a side-effect of using Reals instead of BigInts to carry out the calculation. The simple work-around is to use BigInts.

Do this by converting one or both of your numbers to BigInt, before multiplying. To convert a number on the stack to a BigInt, go to the Integer menu and tap ->big. A "big" badge will indicate the BigInt type.

What you're running into here with Reals, is the size limit of integers that can be accurately represented. In ND1, Reals can represent integers up to 2^53-1.
In practice, this means integers beyond 15 digits are not safe. (Some may be accurately represented, others not.)
If you type in a >15 digits integer on the edit line, it will be automatically made a BigInt (complete with badge), but there's no automatic promotion of arguments to BigInts for results which exceeds what a Real can represent. (Adding such a check would slow down operations on Reals tremendously.)

When multiplying these two numbers, most calculators will give you the result in exponential presentation, with *fewer* accurate digits--but at least the shown digits are correct. This is important, and I will look into doing this in ND1, too, while keeping all but the unsafe digits for a longer, better mantissa.

Thank you for reporting.

Oliver

P.S. Read this if you're an HP 49, 50g user:
ND1 breaks from type conventions in HP 49, 50g in two ways:
a) lists, and all kinds of vectors (symbolic, complex, real) are unified into one array type
b) numbers entered without a period are Reals, and not BigInts
(So, if you expected your numbers to be BigInts, know that they weren't. ND1 breaks with 49, 50g compatibility in this respect because this particular convention caused confusion and led to inefficient RPL code being written (when people forget to type the period).)
oliver
Site Admin
 
Posts: 433
Joined: Sat May 01, 2010 2:11 pm

Re: Bigint Problem

Postby oliver » Fri Jul 15, 2011 4:53 pm

This problem has been addressed in v1.3.10. The calculator now returns 1.52415792440177e+16.
As mentioned in the original reply, an Integer result is obtained if one or both of the inputs are BigInts.
oliver
Site Admin
 
Posts: 433
Joined: Sat May 01, 2010 2:11 pm


Return to Bugs

Who is online

Users browsing this forum: No registered users and 1 guest

cron