From: Alex Gutteridge <alexg@...>
Date: 2012-09-20T17:22:40+09:00
Subject: Re: FloatDomainError while rounding

On 20.09.2012 05:04, raj g. wrote:
> Alex Gutteridge wrote in post #1076502:
>> On 18.09.2012 16:51, Alex Gutteridge wrote:
>>>> Raj
>>>
>>> In that case @num must be Infinity (or -Infinity). Possibly @a or 
>>> @b
>>> are 0 and so the logarithms you are calling earlier generate
>>> -Infinity. Either way, as Jesus says the simplest thing is to just
>>> print out all the intermediate variables until you see what's 
>>> wrong.
>>
>> Apologies, I didn't read your second paragraph closely enough. If 
>> your
>> number is above the Float maximum limit it will also die with this 
>> error
>> (so not necessarily Infinity, but a large number):
>>
>>>> (Float::MAX_10_EXP).round
>> => 308
>>>> (1E308).round
>> =>
>>
> 
> 100000000000000001097906362944045541740492309677311846336810682903157585404911491537163328978494688899061249669721172515611590283743140088328307009198146046031271664502933027185697489699588559043338384466165001178426897626212945177628091195786707458122783970171784415105291802893207873272974885715430223118336
>>>> (1E309).round
>> (irb):23: warning: Float 1E309 out of range
>> FloatDomainError: Infinity
>>          from (irb):23:in `round'
>>          from (irb):23
>
> Thanks for the explanation. So How can we avoid this problem? Is 
> there
> any solution for this to avoid this FloatDoaminError?

Are you really sure that such a large number is the correct output from 
your algorithm?

If so, then AFAIK your only option is to use BigDecimal in your 
calculations rather than Float:

http://www.ruby-doc.org/stdlib-1.9.3/libdoc/bigdecimal/rdoc/BigDecimal.html

-- 
Alex Gutteridge