> A variant of the float to string conversion that tries to preserve all the
> binary information would certainly be useful, but not as the default to_s
> method.   >
> irb> .1 - 3.0
> -0.8999999999999999
>
> would not be "surprising"! well, at least annoying)

At least users would be able to understand this:
>> 2.1 - 3.0
=> -0.9
>> (2.1-3.0) == -0.9
=> false

Since it would become
>> 2.1 - 3.0
=> -0.8999999999999
>> (2.1-3.0) == -0.9
=> false

I'd prefer the second--at least I'm reminded as to what's going on and
don't have to constantly pester ruby talk with questions like "is this
normal?" :)

However, this would seem surprising
>> 0.9
=>  0.90000000000000002

It seems slightly more annoying and verbose but more honest.
That being said, as someone above mentioned, the current float default
output is to "prettify" by stripping the ending few bits [I guess].
Perhaps we should keep this around as a new method name.  Like
to_human_string or something [?]

> One very elegant fix would be to invent a DecimalFloat type.  
> limited precision, as Floats do, but 0.1 would be represented precisely, as
> we finger counting humans expect it to be.  §  > in any other language.    
That would be way cool.  Something that elegantly "became" BigDecimal
whenever it hit possible loss of precision.  My guess is that this
would be almost immediately for almost everything, though :)

Ex:
>> 0.5.class
=> Float
>> 0.9.class
=> BigDecimal




>
> What most numerical methods folks do in practice, is to define equality of
> floats over an epsilon interval that is context dependent.    > insist on counting dollars with Floats, any values equal to within 1/2 a
> penny (or 0.005 dollars) might be considered "equal" for all practical
> purposes.       > Here's one way to avoid that:
>
> require 'delegate'
> class Money < DelegateClass(Float)
>   >
>   >   >     >    
> 
>  
>
>    >   
>  
> end

That would be an option.  By default use an epsilon of some quantity.
The only problem with this is the one you pointed out--...who can
determine a default epsilon?

I'm thinking that for now making the default string output be one s.t.
some_float == eval("#{some_float}")
would be good [and possibly retaining the 'human readable' through
some other name, not even to be used with inspect, just something not
existing].
Thoughts?
-=r