[Gcl-devel] Re: printing IEEE 754 special values (inf, nan)

[Raymond Toy]

Robert Dodier wrote:
> Hello Camm,
>
>
> > Greetings!  I could probably implement 'Inf' an 'NaN' as the print
> > forms of these float pretty trivially -- the question is whether the
> > reader must be modified to read them in, and what are the compliance
> > issues if any.
>
>
> Well, maybe some special symbols can be defined so
> that floating point inf, -inf, and nan are printed as
> those symbols.
>
> For comparison, CMUCL has #.EXT:DOUBLE-FLOAT-POSITIVE-INFINITY,
> #.EXT:DOUBLE-FLOAT-NEGATIVE-INFINITY, and #<DOUBLE-FLOAT Quiet NaN>.
> This last one is unreadable, right? I'm guessing that's
> because nan is not a single bit pattern but a range of bit patterns.
> However, I don't see why nan couldn't be read as a specific
> value; all nan's have the same properties.

I think that if you have readable NaN's, you should make it so that what 
you read is what was written.  It would certainly break people's 
expectation if all printed NaN's got read as the same NaN.

Also, I think implementations are free to encode whatever they want in 
the bits, so it might be important for the implementation.  But I might 
be wrong here.

Ray