This isn't a Mojo::JSON problem, nor a Math::BigInt or Math::BigFloat
problem. And it won't be a JavaScript problem either. It is a problem of
what happens when a floating point number is represented internally in
binary format, regardless of the language.
Any JSON library on any platform is going to see {"int":5e22} as a floating
point number (because 5e22 exceeds the size of any common platform's native
integers). Most platforms will treat that internally as a double. The
internal representation of 5e22 is inherently non-terminating. You and I
see 1/3rd, and intuitively know that it is represented in decimal format as
0.33333333333......... (a non-terminating sequence). We have no problem
making the mental leap that 33 cents is about 1/3rd of a dollar, even
though we know in the back of our minds from elementary school that it's
not exactly 1/3rd of 1.
Well, in the world of binary-representations of floating point numbers, any
number that cannot be represented by m/(2^m) forms a non-terminating binary
sequence, and as such, cannot be represented precisely with a finite number
of bits. This is the case in any language that stores floating point
numbers in an internal binary format.
So when you send {"int":5e22} to your JavaScript (or to Python, or even
back to Perl), the receiving end will translate that to an internal
representation of the floating point number 5e22. The internal binary
representation, being non-terminating, is imprecise, and would be converted
back to base-10 as 4.999e+22.
Presumably you've been using Math::BigInt and Math::BigFloat because
precision is of some importance in what you're working on. When you
serialize it to a plain-old-datatype, you are losing that precision, and
the receiving party has no choice but to accept that loss.
If you serialize to a string, the receiving party still has an opportunity
to preserve precision by converting the string to that language's concept
of a BigInt or BigFloat.
In the general case, stringification is the better solution, and it might
be so in your case as well; you will just have to deal with converting the
string losslessly at the receiving end, if that's important to you.
On Mon, Jun 30, 2014 at 3:05 PM, Tekki <[email protected]> wrote:
> Am Montag, 30. Juni 2014 22:48:07 UTC+2 schrieb David Oswald:
>
>>
>> I think what you were wanting is numification, so your TO_JSON methods
>> would call ->numify(). But then you will *not* be guaranteed a loss-less
>> round trip.
>>
>>
>>
> I will try it this way. I know that there is no guarantee, especially as
> JavaScript on the client side is involved. But like that I know where to
> search when new problems arise. Thanks for the help!
>
> --
> You received this message because you are subscribed to the Google Groups
> "Mojolicious" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To post to this group, send email to [email protected].
> Visit this group at http://groups.google.com/group/mojolicious.
> For more options, visit https://groups.google.com/d/optout.
>
--
David Oswald
[email protected]
--
You received this message because you are subscribed to the Google Groups
"Mojolicious" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/mojolicious.
For more options, visit https://groups.google.com/d/optout.
