[XSL-LIST Mailing List Archive Home] [By Thread] [By Date]

Re: [xsl] Implicit casting from double to float

---

---

Thanks, David for the remarks.

I think, this is how Saxon implements xs:double to xs:float casts (the
implicit casts). I did some tests today about this ...

If the double value can be accommodated in the float range, a float
variable can hold the double value (as a float).

But if the double value is larger than the maximum float value, then
the result of the cast is an infinite value (INF or -INF in XPath 2.0
terms).

I can see, that specs do not disallow this. So this should be conformant I feel.

On Wed, Jan 7, 2009 at 6:19 PM, David Carlisle <davidc@xxxxxxxxx> wrote:
>
> PS
>
>> However I can't actually verify that this should work in XSLT.
>
> I think it's clear that the intent is that it should work, it would be
> bizarre to allow automatic promotion from decimal to float (losing
> arbitrary amounts of precision) but not to allow promotion from double
> to float (losing a bounded amount of precision). Basically XPath works
> as if there is a numeric type from which decimal/double/float are all
> derived, and which if it existed would make many of these conversions
> much simpler to specify. However it wouldn't be the first time that
> there is a difference between what was intended to be specified and what
> the specification actually says...
>
> David



-- 
Regards,
Mukul Gandhi

---