[
Lists Home |
Date Index |
Thread Index
]
At 03:03 PM 12/4/2002 -0500, Rich Salz wrote:
>>Well, there's a problem you will encounter earlier. Float and double are
>>restricted to 32 and 64 bits respectively:
>
>Only in their value space, not their lexical space.
Hmmmm....you're right, in the value space this number would be rounded, and
it is not invalid to have more digits after the decimal point than will
actually be used.
>My example -- with and without the type attribute -- is legit. And shows
>the problem.
Yes, you are right - using the *wrong* datatype can lead to loss of
precision. There are three simple solutions here. The first two require you
to be in control of the data:
1. Use the right datatype. It's xs:decimal, which gives you arbitrary
precision.
2. Use no datatype, and let the application figure out how to deal with the
data.
The third can be used even if someone supplied the wrong datatype:
3. Ignore the specified datatype in the software that processes the data
and access the lexical representation. For XQuery or XSLT, you would use
//pi-element/text().
Either 2 or 3 require a lot more work from the programmer. If you are in an
environment that can use arbitrary-precision math, it's easier to use
existing libraries and to have your parser recognize the data and deal with
it appropriately.
So I think this basically demonstrates that (a) if you have the wrong
datatype, and aren't paying attention, it can lull you into loss of
precision, and (b) if you have no datatype or the wrong datatype, you can
still get the information you want, but it may involve a lot more work. The
work required is the same in either case, since both involve operations
based on the lexical form.
Jonathan
|
