Re: [xml] More floating point "issues"

[Daniel Veillard <veillard redhat com>]

On Tue, Apr 09, 2002 at 03:27:24PM +0100, Richard Jinks wrote:
> Hi
>
> Last problem with floating point numbers, I hope...
> (Well, these are the last of the regression tests I've been playing with,
> anyway!)
>
> According to the XPath spec, scientific notation is not allowed at all 
> (i.e. 1e-4).
> This applies both to converting numbers to strings, and strings to numbers.
> Libxml currently outputs "1e-6" instead of "0.000001", and will happily
> accept "1e-6" as a valid number instead of converting it to "NaN".

  Was beaten to death one year ago. Yes we know it's a violation
of the XPath 1.0 spec, it's part of XPath 2.0 requirement to support those,
and in this specific case I traded conformance for a more sensible behaviour.

> Also, when converting from numbers to strings ONLY, it is valid to have a
> number with no digits before or after the decimal point (but not both),
> i.e. ".4" and "4." are both valid, "." isn't.
> Of these cases, libxml only complains about numbers with no leading digits
> (".4") giving an "Invalid expression" error.

  Hum, okay,

> I've done a fix for all three cases, but I wanted to check with you about
> the problem with scientific numbers, esp since you've put specific cases 
> in the code to handle them. My guess is that libxml should probably never 
> output numbers in this format (i.e. should always output "0.000001"), 
> but do you want to stop this notation being used when inputting numbers?

  Basically patch gratefully accepted for the 2 latter cases but the first 
one would be a problem I'm afraid,

Daniel

-- 
Daniel Veillard      | Red Hat Network https://rhn.redhat.com/
veillard redhat com  | libxml GNOME XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/