Re: [xsl] XPath: comparison of -0.0e0 and +0.0e0

["Imsieke, Gerrit, le-tex" <gerrit.imsieke@xxxxxxxxx>]

On 2011-06-19 23:08, Stefan Krause wrote:
> > [b&] Out of interest,
> > why do you need to distinguish them?
>
> Because atan2(y,x) treats them different, see [1]. I try to implement this behavior in XSLT 2.0.

I think its strange that the spec requires ambiguous results of atan2 
for zero-valued arguments:

The expression math:atan2(+0.0e0, -0.0e0) returns 0.0e0.
The expression math:atan2(+0.0e0, -0.0e0) returns math:pi().

The expression math:atan2(-0.0e0, -0.0e0) returns -0.0e0.
The expression math:atan2(-0.0e0, -0.0e0) returns -math:pi().

Then atan2 is no longer a function, mathematically.

If, in an attempt to disambiguate the function, the B1O results were not 
required, you could simply return
-x*y for x=B10, y=B10.

If the spec is flawed and should actually read:

The expression math:atan2(+0.0e0, +0.0e0) returns 0.0e0.
The expression math:atan2(+0.0e0, -0.0e0) returns -0.0e0.
The expression math:atan2(-0.0e0, +0.0e0) returns math:pi().
The expression math:atan2(-0.0e0, -0.0e0) returns -math:pi().

then I donbt see another way than reverting to some tricks such as cast 
each argument to xs:string and look up the corresponding result.

Gerrit