convert an xs:float value to a hex-string representation of its binary value

Question

using xsl 2.0, how would you convert an xs:float value to a hex-string representation of its binary value? i have no problem doing this for an integer (divide by 16 recursively and concatenate chars 0-9A-F), but float/double is stumping me.

<xsl:function name="my:float-to-hex" as="xs:string">
    <xsl:param name="in" as="xs:float"/>
    <xsl:sequence select="magic-here($in)"/>
</xsl:function>

a valid answer may be, "this is not possible," and that would be valuable to me if it were true. i'm open to other suggestions, but please don't depart from XSL. i am fully aware that there are more than one ways to skin this cat.

to clarify, the expected output would be the same as the output from this C code:

float f = 28.25f;
char *ptr = &f;
for(int i = 0; i < sizeof(float); i++) { printf("%02X", *(ptr + i)); }

Answer 1

I would go out on a limb and say that this is impossible in vanilla XSLT. With access to user-definable extension functions it is easy to offshore it to some language better suited, as you already indicated yourself by magic-here($in).

I don't think XSLT has a way to cast a number into a byte array. I can't point out a relevant section in the spec to prove it, though. Maybe someone else digs up the right paragraph.

XSLT is supposed to be deterministic and side-effects free, and as such it is highly unlikely that it provides a native way to access implementation specific details like the internal number representation.