[xml] xpath number()/string() and locale

[Lars Oppermann <lars oppermann sun com>]

Hi all,

I think that the way the string() function in libxml2 xpath works is 
wrong according to the xpath specification. Please have a look at the 
following thoughts and let me know what you think so I can put together 
a patch.

Consider the following xml document:

<?xml version="1.0"?>
<root>1.25</root>

and the following xpath expression:

number(string(number(/root/text())))

Should LC_NUMERIC (or LC_ALL) be set to a language that uses a decimal 
seperator other then the '.' character the above expression will result 
in NaN being returned, since the string() function will use snprintf() 
in order to format the number to a string. The resulting string however 
cannot be used as a number again.

http://www.w3.org/TR/xpath states:

3.7
[30] Number ::= Digits ('.' Digits?)? | '.' Digits

4.2:
...otherwise, the number is represented in decimal form as a Number 
including a decimal point...
NOTE: The string function is not intended for converting numbers into 
strings for presentation to users. The format-number function and 
xsl:number element in [XSLT] provide this functionality.

With the string() function ultimately using snprintf() in order to 
format it's outoput it does not fullfill the requirement of representing 
a number in a language neutral form.
Also the scientific notation that might be generated from 
xmlXPathFormatNumber() is not allowed by the specification since it does 
not conform to 3.7[30]

I would rewrite the xmlXPathFormatNumber() in a way where it just dumps 
out the numbers according to xpath 3.7 and leave the rest to 
xslt:format-number. What do you think?

Cheers,
~Lars


--
Lars Oppermann <lars oppermann sun com>               Sun Microsystems
Software Engineer - StarOffice                           Sachsenfeld 4
Phone: +49 40 23646 959                                D-20097 Hamburg
Fax:   +49 40 23646 550                  http://www.sun.com/staroffice