Re: Keeping numeric entities intact when parsing and serializing?

On Sat, 10 Oct 2009 01:22:30 +0200, Petr Pajas <[email protected]>  
wrote:

> 2009/10/10 Nicolas Mendoza <[email protected]>:
>> On Fri, 09 Oct 2009 23:58:53 +0200, Petr Pajas <[email protected]>
>> wrote:
>>
>>> 2009/10/9 Nicolas Mendoza <[email protected]>:
>>>>
>>>> On Fri, 09 Oct 2009 22:51:32 +0200, Aristotle Pagaltzis
>>>> <[email protected]>
>>>> wrote:
>>>>
>>>>> * Nicolas Mendoza <[email protected]> [2009-10-09 17:55]:
>>>>>>
>>>>>> Is there some way to keep the entities intact when parsing and
>>>>>> serlalizing numeric entities?
>>>>>
>>>>> Why would you want such a thing?
>>>>>
>>>>
>>>> Because I'm feeding it data and I want it to come out the same way?  
>>>> Just
>>>> like &amp; does. (I want to distinguish an incoming &#39; and "'". So  
>>>> I
>>>> don't want it to alter my valid XML data, basically.)
>>>>
>>>> Actually it's a bit surprising that libxml2 does that.
>>>
>>> Huh? There is absolutly nothing surprising about that, it is an XML
>>> parser? Programming XML would be hell if XML parsers didn't do this.
>>>>
>>>> From the XML point of view, &#39; and ' are the same thing! Read the
>>>
>>> XML spec.
>>>
>>> To be fair, there are few cases when particular formatting matters,
>>> but those are (supposed to be) dealt with by XML C14N (and possibly
>>> XML Encryption and XML Signature).
>>>
>>
>> I think I can sympathize with your sentiments, but I'm not sure I can  
>> agree
>> 100% that altering in-data is the optimal way of functioning.
>
> The data is the charcters, not the way they are serialized in XML.
> What you want is of the same nature as asking the parser to remember
> the original whitespace within XML tags, e.g. to distinguish
>
> <foo bar="baz"/>
>
> from
>
> <foo
>        bar="baz"
>     />
>
> No widely-adopted XML API can preserve this distinction, nor it can
> preserve the distinction between a character and the corresponding
> numerical entity. XML APIs typically exchange content, not
> representation.
>
>> No matter if they are the same in theory.
>
> What theory that would be? I'm not talking about theories, I'm talking
> about the XML 1.0 spec.
>
>> Why would it be hell if XML parsers didn't convert numeric entities to
>> ASCII, UTF-8 (or whatever charset is possible/available at the time) on
>> serialization?
>
> It would be hell because it would be like having no parser at all.
>
> Also note that parsers don't "convert" anything on serialization, but
> already during parse. In fact, parsers don't serialize (serializers
> do); they parse, i.e. read the input and decode the content into some
> structured form, passing it via some API to a handler (possibly a
> serializer or an application). Content, not the way it is encoded
> (alghough, to be fair, some do parsers send the original as well, e.g.
> via offsets to the source stream; from the implementation point of
> view this may be costly, from the practical point of view it is seldom
> useful).
>
> So to sum up: if you don't want a parser to parse your input, then
> don't use one! Just process the XML as text (e.g. using regexps,
> lexer, tokenizer, or a parser that gives you real low-level access),
> since apparently you are not interested in the content of the XML
> document but one particular textual representation of the content in
> XML.
>
> -- Petr

So, given that we accept this, why is &amp; not converted to "'"?

-- 
Nicolas Mendoza
http://my.opera.com/nicomen
_______________________________________________
Perl-XML mailing list
[email protected]
To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs