Re: [xml] ampersand conversion

[Igor Zlatkovic <igor zlatkovic com>]

On 17.05.2005 15:20, oliverst online de wrote:
> Hi,
>
> this code is giving me "error : unterminated entity reference" with
> libxml2-2.6.16
>
> xmlNodePtr node = xmlNewNode(NULL, BAD_CAST "test");
> if( node ) {
>         const char* str = "<>\"'&";
>         xmlNodeSetContent(node, BAD_CAST str);
> }
>
> I know, that some specific character have to be encoded/quoted, but all
> others are already being encoded internally and I would just have to
> replace the ampersand by myself. Is this a wanted behavior? Because it
> seems a bit inconsistent to me encoding all the others, but not the
> ampersand.

I haven't checked the code, so this is without warranty, but this is how 
I see it. xmlNodeSetContent creates a child node of type text and puts 
the supplied content in it. It does not parse the supplied content as 
XML and then inject the result into the tree. For the latter to work in 
any sane way, the supplied content would have to be at least a valid 
document fragment.

I cannot use a '>' or a '<' within a text node, so these are reasonably 
quoted. But I can very well use entity references or character 
references and expect them to be handled accordingly. Therefore, leaving 
ampersands alone when quoting things in a text node is only sane, 
because I would have no chance to insert an entity reference, or a 
character reference, in the newly created text child otherwise.

For consistency, if at all, libxml could require you to use a

  const char* str = "&lt;&gt;\"'&amp;";

instead of what you have used, I feel this way would be a lot more safe 
than automagically supplying an &amp; for every ampersand that triggers 
the unterminated entity reference error.

Ciao,
Igor