Re: [xml-dev] XML parser model, when are &foo; inserted?

[David Carlisle <davidc@nag.co.uk>]

On 03/12/2012 23:19, ольга крыжановская wrote:
> Michael, how does nesting work? Can I just take the text of &foo;,
> expand that until I run out of &...; parts and insert the result in
> the text stream?
>
> Olga



No because if you have

<!ENTITY foo "<x>aaa"  >

with a replacement text that has a start tag but no close, you can not 
flag an error at that point as the document is still well formed if the 
entity is not referenced.

<!DOCTYPE x [
<!ENTITY foo "<x>aaa">
]>
<x>
</x>


is OK, but

<!DOCTYPE x [
<!ENTITY foo "<x>aaa">
]>
<x>
&foo;</x>
</x>



is not well formed:

$ rxp -sx bad2.xml
Error: Element ends in different entity from that in which it starts
  in unnamed entity at line 5 char 9 of file:///c:/tmp/bad2.xml


However if you just expanded the foo without doing the check, and got


<!DOCTYPE x [
]>
<x>
<x>aaa</x>
</x>

It would, again, be well formed.

So you need to keep track of entity boundaries you can not simply expand 
all the entities as macros and then parse the result.

David