Re^4: Repair malformed XML

Not quite. If the DTD tells you, for example, that element a may contain elements b, c, or d, and that b can contain e and f, then if it looks like element a contains one b, and two e's, you can be pretty sure that the b was close improperly (if at all), and the e's should be in b.

There are still many possibilities for confusion. But a heuristic that started with the DTD could do quite a good job. I'm not going to pretend it would be easy and/or fun ... but in theory the information may be there that could do a good job - and, if the DTD does not allow overlaps (such as a and b both allowing d's, so that the d can either be a child or a grandchild of a), you may even be able to do a perfect job.

Comment on Re^4: Repair malformed XML

Replies are listed 'Best First'.
Re^5: Repair malformed XML by Anonymous Monk on Feb 03, 2005 at 17:38 UTC
Ah, I see. You mean (X)HTML, and 'a' is `P`, 'b' is `SPAN`, 'c' is `CODE`, 'd' is `SMALL`, 'e' is `EM` and 'f' is `STRONG`. So, now you encounter: `<P>foo <SPAN> bar baz <EM> qux </EM> <EM> quux </EM> </P>` [download] Now, assuming tags aren't to be inserted inside words, I still can find five places to put in `</SPAN>`: before 'bar', between 'bar' and 'baz', after 'baz', between the two `EM` elements, and before the `</P>` tag. Now, if you have a DTD that says that the only possible content of a 'b' is exactly two 'e's, you know where the missing closing tag should have been. Note also that if you have a DTD where you can always unambigiously deduce where a missing closing tag should have gone, the closing tag is redundant - and if it were an SGML DTD instead of an XML DTD, the closing tag would have been optional. (And that would have solved the problem instantly - the document would be conforming).	[reply] [d/l]

Replies are listed 'Best First'.

Re^5: Repair malformed XML
by Anonymous Monk on Feb 03, 2005 at 17:38 UTC

P

SPAN

CODE

SMALL

EM

STRONG

  <P>foo <SPAN> bar  baz <EM> qux </EM> <EM> quux </EM> </P>
[download]

</SPAN>

EM

</P>

Now, if you have a DTD that says that the only possible content of a 'b' is exactly two 'e's, you know where the missing closing tag should have been.

Note also that if you have a DTD where you can always unambigiously deduce where a missing closing tag should have gone, the closing tag is redundant - and if it were an SGML DTD instead of an XML DTD, the closing tag would have been optional. (And that would have solved the problem instantly - the document would be conforming).

[reply]
[d/l]