Re: Validating with XML::LibXML::RelaxNG?

Fixed version now in SVN ( svn co svn://axkit.org/XML-LibXML/trunk )

-- Petr


2009/7/16 Petr Pajas <[email protected]>:
> 2009/7/7 Vaclav Barta <[email protected]>:
>> On Tuesday 07 July 2009 11:56:53 Petr Pajas wrote:
>>> I just looked into the libxml2 source code to find out that it stores
>>> some per-node psvi information based on previous successful
>>> validation. If the node->psvi pointer equals the pointer to the
>>> definition it is currently being tested against, the validation of
>>> that node is skipped. This means that nodes that were once valid
>> ...
>>> Not sure how to deal with this in XML::LibXML. Options include:
>> If you're concerned about efficiency, I suppose you can maintain a 3-state
>> flag per document:
>>  - either it hasn't been validated before (and therefore has no validation
>> info)
>>  - or it has been validated but hasn't changed since then (so the validation
>> info is correct)
>>  - or has changed after validation - this is the only case when the next
>> validation must clear the cached info before it starts
>
> In fact I can never be completely sure about the second case, since
> the validator simply compares node->psvi's as pointers. It you free
> one schema and load another one, there is a nonzero (though little)
> chance that some node->psvi pointers will again point to a grammar
> rule from the new schema it is supposed to match and the validator
> will skip the actual check. Esp. if the first and second schema are
> very similar.
>
> This is in fact nice, since it reduces the situation to just two
> cases: validated before / never validated. When a document is
> validated, I just have to taint the document node and keep it tainted
> forever. This means I won't have to worry about linking nodes with the
> same owner document to the document tree (i.e. the remove nodes,
> validate, append nodes, re-validate scenario does not influence the
> status). However, when a node from a tainted document is imported to
> the current document tree, the current document must also get tainted
> (or, the psvi information from the imported subtree must be cleared).
>
> -- Petr
>
>
> -- Petr
>
_______________________________________________
Perl-XML mailing list
[email protected]
To unsubscribe: http://listserv.ActiveState.com/mailman/mysubs