AW: AW: [Metafacture] Character classes in data source

[Christoph, Pascal]

Hi Christoph,

just opened #143[1], because I discovered some strange behaviour using simple
character classes in conjunction with the WildcardTrie "OR pattern".
(If #143 is a proper fix, we can modify WildcardTrie to support character
classes natively if you like).
In the long run I would love to see the possibility to use directly regex in
the "data source" field of metamorph. We then could maybe discard the
WildcardTrie as a whole - don't know about the sideffects of this, though.

cheers -o

[1]https://github.com/culturegraph/metafacture-core/pull/143

Am 28.10.2013 18:05 schrieb Böhme, Christoph :

> Hi Pascal,
> 
> As Markus merged the current implementation already, we can leave it for now. I think, for a more solid integration of the new matching patterns, the implementation of the WildcardTrie should be modified to support character classes natively. I cannot see a reason at the moment why this should not be part of the standard implementation. 
> 
> Alternatively, Metamorph could be extended support different types of matchers. The user could then use an attribute on the <metamorph> element to choose a matcher class that should be used in a certain script. I am not convinced though that this is a good idea whe looking at the future of Metamorph because the way a certain matcher works may affect other behaviours of the language.
> 
> Cheers,
> Christoph
> 
> 
>> -----Ursprüngliche Nachricht-----
>> Von: metafacture-bounces at lists.dnb.de [mailto:metafacture-
>> bounces at lists.dnb.de] Im Auftrag von Christoph, Pascal
>> Gesendet: Donnerstag, 17. Oktober 2013 11:39
>> An: Böhme, Christoph
>> Cc: metafacture at lists.dnb.de
>> Betreff: Re: AW: [Metafacture] Character classes in data source
>> 
>> Hi Christoph,
>> 
>> Am 16.10.2013 19:34 schrieb Böhme, Christoph :
>> 
>> > it is currently not possible to use character classes. You would have to extend the
>> WildcardTrie class [1] to support more elaborate matching patterns. I reckon it
>> should not be incredibly difficult to add support for character classes. We do not
>> have any plans for implementing this at the moment, though. Perhaps you could
>> give it a try yourself?
>> 
>> Yes, I have done this already, see the pull request[1].
>> I am not sure how to embed that new functionality, though. As you can see in
>> [1] I made a wrapper, for the WildardTrie is final and thus cannot be extended.
>> #135 provides a metamorph test, too. To make it simple for now, the new class
>> is instantiated at the WildcardDataRegistry , which is obviously not a good
>> place resp. needs renaming.
>> I would be glad if you have a more solid notion of integrating this new class.
>> Renaming is also fine with me.
>> 
>> -o
>> 
>> [1] https://github.com/culturegraph/metafacture-core/pull/135
>> 
>> > Best,
>> > Christoph
>> >
>> >
>> >
>> > [1] https://github.com/culturegraph/metafacture-
>> core/blob/master/src/main/java/org/culturegraph/mf/util/tries/WildcardTrie.java
>> >
>> >
>> >
>> >> -----Ursprüngliche Nachricht-----
>> >> Von: metafacture-bounces at lists.dnb.de [mailto:metafacture-
>> >> bounces at lists.dnb.de] Im Auftrag von Christoph, Pascal
>> >> Gesendet: Mittwoch, 16. Oktober 2013 10:42
>> >> An: metafacture at lists.dnb.de
>> >> Betreff: [Metafacture] Character classes in data source
>> >>
>> >> Hello *,
>> >>
>> >> in metamorph, is there a possibility to use more then wildcards ('*','?')
>> >> defining the data source? I would like to use simple character classes. For
>> >> instance, something like this:
>> >>
>> >> <rules>
>> >> 	<data source="902-1.9|907-1.9|.../>
>> >> </rules>
>> >>
>> >> should be expressable by this:
>> >>
>> >> <rules>
>> >> 	<data source="9[0123445678][27]-[-1].9/>
>> >> </rules>
>> >>
>> >> -o
>> >>
>> >> _______________________________________________
>> >> Metafacture mailing list
>> >> Metafacture at lists.dnb.de
>> >> http://lists.dnb.de/mailman/listinfo/metafacture
>> >
>> 
>> _______________________________________________
>> Metafacture mailing list
>> Metafacture at lists.dnb.de
>> http://lists.dnb.de/mailman/listinfo/metafacture
>