This was my original implementation but it was changed to this explicit API with the reasoning that doing this caching should be an explicit act of intent and not just a simple flag working globally.

Note, if more than just xml based metadata could be cached that would be awesome - but that might be better solved via something like AS7 annotation index mechanism.

2011/4/13 Steve Ebersole<steve(a)hibernate.org>: > On 04/12/2011 06:01 PM, Max Rydahl Andersen wrote: >> This was my original implementation but it was changed to this explicit API with the reasoning >> that doing this caching should be an explicit act of intent and not just a simple flag working globally. > > But isn't the user explicitly saying "cache them all" the same on this > point? To me it really comes down to whether there is a valid use case > for caching some, but not all. +1 for all or nothing. If it's possible / makes sense, and as I see the implementation is checking anyway for file timestamps, I don't see why any more fine grained control could be needed. Should it be on by default? (BTW looking at implementation it seems the IOStreams are never closed - I'm making a patch) >> Note, if more than just xml based metadata could be cached that would be awesome - but that might be >> better solved via something like AS7 annotation index mechanism. > > funny you should say that... ;) > > -- > Steve Ebersole<steve(a)hibernate.org> > http://hibernate.org > _______________________________________________ > hibernate-dev mailing list > hibernate-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hibernate-dev >

Do we want to make caching of files a flag for processing all of the xml files? Currently we have methods: addFile(String path) addFile(File file) addCacheableFile(String path) addCacheableFile(File file) Could we instead condense these and allow configuration of whether to try and cache xml files?

im confused - are we seriously looking at using jaxb for parsing the xml configurations even though they are doing what they can to remove any jaxb stuff from AS7 codebase because of the bad performance and maintenance overhead (especially when its part of the server) ? /max On Apr 13, 2011, at 11:15, Hardy Ferentschik wrote: > On Tue, 12 Apr 2011 21:26:01 +0200, Steve Ebersole<steve(a)hibernate.org> > wrote: > >> Do we want to make caching of files a flag for processing all of the xml >> files? >> >> Currently we have methods: >> addFile(String path) >> addFile(File file) >> addCacheableFile(String path) >> addCacheableFile(File file) >> >> Could we instead condense these and allow configuration of whether to >> try and cache xml files? > > Besides my other comments regarding low priority and xml maybe not being > the right > level of caching, I would also prefer a configuration approach. > > --Hardy > _______________________________________________ > hibernate-dev mailing list > hibernate-dev(a)lists.jboss.org > https://lists.jboss.org/mailman/listinfo/hibernate-dev /max http://about.me/maxandersen

On Apr 15, 2011, at 17:49, Steve Ebersole wrote: > What causes the bad performance? You do realize there are 3 aspects to performing jaxb right? You have the actual jaxp parsing, the validation and then finally the binding. > > Conversely, with DOM you have jaxp parsing, validation and DOM model binding. > > So what is the big difference? I'm not an expert on it but what I understood from Jason on why they (as I understood it) removed all JAXB parsing was that a) it was too slow b) not flexible c) the code got very messy to support multiple versions/variations (probably related to b) This was in comparison to the stax(?) parsers they moved to as far as I understood it. /max > > On 04/15/2011 10:45 AM, Max Rydahl Andersen wrote: >> im confused - are we seriously looking at using jaxb for parsing the xml configurations even though >> they are doing what they can to remove any jaxb stuff from AS7 codebase because of the bad performance >> and maintenance overhead (especially when its part of the server) ? >> >> /max >> >> On Apr 13, 2011, at 11:15, Hardy Ferentschik wrote: >> >>> On Tue, 12 Apr 2011 21:26:01 +0200, Steve Ebersole<steve(a)hibernate.org> >>> wrote: >>> >>>> Do we want to make caching of files a flag for processing all of the xml >>>> files? >>>> >>>> Currently we have methods: >>>> addFile(String path) >>>> addFile(File file) >>>> addCacheableFile(String path) >>>> addCacheableFile(File file) >>>> >>>> Could we instead condense these and allow configuration of whether to >>>> try and cache xml files? >>> >>> Besides my other comments regarding low priority and xml maybe not being >>> the right >>> level of caching, I would also prefer a configuration approach. >>> >>> --Hardy >>> _______________________________________________ >>> hibernate-dev mailing list >>> hibernate-dev(a)lists.jboss.org >>> https://lists.jboss.org/mailman/listinfo/hibernate-dev >> >> /max >> http://about.me/maxandersen >> >> >> > > -- > Steve Ebersole<steve(a)hibernate.org> > http://hibernate.org /max http://about.me/maxandersen

> we are using stax now under the covers as well. > > These are just different models of jaxp parsing: dom, sax, stax.

Right, and the benefit of JAXB is the typed API.