Well, I haven't yet seen anything that tells me that it would be correct for a mapping found in WEB-INF/validation.xml to be resolved from the ServletContext as opposed to the classpath, but since in an EE server the BV impl (here BVal) would live "above" the application code there's a problem regardless in having BVal load the mapping resources, I think, because it won't have awareness of a given webapp's classloader.
However, using Romain's approach of having the actual parsed JAXB ValidationConfigType object be passed to BVal would seem to take care of your issue: the EE server could use JAXB to produce this from WEB-INF/validation.xml, then pick off the mapping elements, provide the modified ValidationConfigType object to the BV bootstrapping, and call #addMapping() for the app-specific resource streams. How does that sound? Matt On Wed, Mar 19, 2014 at 5:25 PM, Michael Blyakher <michael.blyak...@gmail.com> wrote: > From an application perspective I understand that regardless how the > ValidatorFactory is built there would never be a desire to ignore mappings > files specified in validation.xml. The application already knows what it > wants and therefor anything specified should be used from both ways to > specify mappings. > > In an EE app server environment, the server needs to make the > Validator/ValidatorFactory for each module available through injection or > lookup. This means the app server is bootstrapping the ValidatorFactory > itself, using the module deployment descriptors (validation.xml) to create > it before passing it back to the application. With this in mind, the app > server needs to be able to direct bval to specify that the location of > validation.xml will be under WEB-INF for a web module (if it was included > by the app developer). As we discussed earlier, bval doesn't handle this. > > Taking a step back to 1.0 this wasn't an issue, because as long as the EE > app server could handle parsing validation.xml since it knows where/how to > find it and programatically bootstrap the Configuration, it could then call > ignoreXMLConfiguration and nothing would be lost. Now with 1.1, all CDI > integration bval does is lost if the EE app server follows this pattern. > Thus, to utilize the CDI integration piece, bval needs to create all of the > configuration components, but that also means that it needs to parse > validation.xml (or have it be provided to it). > > Now, if something (method TBD) was done to find WEB-INF/validation.xml by > bval, how then would it go about trying to find the mapping files? This is > done the same way that validation.xml was looked for originally before this > workaround/solution, which gets us into the same situation where we > couldn't find WEB-INF/validation.xml if the mapping file is > WEB-INF/my-mapping.xml (I'm curious where the spec indicates that this > location isn't compliant). > > So in short, it's not that I want to be able to ignore mappings altogether. > I was just thinking that if WEB-INF is a valid location for the mapping > file to live, bval won't be able to find it either, so even if a workaround > is provided for finding validation.xml, any mappings specified in xml will > not be found either. The idea of being able to programatically specify that > xml mappings should be ignored is so that the EE app server could convert > them into InputStream's and then somehow indicate to bval that it doesn't > need to do anything with the xml anymore. > > Hopefully all of that rambling makes sense and clarifies the problem I'm > butting into :) > > > > > On Wed, Mar 19, 2014 at 4:28 PM, Romain Manni-Bucau > <rmannibu...@gmail.com>wrote: > >> I think mapping in web-inf is not spec compliant >> >> That said calling ignoreXmlConfig you can already do what you want >> >> Finally i think the spi or assimilated is useless and using api + maybe >> few custom properties should be enough so i wouldnt add it before it sould >> be mandatory. It generally breaks the framework which is not enough tested >> then. >> Le 19 mars 2014 22:04, "Michael Blyakher" <michael.blyak...@gmail.com> a >> écrit : >> >> > I'm prototyping the development efforts for pulling in the 1.1 >> > implementation into an EE app server, so I need be able to press the >> right >> > buttons on bval so that it is able to handle both the mappings files and >> > validation.xml. (I won't be able to control how an application specifies >> > it's mappings, but I need to ensure that specifying them in xml under >> > WEB-INF works) >> > >> > My concern was that I was going to run into the same issues loading the >> > mappings files as with validation.xml from WEB-INF unless the proposed >> > change somehow provided a way to tell bval to skip using the mappings >> found >> > in the provided parsed validation.xml and only use those provided by >> > calling Configuration#addMapping(). Otherwise I would call >> > Configuration#addMapping(), but bval would still try to find the mappings >> > resources itself and fail to do so. Does that make sense? >> > >> > >> > On Wed, Mar 19, 2014 at 12:32 PM, Matt Benson <gudnabr...@gmail.com> >> > wrote: >> > >> > > XML constraint mapping files are separate from xml validation config. >> > > So you either provide them via Configuration#addMapping() or in your >> > > validation.xml (or whatever you override with). >> > > >> > > Matt >> > > >> > > On Wed, Mar 19, 2014 at 11:39 AM, Michael Blyakher >> > > <michael.blyak...@gmail.com> wrote: >> > > > So if I understand this latest proposal correctly, any bootstrapper >> (EE >> > > > servers specifically) will be able to provide the parsed >> validation.xml >> > > > configuration to the ApacheValidatorConfiguration? >> > > > >> > > > If so, how will this work with the mappings config files? If for >> > example >> > > I >> > > > have my constraints defined in WEB-INF/my-mappings.xml, while >> > > bootstrapping >> > > > will I still be able to set the InputStream for that file without >> bval >> > > > trying to do it as well (and not finding this resource at this >> > location)? >> > > > Previously this could be accomplished by specifying >> > > > Configuration.ignoreXMLConfiguration, but I don't quite see how that >> > > would >> > > > work in this case. >> > > > >> > > > Thanks, >> > > > Mike >> > > > >> > > > >> > > > On Wed, Mar 19, 2014 at 11:11 AM, Romain Manni-Bucau < >> > > rmannibu...@gmail.com> >> > > > wrote: >> > > >> >> > > >> Well if we can avoid to fork/branch tomee before next release would >> be >> > > >> awesome but yes it sonds reasonable and avoiding jvm SPI is awesome >> > > >> Romain Manni-Bucau >> > > >> Twitter: @rmannibucau >> > > >> Blog: http://rmannibucau.wordpress.com/ >> > > >> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> Github: https://github.com/rmannibucau >> > > >> >> > > >> >> > > >> >> > > >> 2014-03-19 17:10 GMT+01:00 Matt Benson <gudnabr...@gmail.com>: >> > > >> > Actually, come to think of it, we don't have to do it as a >> > "services" >> > > >> > SPI at all; we can just define the interface and have it be a >> custom >> > > >> > config item for ApacheValidatorConfiguration. This makes it more >> > > >> > explicit and TomEE can just specify when bootstrapping--hopefully, >> > > >> > anyway. We'll see if there are any gotchas and hopefully we can >> get >> > it >> > > >> > working in a TomEE branch or fork before we set it in stone. Okay? >> > > >> > >> > > >> > Matt >> > > >> > >> > > >> > On Wed, Mar 19, 2014 at 11:06 AM, Matt Benson < >> gudnabr...@gmail.com >> > > >> > > >> > wrote: >> > > >> >> Well, in that case I don't see how we can really go wrong there. >> > I'll >> > > >> >> try to remember to do this as I'm hacking BVal in the coming >> weeks >> > > and >> > > >> >> maybe we can then see how it looks in TomEE. >> > > >> >> >> > > >> >> Matt >> > > >> >> >> > > >> >> On Wed, Mar 19, 2014 at 11:00 AM, Romain Manni-Bucau >> > > >> >> <rmannibu...@gmail.com> wrote: >> > > >> >>> that's what I was thinking about but when I hacked 1.1 branch I >> > was >> > > >> >>> really thinking adding it when integrating tomee to avoid a >> > useless >> > > or >> > > >> >>> wrong SPI. >> > > >> >>> Romain Manni-Bucau >> > > >> >>> Twitter: @rmannibucau >> > > >> >>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>> Github: https://github.com/rmannibucau >> > > >> >>> >> > > >> >>> >> > > >> >>> >> > > >> >>> 2014-03-19 16:59 GMT+01:00 Matt Benson <gudnabr...@gmail.com>: >> > > >> >>>> So are you proposing the SPI look more like: >> > > >> >>>> >> > > >> >>>> public interface DefaultValidationConfigProvider { >> > > >> >>>> org.apache.bval.jsr.xml.ValidationConfigType >> > > >> >>>> getDefaultValidationConfig(); >> > > >> >>>> } >> > > >> >>>> >> > > >> >>>> ? >> > > >> >>>> >> > > >> >>>> Matt >> > > >> >>>> >> > > >> >>>> On Wed, Mar 19, 2014 at 10:57 AM, Romain Manni-Bucau >> > > >> >>>> <rmannibu...@gmail.com> wrote: >> > > >> >>>>> Cause: >> > > >> >>>>> 1) TomEE added some features relying on internal config >> > > >> >>>>> (placeholders etc) >> > > >> >>>>> 2) TomEE uses its own model for all EE descriptors whatever >> the >> > > spec >> > > >> >>>>> is >> > > >> >>>>> >> > > >> >>>>> That's not an issue on BVal side but it will need to be >> > integrated >> > > >> >>>>> without forking as much as possible >> > > >> >>>>> >> > > >> >>>>> Romain Manni-Bucau >> > > >> >>>>> Twitter: @rmannibucau >> > > >> >>>>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>>>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>>>> Github: https://github.com/rmannibucau >> > > >> >>>>> >> > > >> >>>>> >> > > >> >>>>> >> > > >> >>>>> 2014-03-19 16:52 GMT+01:00 Matt Benson <gudnabr...@gmail.com >> >: >> > > >> >>>>>> Why can't TomEE rely on BVal for parsing? We should devise >> > > >> >>>>>> something >> > > >> >>>>>> as simple as possible, whatever the case. >> > > >> >>>>>> >> > > >> >>>>>> Matt >> > > >> >>>>>> >> > > >> >>>>>> On Wed, Mar 19, 2014 at 10:45 AM, Romain Manni-Bucau >> > > >> >>>>>> <rmannibu...@gmail.com> wrote: >> > > >> >>>>>>> well this way we'll need another spi for TomEE which can't >> > rely >> > > on >> > > >> >>>>>>> BVal for parsing. That's why I thought sending the parsing >> > > result >> > > >> >>>>>>> >> > > >> >>>>>>> >> > > >> >>>>>>> BTW any urgence on it? >> > > >> >>>>>>> Romain Manni-Bucau >> > > >> >>>>>>> Twitter: @rmannibucau >> > > >> >>>>>>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>>>>>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>>>>>> Github: https://github.com/rmannibucau >> > > >> >>>>>>> >> > > >> >>>>>>> >> > > >> >>>>>>> >> > > >> >>>>>>> 2014-03-19 16:43 GMT+01:00 Matt Benson <mben...@apache.org >> >: >> > > >> >>>>>>>> I was thinking along the lines Michael says. e.g.: >> > > >> >>>>>>>> >> > > >> >>>>>>>> public interface DefaultValidationConfigurationProvider { >> > > >> >>>>>>>> InputStream getDefaultValidationConfiguration(); >> > > >> >>>>>>>> } >> > > >> >>>>>>>> >> > > >> >>>>>>>> Then we use ServiceLoader (functional equivalent for BVal >> > 1.0, >> > > >> >>>>>>>> Java 5) >> > > >> >>>>>>>> to find any available implementations. If none found, we >> fall >> > > >> >>>>>>>> back to: >> > > >> >>>>>>>> >> > > >> >>>>>>>> class StandardDefaultValidationConfigurationProvider >> > implements >> > > >> >>>>>>>> DefaultValidationConfigurationProvider { >> > > >> >>>>>>>> final Properties properties; >> > > >> >>>>>>>> StandardDefaultValidationConfigurationProvider(Properties >> > > >> >>>>>>>> properties) { >> > > >> >>>>>>>> this.properties = properties; >> > > >> >>>>>>>> } >> > > >> >>>>>>>> >> > > >> >>>>>>>> public InputStream getDefaultValidationConfiguration() { >> > > >> >>>>>>>> // look for property pointing to custom resource, else >> > > >> >>>>>>>> META-INF/validation.xml >> > > >> >>>>>>>> // ensure only one such resource >> > > >> >>>>>>>> // return getResourceAsStream(resourceName) >> > > >> >>>>>>>> } >> > > >> >>>>>>>> } >> > > >> >>>>>>>> >> > > >> >>>>>>>> This way TomEE would simply have to provide: >> > > >> >>>>>>>> >> > > >> >>>>>>>> WebApplicationDefaultValidationConfigurationProvider >> > implements >> > > >> >>>>>>>> DefaultValidationConfigurationProvider { >> > > >> >>>>>>>> public InputStream getDefaultValidationConfiguration() { >> > > >> >>>>>>>> return >> > > >> >>>>>>>> >> > > getServletContext().getResourceAsStream("WEB-INF/validation.xml"); >> > > >> >>>>>>>> } >> > > >> >>>>>>>> >> > > >> >>>>>>>> private static ServletContext getServletContext() { >> > > >> >>>>>>>> // TBD >> > > >> >>>>>>>> } >> > > >> >>>>>>>> } >> > > >> >>>>>>>> >> > > >> >>>>>>>> Matt >> > > >> >>>>>>>> >> > > >> >>>>>>>> On Wed, Mar 19, 2014 at 10:28 AM, Romain Manni-Bucau >> > > >> >>>>>>>> <rmannibu...@gmail.com> wrote: >> > > >> >>>>>>>>> Actually I'd expect the SPI to give the processed instance >> > and >> > > >> >>>>>>>>> not the >> > > >> >>>>>>>>> location. That's why i sugegsted to wait a bit for it to >> see >> > > the >> > > >> >>>>>>>>> real >> > > >> >>>>>>>>> need. >> > > >> >>>>>>>>> >> > > >> >>>>>>>>> >> > > >> >>>>>>>>> Romain Manni-Bucau >> > > >> >>>>>>>>> Twitter: @rmannibucau >> > > >> >>>>>>>>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>>>>>>>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>>>>>>>> Github: https://github.com/rmannibucau >> > > >> >>>>>>>>> >> > > >> >>>>>>>>> >> > > >> >>>>>>>>> >> > > >> >>>>>>>>> 2014-03-19 16:10 GMT+01:00 Michael Blyakher >> > > >> >>>>>>>>> <michael.blyak...@gmail.com>: >> > > >> >>>>>>>>>> How would an SPI like this work? Would it allow the EE >> > server >> > > >> >>>>>>>>>> to specify >> > > >> >>>>>>>>>> the location of the validation.xml (maybe in the form of >> an >> > > >> >>>>>>>>>> InputStream)? >> > > >> >>>>>>>>>> >> > > >> >>>>>>>>>> >> > > >> >>>>>>>>>> On Tue, Mar 18, 2014 at 1:59 PM, Romain Manni-Bucau >> > > >> >>>>>>>>>> <rmannibu...@gmail.com>wrote: >> > > >> >>>>>>>>>> >> > > >> >>>>>>>>>>> tomee parses it itself and then create the configuration >> > > >> >>>>>>>>>>> itself. I >> > > >> >>>>>>>>>>> think we can wait tomee starts javaee7 to write it since >> > it >> > > >> >>>>>>>>>>> should be >> > > >> >>>>>>>>>>> very soon (when next release is done) and it would be >> the >> > > main >> > > >> >>>>>>>>>>> and >> > > >> >>>>>>>>>>> more demanding user. >> > > >> >>>>>>>>>>> Romain Manni-Bucau >> > > >> >>>>>>>>>>> Twitter: @rmannibucau >> > > >> >>>>>>>>>>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>>>>>>>>>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>>>>>>>>>> Github: https://github.com/rmannibucau >> > > >> >>>>>>>>>>> >> > > >> >>>>>>>>>>> >> > > >> >>>>>>>>>>> >> > > >> >>>>>>>>>>> 2014-03-18 19:42 GMT+01:00 Matt Benson < >> > mben...@apache.org >> > > >: >> > > >> >>>>>>>>>>> > On Tue, Mar 18, 2014 at 1:01 PM, Michael Blyakher >> > > >> >>>>>>>>>>> > <michael.blyak...@gmail.com> wrote: >> > > >> >>>>>>>>>>> >> Hi All, >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> >> Thanks for the quick replies, and apologies for not >> > being >> > > >> >>>>>>>>>>> >> more specific >> > > >> >>>>>>>>>>> - I >> > > >> >>>>>>>>>>> >> was quoting the EE 7 Platform spec as I am >> particularly >> > > >> >>>>>>>>>>> >> interested in >> > > >> >>>>>>>>>>> using >> > > >> >>>>>>>>>>> >> the bval 1.1 implementation that hasn't been >> officially >> > > >> >>>>>>>>>>> >> released yet. >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> >> But from what I am hearing, it is the responsibility >> of >> > > an >> > > >> >>>>>>>>>>> >> EE server to >> > > >> >>>>>>>>>>> >> handle the WEB-INF case. I can see how this is >> possible >> > > for >> > > >> >>>>>>>>>>> >> the 1.0 >> > > >> >>>>>>>>>>> >> implementation, as the server can parse the >> > > validation.xml >> > > >> >>>>>>>>>>> >> itself and >> > > >> >>>>>>>>>>> >> bootstrap the configuration through the validation >> spec >> > > >> >>>>>>>>>>> >> API's. How would >> > > >> >>>>>>>>>>> >> this be done for the current 1.1 implementation in >> the >> > > >> >>>>>>>>>>> >> bval-1.1 branch >> > > >> >>>>>>>>>>> in >> > > >> >>>>>>>>>>> >> the repository? I don't see how the values for the >> > > >> >>>>>>>>>>> "executable-validation" >> > > >> >>>>>>>>>>> >> element could be provided to the impl through the >> > > >> >>>>>>>>>>> >> validation spec API's. >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> > >> > > >> >>>>>>>>>>> > Well, the >> > > >> >>>>>>>>>>> >> > > >> >>>>>>>>>>> >> > > >> > >> http://bval.apache.org/mvnsite/bval-jsr303/apidocs/org/apache/bval/jsr303/ApacheValidatorConfiguration.Properties.html#VALIDATION_XML_PATH >> > > >> >>>>>>>>>>> > property can be used to point to a different resource >> on >> > > the >> > > >> >>>>>>>>>>> > classpath, but I can't find any mechanism that could >> be >> > > used >> > > >> >>>>>>>>>>> > to hook >> > > >> >>>>>>>>>>> > up WEB-INF/validation.xml, and I can't find how TomEE >> > does >> > > >> >>>>>>>>>>> > it, so >> > > >> >>>>>>>>>>> > AFAICT you have indeed found what I consider a >> problem. >> > > Off >> > > >> >>>>>>>>>>> > the top of >> > > >> >>>>>>>>>>> > my head I think we could solve it by adding a simple >> SPI >> > > to >> > > >> >>>>>>>>>>> > discover >> > > >> >>>>>>>>>>> > the default validation configuration resource. >> Thoughts? >> > > >> >>>>>>>>>>> > >> > > >> >>>>>>>>>>> > Matt >> > > >> >>>>>>>>>>> > >> > > >> >>>>>>>>>>> >> Thanks, >> > > >> >>>>>>>>>>> >> Michael >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> >> On Tue, Mar 18, 2014 at 12:13 PM, Romain Manni-Bucau >> > > >> >>>>>>>>>>> >> <rmannibu...@gmail.com>wrote: >> > > >> >>>>>>>>>>> >> >> > > >> >>>>>>>>>>> >>> Hi >> > > >> >>>>>>>>>>> >>> >> > > >> >>>>>>>>>>> >>> Bval only looks in META-INF but TomEE for instance >> > (more >> > > >> >>>>>>>>>>> >>> generally EE >> > > >> >>>>>>>>>>> >>> servers) handles WEB-INF case. >> > > >> >>>>>>>>>>> >>> Romain Manni-Bucau >> > > >> >>>>>>>>>>> >>> Twitter: @rmannibucau >> > > >> >>>>>>>>>>> >>> Blog: http://rmannibucau.wordpress.com/ >> > > >> >>>>>>>>>>> >>> LinkedIn: http://fr.linkedin.com/in/rmannibucau >> > > >> >>>>>>>>>>> >>> Github: https://github.com/rmannibucau >> > > >> >>>>>>>>>>> >>> >> > > >> >>>>>>>>>>> >>> >> > > >> >>>>>>>>>>> >>> >> > > >> >>>>>>>>>>> >>> 2014-03-18 17:50 GMT+01:00 Michael Blyakher < >> > > >> >>>>>>>>>>> michael.blyak...@gmail.com>: >> > > >> >>>>>>>>>>> >>> > Hi, >> > > >> >>>>>>>>>>> >>> > >> > > >> >>>>>>>>>>> >>> > Where is the validation.xml supposed to be for a >> web >> > > >> >>>>>>>>>>> >>> > archive? The >> > > >> >>>>>>>>>>> bval >> > > >> >>>>>>>>>>> >>> > spec's only indicate the "META-INF/validation.xml" >> > > >> >>>>>>>>>>> >>> > location, but the >> > > >> >>>>>>>>>>> EE >> > > >> >>>>>>>>>>> >>> > platform spec indicates that for a web archive >> this >> > > >> >>>>>>>>>>> >>> > location must be >> > > >> >>>>>>>>>>> >>> > "WEB-INF/validation.xml". >> > > >> >>>>>>>>>>> >>> > >> > > >> >>>>>>>>>>> >>> > EE.5.17 - "The name of the descriptor is >> > > >> >>>>>>>>>>> >>> > WEB-INF/validation.xml for >> > > >> >>>>>>>>>>> web >> > > >> >>>>>>>>>>> >>> > modules and META-INF/validation.xml for all other >> > > types >> > > >> >>>>>>>>>>> >>> > of modules." >> > > >> >>>>>>>>>>> >>> > >> > > >> >>>>>>>>>>> >>> > Given this, I don't see anywhere in the bval 1.0 >> or >> > > 1.1 >> > > >> >>>>>>>>>>> >>> > code that >> > > >> >>>>>>>>>>> handles >> > > >> >>>>>>>>>>> >>> > this. Am I missing something or does this >> > > implementation >> > > >> >>>>>>>>>>> >>> > not handle >> > > >> >>>>>>>>>>> this >> > > >> >>>>>>>>>>> >>> > case for web archives? >> > > >> >>>>>>>>>>> >>> > >> > > >> >>>>>>>>>>> >>> > Thanks, >> > > >> >>>>>>>>>>> >>> > Michael >> > > >> >>>>>>>>>>> >>> >> > > >> >>>>>>>>>>> >> > > > >> > > > >> > > >> > >>
