Absolutely, I agree with Ichiro's point of view, I'm all in for generics. Furthermore, if you apply the generics change only on WikiPlugin, the source still compiles + tests ok. The compatibility concern is valid for a rather strange corner case (somebody using the key value of the map as something different to String).
The place for introducing API changes has been -historically- between major versions, that's why the 2.10. But again, if nobody is being creative with the Map, it could be very well get into 2.9.1. The more I think of it, the more it convinces me to push it into 2.9.1. Also, I'd prefer to favour/encourage active developers/contributors over inactive plugins.. hmmm seems I'm pretty good arguing against myself. Ok. Here it's 1:00am and I must pack in order to get to Sinsheim tomorrow. I'm going with a borrowed laptop so I won't be able to push this in, but I'll check on e-mail so, if everyone's ok about this, we can commit the generics API change on WikiPlugin over the weekend (or before if anybody can make the commit throughout the week). I'd like to hear more opinions though, just to check I'm not missing anything important and that this seems like a "Good Thing (TM)" for everyone. br, juan pablo On Mon, Nov 5, 2012 at 12:03 AM, Glen Mazza <glen.ma...@gmail.com> wrote: > If I understand what you're saying "2.10.x would be the place for that", > that means you're in agreement with Ichiro, it's just you would like to > postpone this change until 2.10.x, correct? Perhaps the additional time to > make sure the change is well thought through would be good anyway. > > JSPWiki already has a build requirement of JDK 5.0 ( > http://incubator.apache.org/**jspwiki/development/building.**html<http://incubator.apache.org/jspwiki/development/building.html>) > so I'm not sure a need to support pre-5.0 versions (i.e., no generics) is > all that important. OTOH, with the proposed change, would that mean > all/most of these plugins: http://www.jspwiki.org/wiki/** > ContributedPlugins <http://www.jspwiki.org/wiki/ContributedPlugins> would > have to be rewritten? (Or is that already the case due to the classnames > being changed to org.apache.xxx?) > > Glen > > > > On 11/04/2012 04:19 PM, Juan Pablo Santos Rodríguez wrote: > >> Hi Ichiro, >> >> com.ecyrd.jspwiki namespace is also a default valid plugin namespace for >> 2.9.x, it is >> translated into org.apache.jspwiki at >> org.apache.wiki.utilClassUtil#**findClass (this was >> part of >> https://issues.apache.org/**jira/browse/JSPWIKI-719<https://issues.apache.org/jira/browse/JSPWIKI-719> >> ). >> >> The <K,V> is just to ensure that older, ungenerified, plugins can remain >> compatible with >> current trunk, while also allowing something like: >> >> MyPlugin implements WikiPlugin<String, Object> >> >> and therefore getting the typed Map at execute(..). I'm not very fond of >> introducing the >> <String, Object> change b/c of the compatibility issue, (in 2.9.x, 2.10.x >> would be >> the place for that) but maybe I'm being over-conservative here. So, other >> thoughts from >> anyone else (specially plugin devs)? >> >> >> cheers, >> jp >> >> >> On Sun, Nov 4, 2012 at 9:39 PM, Ichiro Furusato >> <ichiro.furus...@gmail.com>**wrote: >> >> Hi Juan Pablo, >>> >>> Given that the package name of all plugins is required to change from >>> com.ecyrd.jspwiki >>> to org.apache.wiki, anyone with existing plugins is going to have to >>> fix and recompile them >>> for compatibility anyway, so it's a good time to force generics into the >>> API. >>> >>> I'm not sure why you'd suggest >>> >>> public String execute( WikiContext context, Map<K,V> params ) >>> throws PluginException; >>> >>> as the method signature, as I believe we'd want to constrain K to >>> String and V to any >>> Object, so I'd advocate for Map<String,Object> myself (and all of my >>> plugins would >>> conform with that simple change). Perhaps you have a reasons I don't >>> understand. >>> >>> I've taken over the CeryleWikiPlugin set and while I'm not going to >>> commit to upgrading >>> the entire set, I'm willing to release most of them, depending on my >>> time constraints. >>> Some of rather obscure, but a number are pretty handy (such as the user >>> tagging >>> system). >>> >>> Cheers, >>> >>> Ichiro >>> >>> On Mon, Nov 5, 2012 at 8:31 AM, Juan Pablo Santos Rodríguez >>> <juanpablo.san...@gmail.com> wrote: >>> >>>> Hi Ichiro, >>>> >>>> in fact, PluginManager#parseArgs builds a HashMap<String, Object> but >>>> returns an >>>> ungenerified Map, most probably because this way binary compatibility >>>> >>> with >>> >>>> older, >>>> ungenerified plugins is ensured. Besides, as the plugin system hasn't >>>> had >>>> any issues >>>> since a long long time, the focus had been put elsewhere. >>>> >>>> But yes, aside from the compatibility issue, constraining to <String, >>>> Object> makes >>>> sense to me, as most probably this is how the Map is used by everyone >>>> (anyone not >>>> doing this way please correct me). >>>> >>>> hmm right now I'm thinking that WikiPlugin's signature could be changed >>>> >>> to >>> >>>> something like: >>>> >>>> public interface WikiPlugin <K, V> { >>>> public String execute( WikiContext context, Map<K,V> params ) >>>> throws >>>> PluginException; >>>> } >>>> >>>> so you could override the Map types on your plugins. Makes sense? >>>> >>>> >>>> br, >>>> juan pablo >>>> >>>> On Sat, Nov 3, 2012 at 10:23 AM, Ichiro Furusato >>>> <ichiro.furus...@gmail.com>**wrote: >>>> >>>> Hi, >>>>> >>>>> In looking at the 2.9.0 code I note that it doesn't use generics in a >>>>> number of >>>>> rather key places, such as WikiPlugin.execute(), so I'm guessing it >>>>> >>>> hasn't >>> >>>> been updated in a few years (at least in that regard), perhaps due to >>>>> >>>> the >>> >>>> somewhat longstanding status of the project going into Apache, or for >>>>> legacy- >>>>> compatibility reasons. >>>>> >>>>> Could someone please fill me in as to any particular reason why the Map >>>>> argument couldn't be constrained to <String,String> or <String,Object>? >>>>> I've >>>>> got quite a few plugins that are now incompatible with a generics-less >>>>> WikiPlugin API, i.e., >>>>> >>>>> public String execute( WikiContext context, Map<String,Object> >>>>> params ) throws PluginException; >>>>> >>>>> ...and rather than remove all my generics I'm wondering if it'd be >>>>> >>>> possible >>> >>>> to update the API. Yes, I know that is a brash assumption. >>>>> >>>>> Thanks much, >>>>> >>>>> Ichiro >>>>> >>>> >
