Romain, usually PropertySources should have different ordinals. The trick with the fqn Classname and the property source name is just to at least get a reproducible result during server restarts. If they change the package name then it's their fault.
PropertySources with the same ordinal are also only an issue if they contain the SAME config keys!. LieGrue, strub > On Wednesday, 28 January 2015, 11:57, Romain Manni-Bucau > <[email protected]> wrote: > > inline ;) > > > 2015-01-28 11:46 GMT+01:00 Tresch, Anatole > <[email protected]>: >> Well, for me it is questionable if we need this: >> >> 1) I don’t see much advantages adding another abstraction for that >> 2) we HAVE a well defined order currently based on prio and class name, so > there are no conflicts! >> > > class name (worse when using fqn vs simple name) is not something you > can bet on, too fragile + you can get multiple JsonPropertySource with > the same ordinal and same values for instance. You'd need to use > idendityhashCode to be deterministic then. > >> BTW throwing exception in a EE environment can have disastrous > consequences, so it is not always a good idea to just throw an > exception...especially since it may result on some assembly htat might be out > of > control of a developer or operator. So I would suggest writing w warning is > the > better way to handle it. >> > > In EE I don't expect to use it "like it" but in a CDI extension at > minimum so not a big deal, will hopefully make the deployment fail. > > >> Cheers, >> Anatole >> >> >> >> -----Original Message----- >> From: Romain Manni-Bucau [mailto:[email protected]] >> Sent: Mittwoch, 28. Januar 2015 10:23 >> To: [email protected]; Mark Struberg >> Subject: Re: Multiple PropertySources with the same ordinal >> >> Why don't we just introduce a sorter API? Then ordinal would be @Order >> and used as default only. This would allow us to fail when there is a >> conflict since it is solvable on user side and shouldn't happen >> anyway. >> >> >> Romain Manni-Bucau >> @rmannibucau >> http://www.tomitribe.com >> http://rmannibucau.wordpress.com >> https://github.com/rmannibucau >> >> >> 2015-01-28 10:08 GMT+01:00 Mark Struberg <[email protected]>: >>>> Did you read the Javadocs? It is clearly written that we sort >>>> 1) by ordinal >>>> 2) by fully qualified class name. >>> >>> Side note as I'm atm busy with a $$dayjob issue. We might > additionally need to sort via the PropertySource name. That would be > important > if you have 2 PropertyFilePropertySource instances with different URLs. They > should get a well defined sorting as well. >>> >>> >>> After that there really should be no == in the sorting anymore, right? >>> >>> LieGrue, >>> strub >>> >>> >>> >>> >>> >>>> On Tuesday, 27 January 2015, 21:26, Anatole Tresch > <[email protected]> wrote: >>>> > -1 for failing. >>>> >>>> Did you read the Javadocs? It is clearly written that we sort >>>> 1) by ordinal >>>> 2) by fully qualified class name. >>>> >>>> So there is an order that even is not dependend on classloaders. >>>> >>>> >>>> >>>> 2015-01-27 19:52 GMT+01:00 Oliver B. Fischer > <[email protected]>: >>>> >>>>> Hi, >>>>> >>>>> I am just writing some unit tests. One of them adds multiple > property >>>>> sources with the same key and the same ordinal. As we can not > decide >>>>> which one is the right one we should throw an exception. >>>>> >>>>> WDYT? >>>>> >>>>> Oliver >>>>> >>>>> -- >>>>> N Oliver B. Fischer >>>>> A Schönhauser Allee 64, 10437 Berlin, Deutschland/Germany >>>>> P +49 30 44793251 >>>>> M +49 178 7903538 >>>>> E [email protected] >>>>> S oliver.b.fischer >>>>> J [email protected] >>>>> X http://xing.to/obf >>>>> >>>>> >>>> >>>> >>>> -- >>>> *Anatole Tresch* >>>> Java Engineer & Architect, JSR Spec Lead >>>> Glärnischweg 10 >>>> CH - 8620 Wetzikon >>>> >>>> *Switzerland, Europe Zurich, GMT+1* >>>> *Twitter: @atsticks* >>>> *Blogs: **http://javaremarkables.blogspot.ch/ >>>> <http://javaremarkables.blogspot.ch/>* >>>> >>>> *Google: atsticksMobile +41-76 344 62 79* >>>> >
