On Tue, Aug 21, 2012 at 10:18 PM, Vincent Massol <vincent(a)massol.net> wrote:
On Aug 21, 2012, at 2:18 PM, Fabio Mancinelli wrote:
On Fri, Jul 27, 2012 at 10:12 AM, Ludovic Dubost
<ludovic(a)xwiki.com> wrote:
As part of rest improvements to display pretty
names of users and
other improvements, I'm getting CLIRR errors because of API changes of
the model and of public class:
1/ Model CLIRR error because the version field has been moved to
PageSummary from Page. Page extends PageSummary. I need the version
field also in representations sending back only PageSummaries.
Unfortunately CLIRR does not realize that the version field is still
there when moved to the super class. I believe it's safe to ignore
this error. Howerver I've put ignore all errors on the Page class as I
don't have a way to ignore this specific error
Yep, I think it's safe. We're adding stuff in a representation (page
summary) and keeping it also in the other, so API-wise it's ok.
Note that CLIRR doesn't have false positives so if it complains it means there's
a breakage. The only decision to take is whether it's an "acceptable"
breakage, i.e. we voluntarily break assuming that nobody was using it and accept that a
few users might get broken.
Actually I was talking of the REST-API ... Re CLIRR, what is it
complaining for exactly? If I move a method in a superclass keeping
the same access level, it is still available in the subclasses for
clients calling it. So why complain?
2/ CLIRR errors because of parameter additions to
objects that are
used (I think) only internally by the REST server API. Here are the
errors:
[ERROR] org.xwiki.rest.DomainObjectFactory: In method 'public
org.xwiki.rest.model.jaxb.Attachment
createAttachment(org.xwiki.rest.model.jaxb.ObjectFactory,
java.net.URI, com.xpn.xwiki.api.Attachment, java.lang.String,
java.lang.String)' the number of arguments has changed
The DomainObjectFactory is actually a utility class that is used to
build REST-model objects from XWiki-model objects.
It has been created just to prevent code duplication in resource
implementations.
The question is whether this is supposed to be a SPI or not.
No. It's just an utility class where to put the code for building
model objects that is used in different places.
The idea is the same of
https://github.com/fmancinelli/xwiki-platform/blob/master/xwiki-platform-co…
Now I think
it's unlikely that somebody uses it outside the REST
module (a quick grep confirmed this for platform).
The only use case for a developer of a module to use this class is if
she wants to return a REST-model object and build it using the utility
methods.
I think this is quite unlikely.
AFAIU all parameters additions are about "pretty names"
(
https://github.com/ldubost/xwiki-platform/compare/master...bd49bcc84e1dec3d…)
If we want to be conservative we might do the following: we can add
the new methods and preserve the old ones making them call the new
ones with default parameters.
* false in methods like this
https://github.com/ldubost/xwiki-platform/compare/master...bd49bcc84e1dec3d…
* null, false in methods like this
https://github.com/ldubost/xwiki-platform/compare/master...bd49bcc84e1dec3d…
. This implies that in the new implementation the if statement should
also check for null values (like in this case:
https://github.com/ldubost/xwiki-platform/compare/master...bd49bcc84e1dec3d…)
We could also think about whether continuing to keep this class in the
public API. It could make sense but I think that nobody will ever use
it so we can start to @deprecate it and eventually move it in internal
packages.
Based on what you say I'd say that these classes/methods were put public by mistake
and should all be moved to the internal package without going through deprecation.
Of course this means no other other XWiki modules should use them either since
they're internal.
Yep.
Also this means that we don't provide any SPI
either since SPI are user-public. Is that what we want?
Yep
Thanks,
Fabio
Thanks
-Vincent
_______________________________________________
devs mailing list
devs(a)xwiki.org
http://lists.xwiki.org/mailman/listinfo/devs