15 Nov
2018
15 Nov
'18
1:23 p.m.
On Thu, Nov 15, 2018 at 11:13 AM Thomas Mortagne
<thomas.mortagne(a)xwiki.com> wrote:
On Thu, Nov 15, 2018 at 11:06 AM Adel Atallah <adel.atallah(a)xwiki.com> wrote:
This is not enforced by the code. You know that features are
associated to groups *because* they are bound to the same property.
Anyway I'm +1 to do it this way.
Ok so it seems like we are getting back to the proposition we made with Vincent.
We need one annotation to enforce the dependence between parameters
(reference and type in our example) and another one that can be used
to *deduce* conflicting parameters.
I don't understand how a hierarchy of groups
can help us specify a
dependence between parameters.
I don't think it does, it's just that since we are defining groups
having subgroups would be useful visually.
A parameter is either in the same group
as another one or it is not. The hierarchy seems to focus on problems
that we are not trying to solve here.
The original proposal was similar to what Thomas proposed, but without
hierarchy:
@Alternative("reference")
@Group("entityReference")
reference
@Alternative("reference")
@Group("entityReference")
type
@Alternative("reference")
page
@Alternative("reference")
document
where "Alternative" is the same as "Feature". Now Marius didn't
agree
with that because the "Alternative" annotation should not be bind to
"reference" and "type" parameters but to the group
"entityReference"
And as I said in my proposal the features are associated to the group,
not the properties. I agree that associating it to the property (and
ending up with half of a group conflicting with half of another) does
really make sense.
,
which is not possible to do without creating other classes. I don't
think this is an issue to put the "Alternative" annotation on
"reference" and "type" because we should have all the necessary
information to *deduce* the conflicting parameters. It's true that
removing the "Alternative" annotation of one of "reference" or
"type"
should produce the same result though, which could be confusing.
On Thu, Nov 15, 2018 at 10:23 AM Marius Dumitru Florea
<mariusdumitru.florea(a)xwiki.com> wrote:
>
> On Thu, Nov 15, 2018 at 10:51 AM Thomas Mortagne <thomas.mortagne(a)xwiki.com>
> wrote:
>
> > I'm also really not a fan of having to implement a component just to
> > indicate that two groups of properties are conflicting.
> >
> > +1 for making @Group support a hierarchy, that's indeed nice.
> >
> > For for conflicting we need a dedicated annotation IMO.
> >
> > So starting from your previous example I would expect something like:
> >
> >
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> >
> > @PropertyGroup("target")
> > @PropertyFeature("reference")
> > page
> >
> > @PropertyGroup({"target", "entityReference"})
> > @PropertyFeature("reference")
> > reference
> >
> > @PropertyGroup({"target", "entityReference"})
> > type
> >
> > @PropertyGroup("target")
> > @PropertyFeature("reference")
> > document
> >
> >
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
> >
>
> I don't think this is complete. The following doesn't make sense:
>
> {{include page="..." type="..."/}}
>
> and neither this:
>
> {{include document="..." type="..." /}}
>
> So it's not the reference parameter alone that provides the
"reference"
> feature. The pair / group of parameters (reference and type) are providing
> the "reference" feature. This is why I think there is the need to specify
> the "feature" on the sub group "entityReference" not on the
parameter. And
> to do this we need another class..
>
> >
> >
> > or
> >
> >
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
> >
> > @PropertyGroup("target", features = "reference")
> > page
> >
> > @PropertyGroup({"target", "entityReference"}, features =
"reference")
> > reference
> >
> > @PropertyGroup({"target", "entityReference"})
> > type
> >
> > @PropertyGroup("target", features = "reference")
> > document
> >
> >
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
> >
> >
>
> > * PropertyGroup define the hierarchy (also proposed a String[] instead
> > of String based value to show all possible ways to pass the hierarchy
> > value)
> >
>
> +1 for this
>
>
> > * PropertyFeature (name is negotiable :)) or PropertyGroup
"features"
> > field associate the group with a set of unique "features". This is
the
> > same logic than for extensions where several groups with with a shared
> > feature are in conflict
> >
>
> You're not associating the feature to the group. That is the problem IMO.
> You are associating the feature to the parameter. For instance:
>
> @PropertyGroup("foo", features = "input")
> one
>
> @PropertyGroup("foo", features = "output")
> two
>
> Is the "input" and "output" feature associate to the
"foo" group or to the
> parameters one and two respectively?
>
> Thanks,
> Marius
>
>
> >
> > We could also decide to support only one feature per group right now
> > since we don't yet have the need for several but it felt more natural
> > like this.
> >
> > On Thu, Nov 15, 2018 at 8:04 AM Vincent Massol <vincent(a)massol.net>
wrote:
> > >
> > >
> > >
> > > > On 15 Nov 2018, at 08:02, Vincent Massol <vincent(a)massol.net>
wrote:
> > > >
> > > >
> > > >
> > > >> On 15 Nov 2018, at 06:29, Marius Dumitru Florea <
> > mariusdumitru.florea(a)xwiki.com> wrote:
> > > >>
> > > >> On Wed, Nov 14, 2018 at 5:12 PM Vincent Massol
<vincent(a)massol.net>
> > wrote:
> > > >>
> > > >>> I thought about something like this but I discarded it as I
find this
> > > >>> complicated for something that should be relatively simple.
> > > >>
> > > >>
> > > >> I don't think it's that complicated because:
> > > >>
> > > >> * Conflicting parameters should be an exception, not the rule.
What
> > other
> > > >> macros, besides include / display, need this?
> > > >> * If you just want to group macro parameters for display then
you
> > only need
> > > >> to use the @Group annotation. You don't need to implement a
> > ParameterGroup.
> > > >> The ParameterGroup is needed only for conflicting parameters
(ATM).
> > > >
> > > > Sure but it’s still 10x more complicated than just having everything
> > in one place in the parameters class with annotations as was suggested
> > initially.
> > >
> > > And requires unnecessary component instances that will stay in the EM
> > for no need. The way to describe the descriptor is transient and only
> > serves to generate the macro descriptors. In the end what’s important is
> > the descriptor format.
> > >
> > > Thanks
> > > -Vincent
> > >
> > > >
> > > > Thanks
> > > > -Vincent
> > > >
> > > >>
> > > >> Thanks,
> > > >> Marius
> > > >>
> > > >>
> > > >>> I’d prefer to have some simple annotations if possible. In
other
> > words, if
> > > >>> feels a bit of over-engineering for the need. Now I have to
admit
> > that I
> > > >>> stopped following this thread after the original proposal so
maybe
> > I’m just
> > > >>> completely off :)
> > > >>>
> > > >>> Thanks
> > > >>> -Vincent
> > > >>>
> > > >>>> On 14 Nov 2018, at 15:51, Marius Dumitru Florea <
> > > >>> mariusdumitru.florea(a)xwiki.com> wrote:
> > > >>>>
> > > >>>> WDYT about:
> > > >>>>
> > > >>>> -----8<----- IncludeMacroParameters ----------
> > > >>>> @Group("target")
> > > >>>> page
> > > >>>>
> > > >>>> @Group("target/entityReference")
> > > >>>> reference
> > > >>>>
> > > >>>> @Group("target/entityReference")
> > > >>>> type
> > > >>>>
> > > >>>> @Group("target")
> > > >>>> document
> > > >>>>
> > > >>>> section
> > > >>>>
> > > >>>> context
> > > >>>> ----->8---------------
> > > >>>>
> > > >>>> That is: specify *only* the group hierarchy in the macro
parameter
> > > >>>> descriptor. This would produce the following hierarchy:
> > > >>>>
> > > >>>> * <target>
> > > >>>> ** page
> > > >>>> ** <entityReference>
> > > >>>> *** reference
> > > >>>> *** type
> > > >>>> ** document
> > > >>>> * section
> > > >>>> * context
> > > >>>>
> > > >>>> Next, for the cases where we want to customize the
behavior of a
> > group,
> > > >>> we
> > > >>>> introduce a component role ParameterGroup. For instance,
for the
> > "target"
> > > >>>> parameter group of the Include Macro we would create
> > > >>>>
> > > >>>> @Named("include/target")
> > > >>>> public class TargetParameterGroup implements
ParameterGroup {}
> > > >>>>
> > > >>>> To specify that the members of a parameter group are
exclusive we
> > can
> > > >>>> either use a method in the ParameterGroup interface
(e.g.
> > isExclusive())
> > > >>> or
> > > >>>> use an annotation on the implementation
TargetParameterGroup.
> > > >>>>
> > > >>>> Thanks,
> > > >>>> Marius
> > > >>>>
> > > >>>>
> > > >>>> On Tue, Nov 13, 2018 at 12:03 PM Adel Atallah <
> > adel.atallah(a)xwiki.com>
> > > >>>> wrote:
> > > >>>>
> > > >>>>> Hello,
> > > >>>>>
> > > >>>>> I'd like to briefly summarize the situation so
that we can make
> > some
> > > >>>>> progress.
> > > >>>>>
> > > >>>>> What we have:
> > > >>>>> * We define "parameters" in a macro by
creating a Java Bean, which
> > > >>>>> provides all the getters and setters of the
parameters we want.
> > > >>>>> * We can use annotations on these getters/setters to
define some
> > > >>>>> behavior or metadata for these parameters
(description, mandatory,
> > > >>>>> deprecated...)
> > > >>>>>
> > > >>>>> What we want:
> > > >>>>> * Being able to handle conflicting parameters. For
instance when we
> > > >>>>> deprecate a parameter and add a new one to replace
it, we should be
> > > >>>>> able to either use the deprecated parameter or the
new one but not
> > > >>>>> both.
> > > >>>>> * We also want to group parameters that are related
to each other.
> > > >>>>>
> > > >>>>> What we proposed:
> > > >>>>> * Use annotations on the parameters to express the
conflict.
> > > >>>>> * Marius proposed to see the problem as a boolean
expression such
> > as:
> > > >>>>> (page XOR (reference AND type) XOR document) OR
section OR context.
> > > >>>>> This would translate as: the user can use the
'section' and/or
> > > >>>>> 'context' parameters (if they want), can use
only one of these
> > > >>>>> parameters: 'page', ('reference' and
'type') or 'document', where
> > > >>>>> 'reference' and 'type' depend on each
other and you can't use one
> > > >>>>> without the other.
> > > >>>>> * You can see on previous e-mails the kind of
annotations we
> > proposed
> > > >>>>> to solve the issue.
> > > >>>>>
> > > >>>>> Thanks,
> > > >>>>> Adel
> > >
> >
> >
> > --
> > Thomas Mortagne
> >
--
Thomas Mortagne