----- Original Message -----
From: "Gilad Chaplik" <gchaplik(a)redhat.com>
To: "Tomas Jelinek" <tjelinek(a)redhat.com>
Cc: engine-devel(a)ovirt.org
Sent: Wednesday, January 30, 2013 11:42:22 AM
Subject: Re: [Engine-devel] [engine-devel] frontend builders proposal
----- Original Message -----
From: "Tomas Jelinek" <tjelinek(a)redhat.com>
To: "Gilad Chaplik" <gchaplik(a)redhat.com>
Cc: engine-devel(a)ovirt.org
Sent: Wednesday, January 30, 2013 11:11:41 AM
Subject: Re: [Engine-devel] [engine-devel] frontend builders proposal
[adding missing part from the thread]
I agree that this refactoring can significantly help us reduce code
complexity,
there is another issue that your suggestion doesn't address,
but we may want take the opportunity to address it if we are already
considering refactoring for this code:
> > this dialog demonstrates the greatest difference (IMHO)
between
> > server side pages to applets,
> > the back and forth filling the form by retrieving all elements one
> > by one.
> > I would think of a concept similar to server side pages, i.e.
> > retrieving all
> > data, visibility and even validations (compat?), in a single
> > request, and let the
> > server have the logic.
> Not sure if I have understood you correctly - do you mean moving the
> frontend logic (e.g. visibility of fields according to some
> selection) and also the frontend models
> to the server? I would say it is better to have this on FE.
IMHO, we should uiqeury (or something similar) or uiplugin (or striving towards it), that
get the entire dialog data (with /without ui).
I'm not sure what is the benefit with your refactoring, can you elaborate on that?
The plan is to decouple the FE from the BE more in the future by changing the
current API to call the REST. My patch is actually a step
to this direction because we will not be able to use the backend models anymore and will
need to use DTOs. If we will have the building logic centralized,
it will be much more simple to refactor them to use the new DTOs. Any coupling with the
server (like getting the dialog data instead of business data) is
a step from the moving to generic REST API, and since this is a long term plan, I would be
very careful.
but let see what others think.
>
> However having the validations common between FE and BE would be
> great. We could prepare a different patch for this.
I think that common validations are nice to have, and it is least of
our problems.
agree
>
>
> ----- Original Message -----
> From: "Gilad Chaplik" <gchaplik(a)redhat.com>
> To: "Tomas Jelinek" <tjelinek(a)redhat.com>
> Cc: engine-devel(a)ovirt.org, "Eli Mesika" <emesika(a)redhat.com>
> Sent: Monday, January 28, 2013 5:08:41 PM
> Subject: Re: [Engine-devel] [engine-devel] frontend builders proposal
>
> ----- Original Message -----
> > From: "Eli Mesika" <emesika(a)redhat.com>
> > To: "Tomas Jelinek" <tjelinek(a)redhat.com>
> > Cc: engine-devel(a)ovirt.org
> > Sent: Saturday, January 26, 2013 10:58:57 PM
> > Subject: Re: [Engine-devel] [engine-devel] frontend builders
> > proposal
> >
> >
> >
> > ----- Original Message -----
> > > From: "Tomas Jelinek" <tjelinek(a)redhat.com>
> > > To: engine-devel(a)ovirt.org
> > > Sent: Friday, January 25, 2013 11:55:43 AM
> > > Subject: [Engine-devel] [engine-devel] frontend builders proposal
> > >
> > > Hi All,
> > >
> > > as many of you may know, the way how the frontend and backend
> > > models
> > > are built on frontend (uicommonweb project) is not really ideal.
> > > Currently this logic is copy pasted over and over again to
> > > different
> > > places where it is needed with minor changes to fulfill the
> > > specific
> > > requirements. It is not only aesthetically problematic, but I
> > > recall
> > > tons of bugs caused by introducing a new field and forgetting to
> > > add
> > > it to every place it is used in GUI.
> > >
> > > Now, as there will be big changes in the VM/Template models
> > > (
http://www.ovirt.org/Features/Instance_Types), so the way how
> > > the
> > > VM, Template, Pool and also the newly created Instance Types
> > > models
> > > are being built has to be touched anyhow, it is a great
> > > opportunity
> > > to rethink the way how we do it.
> > >
> > > I have created a simple infrastructure
> > > (
http://gerrit.ovirt.org/#/c/10874/) which could be used for it,
> > > and
> > > a PoC patch which uses this infrastructure
> > > (
http://gerrit.ovirt.org/#/c/11354/). Please note that the PoC is
> > > not really impressive in means of removing duplications, I wanted
> > > to
> > > start with the simplest possibility.
> > >
> > > The principles behind the infrastructure:
> > > - have small, well named, easy to understand and reuse builders
> > > - this builders can be chained together or embedded to each other
> > > to
> > > build the full resulting object (composite pattern)
> > > - this builders can be asynchronous, and the next builder in the
> > > chain has to be executed only when the current is completely done
> > >
> > > The structure:
> > > - the base is an interface called Builder which has a method
> > > build(source, destination, rest)
> > > - the builder implementing this interface
> > > + will get the source and destination objects
> > > + copies whatever he wants from source to destination
> > > + when done, executes build on the first element of the rest
> > > + this may sound awkward, but this is the way how the async
> > > calls
> > > can be "linearized" in a general way, not embedding
anonymous
> > > class into anonymous
> > > class into anonymous class... as we do it today.
> > > + for synchronous builders, there is a BaseSyncBuilder which
> > > takes care of this boilerplate calling of next and exposes a
> > > simple method
> > > build(S source, D destination)
> > > + to simplify the creating and running the chain of builders,
> > > there
> > > is a BuilderExecutor class (can be created as sync or async)
> > >
> > > So, a simple example - even more simple than the PoC patch :)
> > >
> > > //create the first builder
> > > class FirstLetterBuilder extends BaseSyncBuilder<String,
> > > StringBuilder> {
> > > @Override
> > > protected void build(String source, StringBuilder destination) {
> > > // copy the first letter to the destination
> > > destination.append(source.charAt(0));
> > > }
> > > }
> > >
> > > //create the second builder
> > > class SecondLetterBuilder extends BaseSyncBuilder<String,
> > > StringBuilder> {
> > > @Override
> > > protected void build(String source, StringBuilder destination) {
> > > // copy the second letter to the destination
> > > destination.append(source.charAt(1));
> > > }
> > > }
> > >
> > > // usage
> > > ...
> > > // create the destination object
> > > StringBuilder res = new StringBuilder();
> > >
> > > // configure the executor with the two builders
> > > BuilderExecutor<String, StringBuilder> executor = new
> > > BuilderExecutor<String, StringBuilder>(
> > > new FirstLetterBuilder(),
> > > new SecondLetterBuilder()
> > > );
> > >
> > > // execute the builder chain ("ab" is the source, res the
> > > destination)
> > > executor.build("ab", res);
> > >
> > > // use the result
> > > ...
> > >
> > > That's it. And the nice part is, that this FirstLetterBuilder and
> > > SecondLetterBuilder can be reused anywhere or combined with any
> > > other builders.
> > >
> > > Any comments on this will be more than welcome!
> >
> > great and really simplifies work and eliminate bugs resulted from
> > copy/past code
> > gave +1
> > Thanks
> > Eli
>
> Hi guys,
>
I agree that this refactoring can significantly help us reduce code
complexity,
there is another issue that your suggestion doesn't address,
but we may want take the opportunity to address it if we are already
considering refactoring for this code:
> this dialog demonstrates the greatest
difference (IMHO) between
> server side pages to applets,
> the back and forth filling the form by retrieving all elements one by
> one.
> I would think of a concept similar to server side pages, i.e.
> retrieving all
> data, visibility and even validations (compat?), in a single request,
> and let the
> server have the logic.
>
> Thanks,
> Gilad.
>
> > >
> > > Thank you,
> > > Tomas
> > > _______________________________________________
> > > Engine-devel mailing list
> > > Engine-devel(a)ovirt.org
> > >
http://lists.ovirt.org/mailman/listinfo/engine-devel
> > >
> > _______________________________________________
> > Engine-devel mailing list
> > Engine-devel(a)ovirt.org
> >
http://lists.ovirt.org/mailman/listinfo/engine-devel
> >
>