« Return to Thread: GlazedList GWT
Re: GlazedList GWT
by James Lemieux
::
Rate this Message:
Jesse and I chatted about the two questions (naming and package structure) and have some ideas:
1. We would like to offer (not require though) you the chance to store your source code in GL's CVS repository so that the two projects are developed in close lockstep. This is just an offer, and you are of course free to host your source code wherever you like, but we thought that it may be easier if your build system literally built from our sources in the same source tree. Our thinking is that:
a) any developments in the GL core that could affect GLGWT will be noticed and dealt with immediately
a) development will occur in a more joint fashion since developers have access to working versions of GL and GLGWT built in lockstep... there is no ambiguity about which GL and GLGWT jars "work with each other"
b) all GWT source and artifacts would live under \gwt at the root level, so you'd "have your own space", but your build files are free to refer to ..\source\ca\odell... directly, and your build system (e.g. ANT build.xml if that's what you use) can pick and choose what classes/interfaces you can share unchanged with the GL core, and what you may want to fork...
2. Regarding package naming, Jesse's thoughts are that it is conceivable that one day we could invent a type of GWT EventList that actually spans from server to client... i.e. I call myGWTBasicEventList.add("Hello World"); on the server in Java using standard GL and all *clients* receive that ListEvent and propagate it naturally throughout the Javascript side of the pipeline in GLGWT. The power is... unimaginable. :-) If such a beast were to ever come to fruition, there's a possiblity we'd want to do some GWT-style class substitution, and it may be nicer for us to have similar/identical package naming in that case.... thoughts? But, I also admit this is pie-in-the-sky speculation, and not a hard-and-fast reason for you to mandatorily adopt our package structure...
3. One of the artifacts produced in a "full GL build" would thus be something like "glazedlists-1.9.0_gwt.jar" which would be a jar of .java files produced from the \gwt area of source control.
4. You could use our Google Site for all of your community documentation, etc. Or, if you wanted, you could create your own public face and just keep the source code in our repository. It's really whatever works for you.
5. If we do want to be bound tightly, that would probably necessitate that you use GL style interfaces over Google Collections (e.g. Matcher vs. Predicate). Of course we can trivially make adapters between the two, but this is still a point that should be made now so there is no ambiguity.
Please don't feel we're trying to strongarm you into being assimilated into the GL project. Our motivations are purely for making the best environment for GL and GLGWT to succeed together. If the projects are so closely aligned that GL developers are building GLGWT and vice-versa then I feel we have the best chance at code reuse and refining issues between the projects when they arise. We don't want to steal your thunder here at all...
James (& Jesse)
1. We would like to offer (not require though) you the chance to store your source code in GL's CVS repository so that the two projects are developed in close lockstep. This is just an offer, and you are of course free to host your source code wherever you like, but we thought that it may be easier if your build system literally built from our sources in the same source tree. Our thinking is that:
a) any developments in the GL core that could affect GLGWT will be noticed and dealt with immediately
a) development will occur in a more joint fashion since developers have access to working versions of GL and GLGWT built in lockstep... there is no ambiguity about which GL and GLGWT jars "work with each other"
b) all GWT source and artifacts would live under \gwt at the root level, so you'd "have your own space", but your build files are free to refer to ..\source\ca\odell... directly, and your build system (e.g. ANT build.xml if that's what you use) can pick and choose what classes/interfaces you can share unchanged with the GL core, and what you may want to fork...
2. Regarding package naming, Jesse's thoughts are that it is conceivable that one day we could invent a type of GWT EventList that actually spans from server to client... i.e. I call myGWTBasicEventList.add("Hello World"); on the server in Java using standard GL and all *clients* receive that ListEvent and propagate it naturally throughout the Javascript side of the pipeline in GLGWT. The power is... unimaginable. :-) If such a beast were to ever come to fruition, there's a possiblity we'd want to do some GWT-style class substitution, and it may be nicer for us to have similar/identical package naming in that case.... thoughts? But, I also admit this is pie-in-the-sky speculation, and not a hard-and-fast reason for you to mandatorily adopt our package structure...
3. One of the artifacts produced in a "full GL build" would thus be something like "glazedlists-1.9.0_gwt.jar" which would be a jar of .java files produced from the \gwt area of source control.
4. You could use our Google Site for all of your community documentation, etc. Or, if you wanted, you could create your own public face and just keep the source code in our repository. It's really whatever works for you.
5. If we do want to be bound tightly, that would probably necessitate that you use GL style interfaces over Google Collections (e.g. Matcher vs. Predicate). Of course we can trivially make adapters between the two, but this is still a point that should be made now so there is no ambiguity.
Please don't feel we're trying to strongarm you into being assimilated into the GL project. Our motivations are purely for making the best environment for GL and GLGWT to succeed together. If the projects are so closely aligned that GL developers are building GLGWT and vice-versa then I feel we have the best chance at code reuse and refining issues between the projects when they arise. We don't want to steal your thunder here at all...
James (& Jesse)
On Wed, Apr 15, 2009 at 11:57 AM, Holger Brands <hbrands@...> wrote:
Hey Thomas,
thanks for the infos.
The disadvantage of having a different GL implementation for GWT
is the lack of support for existing domain models that already use
the "normal" Glazed Lists.
You would have to create some kind of DTO model and copy the list
contents for use with a GWT client.
But often, you have to create these DTOs for other reasons anyway,
so I guess it's not a big problem for most people.
Regarding property change support, we've used
GWTx (http://code.google.com/p/gwtx/) with success in the past,
so it might be helpful to you, too.
With regard to google-collections: is it compatible with GWT or is
there a GWT-specific version somewhere ?
As James already said, don't hesitate to make your Glazed Lists GWT
version available somewhere, so other users can give it a try.
As for the base package name, I don't have any preference.
It just shouldn't collide with the original GL package; "ca.odell" is
probably suboptimal in your case anyway...
Thanks for your effort,
Holger
>
> Holger,
>
> Let me try to reply to a few of your questions/thoughts..
>
> 1. About trying to get GL useable in GWT but without a different
> implementation. Its probably possible but only with a pretty severe
> refactoring of the core of GL. Some very useable java-library calls are
> simply not available in GWT, and if you would want to continue using them,
> you'd need to abstract that away.
>
> 2. As for sticking to the EventList interface with a no-op LockAdaptor,
> that's very doable. But trying to keep the same implementations, doesn't
> quite work. My first inclination was going to be to suggest creating a
> sister package called ca.odell.glazedgwt which would really begin with
> almost all new implmentation classes. But many of the originals require
> only minor changes, so its possible to try to tweak those.
>
> 3. Property change and Reflection. Yeah, this is a tough topic in GWT. In
> my companies framework, we get around this issue by generating complex
> ValueObjects with automatic Property accessors. There are a few opensource
> projects that do this ok as well. Essentially the pattern is to work out a
> way to auto-magically (which means 'generator' in gwt), create support for
> access to properties via get(String), set(String, Object) methods on your
> Beans. Gwittr has support for a generator that does this, GXT does too. We
> could have a few options, i believe even working out a way to just require
> that any Bean which GL has support for implement a simplfied interface with
> the requisite methods. It would be great if Google standardized that sort
> of interface in the GWT core, but they haven't.
>
>
> 4. Other API problems. There were definitely a few, and like i originally
> said, I over reduced at first, and I'm only adding back functionality now.
> The great thing, is that the core complex algorithms worked with almost zero
> modification. That's really what encouraged me to continue at the begining.
> If you are into reviewing the results, i'd love to send over repackaged
> version of the library in the state i have it now (with a few changes).
>
>
> Also what do you think about my suggestion/plans to create a dependancy on
> the new google-collections 1.0 release. I think if i did that (which i
> definitely want for my project). I could thin the package out even more, and
> reduce duplication (which really costs you in js apps), with a library most
> uses will likely have. The downside to this, of course, is that it means an
> even further deviation from the GL core.
>
>
> Thanks for the reply, talk to you again soon.
>
> tom.
>
>
> Holger Brands wrote:
> >
> > Thomas,
> >
> > that's interesting news.
> > We intend to provide *some support* for GWT in the future,
> > but we didn't start to plan or implement anything yet.
> >
> > I guess the first step would be to identify the core of GL (without UI
> > stuff)
> > which we could and would like to support on GWT.
> > Ideally, GWT support for this core would not require a different
> > implementation, but I don't know yet if that's feasible at all.
> > (Think about GL code, that you would like to use on the client *and*
> > server)
> >
> > Some quick thoughts:
> >
> > - I wonder if we could stick to the EventList interface and use a no-op
> > LockAdapter for GWT
> >
> > - how to support property change events? (using GWTx perhaps)
> >
> > - how to support/deal with reflection?
> >
> > - what other API problems did you encounter?
> >
> > If you would like to share more details with us, that would be great.
> >
> > Thanks,
> > Holger
> >
> >>
> >> Hey guys,
> >>
> >> I've put together a branch of GL which runs on GWT. I reduced the
> >> functionality quite a bit to the things that i think are applicable to
> >> the
> >> GWT environment (and what i needed there).
> >>
> >> A few examples, of the changes i made:
> >> Definitely good:
> >> 1. Removed all locking, and thread management code. JS is single
> >> threaded.
> >> 2. Altered any use of unsupported jdk APIs to ones that work in GWT
> >> 3. Removed unnecessary swing/swt/networking code which wouldn't build in
> >> gwt
> >>
> >> Arguably not-so-good:
> >> 1. I removed more than i really needed to. Some features could be made
> >> to
> >> work that i removed just to thin out the package. These could be put
> >> back
> >> in without too much cost since GWT is good at only including what you
> >> really
> >> need at runtime.
> >> 2. I renamed a few things that i really didn't need to, this would be
> >> easy
> >> to changed back. Basically i renamed EventList to GlazedList.
> >> 3. I'm in the process of removing duplication in the GL library with the
> >> Google-collections-GWT library, using the google versions to avoid having
> >> two in my applications. At a glance i think this should include moving
> >> Matcher to Predicate, and utilizing their multimap implmentation in place
> >> of
> >> the GL one.
> >>
> >> My version is definitely incomplete, for instance i did not reimplement
> >> the
> >> equivalent swing table adapters for the gwt ones. My project is using
> >> GXT
> >> and so i built adapters to that (with a dependancy on my framework).
> >>
> >> I read the original note from Jesse on the subject of GWT, where he
> >> mentioned that GL can only really work well in an env where the whole
> >> list
> >> is in memory. And while i agree with much of the premise of that, i'd
> >> argue
> >> that JS environments are really much more constrained by how much can be
> >> rendered to DOM than by how many objects and data can be kept in memory.
> >> The fundamentals of what GL offers, really can make a big difference when
> >> developing any UI, even JavaScript ones.
> >>
> >> I'm really interested in some of the additional comments about how GL
> >> often
> >> trades off memory for speed, and any suggestions of where we could
> >> perhaps
> >> provide alternate implementations with different/configurable tradeoffs.
> >>
> >> Thanks again for GL, its really a very smart tool.
> >>
> >> tom.
> >> --
> >> View this message in context:> >> http://www.nabble.com/GlazedList-GWT-tp23002550p23002550.html
> >> Sent from the GlazedLists - Dev mailing list archive at Nabble.com.
> >>
> >>
> >> ---------------------------------------------------------------------
> >> To unsubscribe, e-mail: dev-unsubscribe@...
> >> For additional commands, e-mail: dev-help@...
> >>
> >>
> >
> >
> > __________________________________________________________________________
> > Verschicken Sie SMS direkt vom Postfach aus - in alle deutschen und viele
> > ausländische Netze zum gleichen Preis!
> > https://produkte.web.de/webde_sms/sms
> >
> >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: dev-unsubscribe@...> View this message in context: http://www.nabble.com/GlazedList-GWT-tp23002550p23039520.html
> > For additional commands, e-mail: dev-help@...
> >
> >
> >
>
> --
> Sent from the GlazedLists - Dev mailing list archive at Nabble.com.____________________________________________________________________
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: dev-unsubscribe@...
> For additional commands, e-mail: dev-help@...
>
>
Psssst! Schon vom neuen WEB.DE MultiMessenger gehört?
Der kann`s mit allen: http://www.produkte.web.de/messenger/?did=3123
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@...
For additional commands, e-mail: dev-help@...
« Return to Thread: GlazedList GWT
| Free embeddable forum powered by Nabble | Forum Help |
