Thread: Re: [jdee-devel] Status of project and adding 1.5 generics (Page 2)
Brought to you by:
paullandes
From: Ming-Wei C. <mch...@ui...> - 2008-12-08 22:07:45
|
jo...@ve... writes: > This is what I have done, even if its not quite complete. > My next idea was to use rhino instead of beanshell for reflection etc. > Beanshell would still be used as an interactive java shell in jdee. The idea looks cool! Looking forward to it :) > > One of the benefits of this is that java completion etc will be > available out of the box for all Emacs users in the future. > |
From: Paul L. <la...@ma...> - 2008-12-09 00:45:20
|
Let's revisit this once we nail down how/if/why we're going to use a subordinate scripting JVM environment. jo...@ve... writes: > "Eric M. Ludlam" <er...@si...> writes: > > > Hi, > > > > I'm not exactly sure how beanshell is used in JDEE, but it may be > > possible to use Semantic as the API, and plug the beanshell into the > > Semantic API. All cases where out-of-emacs storage of data is > > involved, such as source files, compiled files, tag tables, etc, there > > is a way to plug in alternate ways to get that data. Thus, if JDEE > > asks beanshell "give me a list of completions with this prefix", there > > is a Semantic API for that, and a way to plug the beanshell into it. > > This is what I have done, even if its not quite complete. > My next idea was to use rhino instead of beanshell for reflection etc. > Beanshell would still be used as an interactive java shell in jdee. > > One of the benefits of this is that java completion etc will be > available out of the box for all Emacs users in the future. > > > > > I expect it is not quite that simple though. I'm happy to advise > > anyone interested in doing such a port. Alternately, it may be that > > existing Semantic tools which work well in C++ may just work as a > > straight substitute needing only minimal tweaking. It's a matter of > > someone trying it out to know what the scope of the job is. > > > > Eric > > > > > >>>> Paul Landes <la...@ma...> seems to think that: > >>Beanshell is a great rapid prototyping tool and I've used it quiet a > >>bit, but I agree, it looks to be dead. From what I've heard, Jython > >>is now several releases behind Python. I'm starting to see a pattern > >>here. > >> > >>But even with beanshell, you have to recompile the source you're > >>working on to get functionality, that is otherwise, missing. For > >>example, `jde-open-class-at-point' won't work for a class you've just > >>created. You also can't use something like > >>`jde-wiz-extend-abstract-class' either because this functionality > >>comes from output of beanshell using the reflection API. To get this > >>functionality on the classes you've just created (correct me if I'm > >>wrong), you have to compile the source and reload classes in > >>beanshell. Even then, I've come across many ClassLoader issues and > >>you'd have to dump the current beanshell session and restart. There > >>are other problems like: > >> > >>- Time to start the beanshell interpreter > >>- Responsiveness of beanshell. > >>- Beanshell not responding (i.e. no closing tick or quote). > >>- Additional output, like that generated as logging, but interpreted > >> as needed Emacs Lisp output consumed by emacs. > >> > >>Has anyone thought replacing the current that uses beanshell and using > >>semantic for it instead. I'm not sure if this is Joakim's idea or > >>not, but I think having a dependence on any subordinate language for > >>any core functionality of the JDEE isn't a good idea for all the > >>reasons given in the context of Beanshell. I don't want to take it > >>away for "value add" like rapid prototyping and interoperability with > >>other languages as it's nice to be able to write your own utilities in > >>your favorite language. However, it's a maintenance concern. > >> > >>I think it's feasible to use semantic for the parsing and ideas from > >>jde-xref for tying things together. Some IDEs like Eclipse are very > >>good at these kinds of things and implementing some of their ideas in > >>the JDEE might be something we should all think about. > >> > >> > -- > Joakim Verona -- Paul Landes la...@ma... |