On 11/28/12 3:01 AM, "Michael Schmalle" <apa...@teotigraphix.com> wrote:
Very interesting Alex. I like this quote:
"The initial versions will be relatively feature-poor compared to
Flex, but the hope is that the architecture is small enough and
modular enough to allow Apache Flex community members, many who are
participating in their spare time, to participate without having to be
immersed in the internals of the framework."
Seems to be the only way out of the forest.
So since you actually have a working prototype, what do you see the
priorities of action here and the levels of contribution?
My priority, after finishing the wiki writeup, is to get Falcon working
better and figure out how to do MXML to JS. Folks are more than welcome to
help here or try to write new components based on agreed upon goals.
I looked at FalconJS. The problem with me is the lowlevel nature of
the SWF format being involved in the cross compilation.
I'm not sure what you mean by that. I don't recall running into any SWF
related restrictions in getting this demo up. FalconJS does know how to
consume a SWF in order to process SWCs so that it can resolve symbols when
compiling your AS files, but I didn't run into anything else, and I am using
the fact that it does not try to generate JS from the SWCs to my advantage.
All references to symbols in the SWC are to be replaced by JS that fulfills
those API contracts.
I'm not an expert on Falcon or FalconJS, but what I think I see is the
AS/JSCompilation unit creates the AST, then a Reducer gets called to reduce
it to output format. For a SWF, that is an ABCReducer, for FalconJS it is
the JSReducer. The Reducers seem to use Emitters to create
InstructionLists. For a SWF the instruction lists are ABC code, for JS, it
appears that each instruction list is one instruction long but that
instruction is a string of JS. All of those instruction lists are
concatenated then a Writer writes them out. For a SWF it writes ABC blocks
and I think manages constant pools. For JS I have it wired to write out
separate JS files, but there is code in there that tries to aggregate them
and push them through the Closure compiler to get optimized/minified.
In a way I
understand why the engineer chose this route but... are we going to
shot oursevles in the foot because we have "one more layer" that
community members would have to be proficient in to work on it?
I think Falcon and FalconJS are well-written enough that there is separation
of concerns. Folks in the community who don't want to know about the
compiler won't have to know. This will be especially true once I change
Falcon to generate data structures instead of code. Then folks who only want
to write AS code can contribute components or parts of components, and those
of us who can do compiler work can. But my hope is that the compiler work
gets done in a way that the main emphasis is on writing AS code, which we
have more community members who are good at that.
Also, I haven't put much time into this thought but, since I have now
read your "Non-Goals" and "Pay as You Go", I would venture to guess at
least exploring a simple implementation using straight AST/IDefinition
would be worth checking out with FlaconJS.
I'm not sure what that means either, but maybe Gordon can offer an opinion.
My problem is we have Flacon which is a huge parser/compiler, MXML
which is another animal and then FlaconJS which using block code just
makes this project insanely complicated for the casual developer.
If we want to own/control the language we need folks who are not afraid to
touch the compiler. At least with Falcon I'm happy with the basic way the
code works. It has not taken me long to be successful in it, compared to
MXMLC. And it was modular enough to allow FalconJS to happen, and it didn't
take me too long to get a feel for FalconJS.
But I want to hide all of that from the casual developer, who I expect will
contribute on the AS/JS side and not ever look at a java file.
he other problem is I have really no knowledge with the latest and
greatest JS techs, so I am seeing what people are saying to even get a
clue as to what the output would be.
Yup, I have no idea what the best output code is. For now, it is working
and enables me to keep on plugging.
Well, a ramble but my post should just say, Should we try a cross
compilation impl that only uses the IDefinition/AST Nodes API to keep
it simple?
Using the proto FalconJS we are STILL tied to Adobe and the SWF/SWC
spec, to my knowledge that is not considered ActionScript correct?
FalconJS is using SWCs but Apache Flex "own" the code that generates SWCs so
IMO, we are not tied to Adobe. We can change anything we want there as long
as the Flash runtime can understand it, or doesn't need to run it.
Mike
Quoting Alex Harui <aha...@adobe.com>:
Hi,
I finally got permission to check in a demo that uses FalconJS into my
whiteboard at [1].
I started a writeup on it on the wiki that I will try to complete tomorrow
at [2].
[1] http://svn.apache.org/viewvc/incubator/flex/whiteboard/aharui/flexjs
[2]
https://cwiki.apache.org/confluence/display/FLEX/Alex's+FlexJS+Prototype
Thanks,
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
--
Alex Harui
Flex SDK Team
Adobe Systems, Inc.
http://blogs.adobe.com/aharui
