Community Meeting

Summary

On the 28th of January 2010 a community meeting was organised. Haxe developers met on the #haxe IRC channel to discuss the language, the compiler, some notable projects, among others. The meeting was hosted by Niel Drummond.

The following was decided/concluded, in order of appearance:

  • "with" is not going to be implemented at any time. It does not bring any real benefits, but does have severe drawbacks when properties are mixed with locals.
  • Operator overloading could make the code less clear/readable, but considered useful for classes such as Point and Int32. It would allow any operator to be customised, including array access. Whether it will be implemented at some point or not is yet to be determined.
  • Nobody objects to the proposed metadata feature. It could be implemented in a way similar to Java. This feature has to be present for the compiler to support SWC input. It will likely be implemented at some point.
  • The proposed macro feature is well received. This too, will likely be implemented at some point.
  • The "compose" keyword might promote composition and therefore cleaner code. It might replace the friend "pattern" in some cases. Whether it will be implemented at some point or not is yet to be determined.
  • YAML support in the Haxe core libraries is wanted. To accomplish this, someone should write a library for it.
  • The PHP target is used by some. It was developed, and is used, because it is in some situations easier to run on a server than Neko binaries.
  • Some are working on targeting Java and Android. They are running into problems with the Dynamic behaviour in Haxe combined with the strictness of Java.
  • Code conventions haven't been discussed, except for that naming conventions are "for Java developers" according to some.
  • We've agreed that we want to have a license for the Haxe core libraries that requires attribution in the source code, but doesn't require attribution in the binaries. The exact license that we are going to use is yet to be determined.

Log

A log of the talk follows:

[hx]Nicolas Hello, World !
aho bonjour
tonypolinelli zzzzzzzzzzzzz!
TheHippo hello....
heinz hi nicolas
iain hi
brunob hi there
bigarobas hi
Lazygekko afternoon :)
alijaya night :D
alijaya 22.00 here >.<
[hx]Nicolas 10 minutes to go ;)
tonypolinelli 2am here- toughen up alijaya ;P
alijaya haha :p
iain 2am is when programmers are *supposed* to be awake :P
tong_ hi
ciscoheat hello
Juan__ wow, tony , that's haxe commitment : )
mbaczynski hello!
pkplas hi everyone
-- pkplas is now known as zjnue
slaskis hey all :)
tonypolinelli gotta see what you guys have to yabber on about
fponticelli tony is not human ... see his facebook profile and you will unde
rstand ;)
[hx]Nicolas link pliz
[hx]Nicolas :)
tonypolinelli bahah- true ;P
tonypolinelli tonypee
Huge offtopic with some time to kill - anyone interested in the "waxe
" wx/haxe integration project
[hx]Nicolas 4pm here, trying to fix a multithread lock bug in a new Tora fea
ture ^^;
Juan__ i am interested in waxe : )
ciscoheat Huge: Yes, very!
slaskis Huge: sounds really interesting i think
fponticelli indeed it is an interesting project and a tutorial would be grea
t ;)
zjnue waxe looks really cool Hugh
Juan__ actually, added the RSS for the SVN commits to my reader
heinz yep interested
justinfront_ yes will it work on a mac ;
fponticelli Juan__ me too ;)
Huge re tutorial - still in early dev, but following the wx path clos
ely (except use properties and closures for much better reading)
Juan__ also been taking a look to wxwidgets, and surprised they also ta
rget WinCe and some other apart from the big 3
EdoRivai Of all people in this room, pimmhogeling wins it cause he's sexy
[hx]Nicolas wx have been out for a looong time
zjnue more offtopic : anyone using eclihx? the eclipsee4 idea looks go
od and flexsclipse project does something similar in as. i think that was an ide
a (online ide) for haxegui also. gershon here?
Huge yes wx has beed around
Pignoufou (Hello :))
pimmhogeling Hello Pin
pimmhogeling Pignoufou,
Pignoufou zjnue: I bet we could implement a web-based IDE via something li
ke CodeMirror
pimmhogeling Is niel-grumpytoad present, yet?
heinz waxe in combination with nme/neash would be a killer framework ;
), native compiled flash apps on the desktop (cross os)
[hx]Nicolas let's make a quick trpoll : what is you Haxe IDE ? :)
* Aduros wakes up early for haxe
zjnue i'll check it out. the googlecode project i mentioned is actuall
y "flexeclipse"
Huge justin - got mac port going. Seems I have to have neko in a ".a
pp" folder - not sure what's going on there.
slaskis [hx]Nicolas: textmate for me
Pignoufou [hx]Nicolas: TextMate
Aduros [hx]Nicolas: vim
iain geany here
tonypolinelli flashdevelop
justinfront_ textmate
mbaczynski FlashDevelop
ciscoheat FlashDevelop
Huge gvim
fmjrey jedit
zjnue old eclihaxe of illey on eclipse here
fponticelli FlashDevelop
brunob FDevelop
tong_ eclipse
[hx]Nicolas still FD 2.x for me
Juan__ textmate / FD here
bigarobas plashdevelop
Pignoufou Talking about IDEs, I would like to see them all using a build s
ystem based on hxml since this is the official Haxe build system
Juan__ nicolas, is it because 2.x still uses haxe compiler for completi
tion?
fmjrey Actually I beefed jedit with my own plugins to make haxe dev eas
ier
fmjrey se http://sourceforge.net/mailarchive/forum.php?thread_name=2010
01281351.44741.jedit-devel@francois.rey.name&forum_name=jedit-devel
[hx]Nicolas 3.x also have haxe completion
alijaya xcode *with no plugin >.<
[hx]Nicolas but no multi-target project
[hx]Nicolas and I like to compile client+server together
tonypolinelli 3.x has so many great features- i just get a post comman
d to call haxe build.hxml for a proper build
Pignoufou [hx]Nicolas: I've quite surprised to see that FD doesn't use hxm
l files... that makes it impossible to move a textmate project to FD
heinz 3.x svn has compiler completion on demand, which brings better p
erformance on big projects
Pignoufou (and this is although we DO have a standard defined)
[hx]Nicolas Filt3r has been working on FD3 Haxe plugin, I think it works wel
l
fponticelli I found having 2 projects in the same directory is convenient in
FD ... you switch with a doubleclick
[hx]Nicolas yes, I agree that Haxe IDE support should at least enable to gen
erate an HXML
fmjrey re: IDE, I think efforts must go into getting a proper *cross-pl
atform* IDE
justinfront_ i am hoping the sort e-texteditor out for linux, the and
roid bundle really shows what power a plugin could have.
[hx]Nicolas we can't have one single IDE that fits everybody needs, so bette
r have wide support ;)
[hx]Nicolas anyway, we will get started soon I guess
pimmhogeling I guess so, too
pimmhogeling Should niel-grumpytoad start off?
justinfront_ Niel can't make it so he asked me to cover
fmjrey i would point out again the idea of running haxe compiler (ocaml
) in java, see http://haxe.org/forum/thread/648
pimmhogeling justinfront_, will you lead the meeting, then?
justinfront_ Nicolas the creator of Haxe would you like to tell us ab
out your future plans for haxe
[hx]Nicolas ah ;) not just me
[hx]Nicolas but anyway, let's get started !
slaskis yay!
justinfront_ * Haxe compiler: present (2.05) and future !
justinfront_ Nicolas Cannasse, Franco Ponticelli, Hugh Sanderso
tonypolinelli n
[hx]Nicolas first, nice to see you there :)
niel-grumpytoad hello, am back
niel-grumpytoad thanks for coming all
[hx]Nicolas hi niel
niel-grumpytoad the principal compiler writers are there
niel-grumpytoad excellent.. so I guess you can begin talking about the n
ew Haxe features ?
[hx]Nicolas sure !
[hx]Nicolas in case you don't know, we have setup a page on the haxe wiki
[hx]Nicolas http://haxe.org/com/features
[hx]Nicolas this is a place were we gather ideas about future language/compi
ler features
Aduros *cheers for swc input*
[hx]Nicolas it helps us to have some feedback about them BEFORE they are act
ually implemented
Huge I must say that I did not rate MultiTypes, but they would make a
good alternative to overloaded constructors for the wx code i've been doing.
[hx]Nicolas and to have enough time to think about pro/cons of each feature
Huge You will be my hero if you can pull off Removal of Temporary Sta
ck Objects
[hx]Nicolas Also, one of the goals of Haxe is to keep the syntax (almost) as
much simple as it is right now
[hx]Nicolas so we definitly don't want to bloat the language
heinz metadata would be fantastic and opens a lot of possibilities
heinz without changing the language
[hx]Nicolas yes, metadata requires a bit more syntax, but does not get in th
e way with other expressions
[hx]Nicolas explaining everything would take quite a lot of time, so maybe w
e should just go with questions/answers ?
[hx]Nicolas Feel free to ask anything about past, current and future Haxe :)
Huge I am interested in the general feeling about pre-processor - or
has it been ruled out in favour of macros?
pimmhogeling I agree with Huge
Pignoufou Instead of developping the language, which is already quite matu
re, wouldn't it be more interesting to develop frameworks?
iain yes, i'm all for the preprocessor
iain (or an interface for preprocessors)
[hx]Nicolas I'm not a big fan of preprocessors, although I us it a lot in C.
pimmhogeling Could you explain why you're no big fan of them?
[hx]Nicolas The main issue is that everybody does some custom syntax changes
to the language with preprocessors
justinfront_ Pignoufou : agreed I would like flex cross platform alte
rnative but maybe later discussion.
Lazygekko as a total beginner, please could the documentation and examples
be looked at?
slaskis wouldn't the macro feature be some kind of preprocessor?
Pignoufou I have to be against preprocessors, as that makes me feel like t
he toolchain may change depending from people
[hx]Nicolas which makes code sharing a big issue
Aduros No scalable language has a preprocessor...
pimmhogeling True
[hx]Nicolas OCaml has something called Camlp4
[hx]Nicolas which enable you to define new syntax rules on-the-fly
[hx]Nicolas that's pretty powerful but maybe too much
Huge You dis' C/C++ as being non-scalable, but in reality, what is th
e biggest prog that has been written in haxe?
[hx]Nicolas I'm quite conservative about syntax since it's the common langua
ge for all Haxe'rs out there
ciscoheat Sorry if it has been discussed before, but there is no mention o
f coroutines on the feature page?
fmjrey preprocessors would make it harder for ide implementation too!
[hx]Nicolas My proposal wrt preprocessors are "macros" (as described on the
future page)
Pignoufou ciscoheat: I think we discussed "concurrent blocks" months (mayb
e years) ago on the ML, but then it got a bit lost somewhere
[hx]Nicolas it doesn't let you change the syntax
tonypolinelli is there any major benefit in allowing getters/setters t
o be defined as specific functions? If they were standardised to getXXX and setX
XX - would this allow for runtime access?
justinfront_ i think we need to keep examples upto date with new func
tionality, for instance the best explaination of using with Haxe is not on the h
aXe site.
[hx]Nicolas but allow to you to write custom transformations from an express
ion A to an expression B, before typing occurs
justinfront_ flash setters getters need solution agree with tony
Huge My main uses for pre-processor would be to automatically create
the "properties", and have constans for use as default values - but this could b
e fixed in other ways
[hx]Nicolas let's take the "with" example that was discussed on the list
[hx]Nicolas it would be possible to do something like Macro.with(myVar,{ x
= 3; y = 5; }) that would generate { myVar.x = 3; myVar.y = 5; }
Huge the hxcpp backend now has runtime properties - it was only 2 lin
es of code, much easier than I thought it would be. Perhaps also not too hard o
n other targets?
pimmhogeling I have to mention that I'm using Haxe for AVM2 (flash9)
projects a lot, but never run in to problems because of the flash getters and se
tters
EdoRivai I agree with pimmhogeling
justinfront_ well with is not really important syntax. Not like issu
es with porting flash to Haxe and having to use _x rather than x as they have in
away3dlite
[hx]Nicolas it's a bit hard to sort through questions, we really need a cust
om thing for next meeting :)
* Dykam is here
pimmhogeling Yes, we do
niel-grumpytoad maybe it is worth looking at the "with" syntax first
pimmhogeling I'd fly everyone here if I had the money
heinz what information about classes and types is available in a macro
, only context stuff?
ciscoheat Turn on moderation and let someone sort out questions? ;)
EdoRivai I'd fly anymhere if I had the money
niel-grumpytoad is it a feature people generally want, or was it a one-t
ime request from someone ?
[hx]Nicolas heinz > I guess we can make all types informations available in
a macro
heinz so we can possible pass a script to a macro which is compiled to
native code
justinfront_ one-time request from someone ?
fmjrey question: sorry to go back to ide, but i feel the "no ide fits e
veryone's" need is too dismissive of the issue. Couldn't there be ways of having
the haxe compiler more supportive of ide implementation?
slaskis niel-grumpytoad: which feature do you mean?
justinfront_ with
Huge As for with syntax - I would be happy to see it in the language
[hx]Nicolas fmjrey > yes, work on it :)
heinz with is just an example which can be implemented by a macro
niel-grumpytoad well, the one posted on the ML - you have a special symb
ol "with"
iain would macros be able to store global data throughout the compila
tion process?
Huge Yes, but not seriously?
[hx]Nicolas iain > yes
iain good to know :)
tonypolinelli fmjrey > it is already very supportive. We just need som
eone to get excited about eclipse. There is also a monodevelop thread for flashd
evelop which has promise of making flashdevelop crossplatform
waneck On the macro topic, couldn't we extend the macro possibilities t
o include also some kind of operator overloading?
slaskis personally I'd never use a 'with' syntax, never used it while de
veloping flash either
-- R is now known as bigarobas
justinfront_ e-textmate/textmate is also possible cross platform and
bundles can be used better.
[hx]Nicolas huge > as for "with" syntax, I'm not a big fan : it's a pure syn
tax feature which doesn't bring any benefit in itself and has severe drawbacks w
hen properties are mixed with locals. I proposed a "copyFields" on the mailing l
ist which enable to copy all fields of an object to another with structural subt
yping
waneck I've ran into some issues where I could use an array syntax [] o
n some platforms (or compiler directives), but not on others...
fmjrey I would work on it, but would take too much time for me and I ha
ve other priorities. But I do believe someone (e.g. a cs compiler student/resear
cher) could do great work with running haxe compiler in java to get on-the-fly c
ompilation, code completion, error reporting, etc.
pimmhogeling I agree with [hx]Nicolas on this one, the with syntax
fponticelli not a fan either
waneck it would be cool if we could substitute those behaviours- array
access, operators, so some things like Int32 class would be more easier to work
with
slaskis the current work on the textmate bundle is really nice i think,
still a bit buggy with the finding of build.hxml for me but autocompletion works
real good
tonypolinelli i agree - forget 'with' - should we move on?
Huge fair enough - I could take it or leave it.
heinz would it be possible to create e4x like extensions for haxe with
a macro
pimmhogeling We're dismissing with once and for all?
Juan__ with has been dismissed forever : )
[hx]Nicolas waneck > there's already "using", but no "+" overloading yet, in
particular because they would break as soon as Dynamic is used
waneck I see,, but using would break as well..
niel-grumpytoad yes, how do macros work nicolas ?
justinfront_ So can we add copyFields to the language
tonypolinelli (maybe just forget it for this discussion.. its 30mins s
o far ;P )
heinz if you just pass a string which is transformed by the makro func
tion
fponticelli using doesn't work on Dynamic
waneck exactly...
Aduros Is it safe yet to turn on JIT when using Tora?
[hx]Nicolas indeed :)
Aduros Nice!
waneck so... many things on haxe would break on dynamic introduction
waneck e.g. getters and setters themselves
waneck also using...
[hx]Nicolas Aduros > we still had a few crashes in Tora+JIT but that's becau
se some app-specific things, it might work better on your box
waneck we just have to know that and live with that limitation
fponticelli waneck, makes sense ;)
[hx]Nicolas waneck > operator overloading often requires method selection si
nce you might want to implement all the possible cases
[hx]Nicolas but OpOver is not hard to do as it
[hx]Nicolas it's just a question wether we want to have it in the language o
r not
[hx]Nicolas and, apart from Haxe.Int32 and maybe some "Point" stuff, I don't
see much use for them
waneck I think that since Haxe is all about being able to target many p
latforms, this would greatly help to overcome some problems on that process
fponticelli well ... just having it for Int32 would be great ;)
Dykam :D
[hx]Nicolas we could also directly support basic ops for Int32 in the compil
er without OpOver
fponticelli that would be enough for me :D
heinz +1 for that
mbaczynski doing heavy math stuff is a pain without operator overloading, w
ould also speed up quick prototyping
pimmhogeling That would be good, but also classes like Point, as you
mentioned, could benefit from it
Dykam why not immediatly expand it, and just made it convention that i
f you want to support the (+) operator, you have to supply a function OpAdd(..,.
.);
Huge re: neko int32 - perhaps you could promote int32 into syntax, an
d implement an extra vaue type in neko. Infact, I don't think there will be a p
erformance loss at all - maybe you can just upgrade neko to 32 bits for all ints
?
Dykam and throw a error when OpAdd does not exist
waneck Also for a possible Int64, for example
niel-grumpytoad From the "features" page, I prefer "Constraints in metho
d parameters" and the library cleanup
zjnue tentative +1 for opover here still, BigInteger, Int64, more free
doms..
mbaczynski like in the D language :) http://www.digitalmars.com/d/2.0/opera
toroverloading.html
mbaczynski a op b -> a.opFunc(b). like 'using' but more lightweight
Dykam for example
justinfront_ Not following all, but swish has an "add" and "plus" whi
ch you can be more clear than "+", that maybe useful?
pimmhogeling niel-grumpytoad, cut off the discussion if it gets frame
y and lengthy, OK?
waneck this could also mean the possibility of being able to implement
arrayAccess?
[hx]Nicolas yes, if we have operator overloading, it will enable to customiz
e all operators including []
Dykam that last one would be super sweet
waneck this would be very great!
[hx]Nicolas (as long as not dynamic)
justinfront_ less clear though
pimmhogeling I agree, both super sweet and less clear
EdoRivai +1
iain +1 for operator overloading
mbaczynski +1
pedromoraes +1
Dykam [hx]Nicolas, why not dynamic? as every acces just gets compiled
to some conventioned name... every class implementing a func with that name, wil
l work with it
[hx]Nicolas however, it makes the programs less readable
pimmhogeling Yes, less readable/clear
justinfront_ +1 people adding +1 document it :)
Dykam never had any problems with less readable apps btw
[hx]Nicolas Dykam > if dynamic, maybe it's an Int
Dykam ah
Dykam that can indeed be a problem
waneck I think that sometimes it makes the program a lot more readable
mbaczynski what is more readable? a.add(b).sub(c).mul(d) or a + b - c * d
Dykam [hx]Nicolas, on JS it is possible to extend the prototype of an
int
pedromoraes me too, especially in math
waneck a simple example is just anything that uses Int32... I get lost
so easily
Dykam heh mbaczynski, very true
Huge The dynamic issue is the same fot using/get/set - we can deal wi
th it.
Dykam indeed
niel-grumpytoad OK, I think Nicolas can think of a way to make it cleare
r..
Dykam though for get/set it is because you can give the get/set functi
on your own name, and the get/setter can differ per property
justinfront_ this overload sounds more suited to a macro language tha
t you could use with Haxe
[hx]Nicolas Dykam > no, because we have to be xplatform, and most Ints are n
ot objects
Dykam indeed...
Dykam then it would just be the same as using/get/set
Huge yes ints are very efficient on c++ - would be a shame to spoint
that.
Dykam Huge, it is wrapped in dynamic already anyways
Dykam in those cases
Huge That said, it could be done at the compiler level like using, so
it may work anyway
[hx]Nicolas Huge > I agree that Dynamic issue is the same as get/set, but th
at's not a reason to spread it all over. The benefit has to be interesting enoug
h to cover the cost ;)
heinz It would be great to use in combination with makros like the way
i posted it in the newsgroup
heinz maybe
heinz Math.openContext('
heinz and so on
[hx]Nicolas yes indeed
alijaya what is more readable? a.add(b).sub(c).mul(d) or a + b - c * d <
<< i think it must be (a+b-c)*d
[hx]Nicolas that's not the question
[hx]Nicolas there's also some issues with type inference
[hx]Nicolas function add(x,y) return x + y;
[hx]Nicolas will type x and y as Int
Dykam atm, yes
[hx]Nicolas it will not allow to pass instances with have overloaded +
waneck but the + operator is already overloaded - it could be a string
e.g.
Dykam [hx]Nicolas, why not? what is preventing such a compiler feature
[hx]Nicolas think about the following
[hx]Nicolas all operations + - * / does type inference with either Int or Fl
oat
Huge (or string for +)
Dykam [hx]Nicolas, currently yes
[hx]Nicolas yes
Dykam what's preventing changing that, besides backwards incompatibili
ty
[hx]Nicolas so with OpOper it will fail unless you typed your variables
[hx]Nicolas we cannot delay this since it requires generating different code
for the compiler
niel-grumpytoad can we give the Haxe features discussion another 10 minu
tes, and switch to the community libraries...
waneck may be could create an interface like Operatable
[hx]Nicolas maybe let's move on right now ?
[hx]Nicolas other questions ?
pimmhogeling Other language features people are intested in?
danielgrad what about allowing overload only for the array access op?
Huge But what is different: function X<A,B>(a:A,b:B) { return a.some
Func(b); }
heinz i think metadata is really important, is this already planned or
only in discussion
justinfront_ e4x someone posted something on the forum for cross plat
form
Huge sorry - we've moved on
Dykam I agree with heinz
[hx]Nicolas everything on the "features" page http://haxe.org/com/features i
s not decided yet
tonypolinelli swc import would be great - timeline?
pimmhogeling Let's first discuss metadata
[hx]Nicolas swc input actually requires metadata and multiple -swf-lib first
[hx]Nicolas since we need to pass as3-specific stuff (protected/private/name
space/etc) to the backend (with metadata)
justinfront_ can you put layouts in something like XAML or FLEX in me
ta data
[hx]Nicolas clarify ?
heinz i think if metadata is implemented than in the java or c# way (w
ith types as arguments, not only strings as in as3)
Dykam primitive types
[hx]Nicolas but yes, it might be interesting to have all core Haxe values as
part of metadata
[hx]Nicolas code generators might use them to perform some specific tasks
[hx]Nicolas but they will also be accessible at runtime
waneck specially with the @macro feature
heinz only core types or also custom classes as types
zjnue is this where full type info can be read live, for say a full ha
xe interpreter?
[hx]Nicolas java annotations syntax is not bad : http://java.sun.com/j2se/1.
5.0/docs/guide/language/annotations.html
[hx]Nicolas zjnue > clarify ?
pimmhogeling Nobody is against adding metadata to the language, or am
I missing someone?
[hx]Nicolas maybe other questions then ? :)
zjnue reading all types, not only ones defined with rtti. this may all
ow to extend hscript with more lang features and support type checking
[hx]Nicolas Huge for haxe/C++ and Franco for Haxe/PHP are here too ^^;
Dykam pimmhogeling, composed?
Huge ready to answer...
pimmhogeling What I would like to ask is, what's [hx]Nicolas' oppinio
n on adding an access level is general
zjnue something like this was on the TODO on some stage i believe. not
sure if still planned or where it is included in features list
pimmhogeling *in general
fponticelli :)
justinfront_ Is it the wrong time to ask about a better solution to l
oading Haxe flash movies into the same domain and trying to improve bad parts of
flash.
[hx]Nicolas zjnue > rtti takes space in generated program, so still be per-c
lass, but we definitely need to find a better way that right now
waneck maybe we could use metadata to implement some features that coul
d only make a difference in a specific platform - @protected for flash, and mayb
e e.g. @final for cpp
zjnue ah, thank you
[hx]Nicolas pimm > what's "an access level" ? :)
Dykam internal, private, public, etc
Huge No point adding it to cpp - I make eveything public!
pimmhogeling public or private, or protected as we've discussed many
times
waneck no, no.. that's what I meant... a @final keyword would only ake
sense for cpp, to generate not-virtual methods
[hx]Nicolas the philosophy of Haxe is not to protect the developer against i
tself
Dykam pimmhogeling and I talked about "composed" or "composing". prope
rties marked with this level would be accesible by itself, inheritors and the cl
asses having a property of this
[hx]Nicolas so public + private seems enough
Dykam it would enable easier composition
pimmhogeling Right
niel-grumpytoad So are there any questions on desired features/feedback/
bugs in community projects like hxigniter, nme, hippo, hxgui, iphone, et al. ?
slaskis Huge: did you ever get nme/neash to compile for osx 10.6?
Aduros [hx]Nicolas: Would it be possible to open source part of Motion
Twin's spod extensions?
Huge I would also like "public static final CONSTANT = 1" so it can b
e used as a default value
niel-grumpytoad Are there some projects that are still _needed_ ?
waneck also, on that matter, we really need a way to define friendly cl
asses on a better way than with anonymous types
justinfront_ I wanted to ask about Android plans
[hx]Nicolas Huge > what about "public static inline CONSTANT = 1" ?
Huge 10.6 I use the -m32 flag for 32 bits. I don't have 10.6, but ti
should work?
fponticelli niel-grumpytoad, we need a cross platform GUI ;)
Dykam http://as3.pastebin.com/m758cba65 composed
[hx]Nicolas waneck > send a formal proposal an the haxe list
Huge Nicolas > ok, never tried that. What about enums without params
?
Pignoufou Ok, Nicolas, maybe I could talk about Android?
pimmhogeling That's Java-ish, Dykam, but yeah good example
niel-grumpytoad heh.. fponticelli you keep asking me for one!
Dykam oh, oops, indeed
fponticelli forever until done :D
niel-grumpytoad meh.
waneck will do ! : )
Dykam C#ish, latest lang I worked with
slaskis Huge: the problem was with SDL i believe
alijaya test
pimmhogeling I invite you all to look at Dykam's example, and tell me
what you think
Dykam pimmhogeling, http://as3.pastebin.com/m4a539c46
Dykam that one
pimmhogeling It needs "var"
Huge Hmm, don't know. I did fix a crash bug on iphone with sdk v3.
Dykam http://as3.pastebin.com/m223c38fb
Huge I may need to be reminded later about details
[hx]Nicolas not bad, but not good either :) replace "composed" by "private"
and you almost have current "friend" feature ;)
Dykam but friend is AND hackish/hackable, and dynamic
Dykam in the back end
pimmhogeling Friend feels really hack-ish
heinz the problem with current friend feature is that it's slow becaus
e of dynamic i think
slaskis Huge: i believe this was where i hit a wall ;) http://markmail.o
rg/search/?q=haxe+robert+sköld#query:haxe%20robert%20sköld%20from%3A%22Robert%
20Sköld%22%20order%3Adate-backward+page:1+mid:ve2wnwivlhlhhplf+state:results
[hx]Nicolas as for backend, it could be optimized.
Dykam and places the accesibility at the compsing class, not the compo
sed one
pimmhogeling Plus it's less clear, if a class has a private field I d
o not expect it to be accessed, not from a "friend" either
Dykam that is a major thing
stickupkid I'm on 10.6 and -m32 does work for c++
EdoRivai I agree with pimmhogeling
waneck also if you need to refactor something, and you have many typede
fs for friendly types, it's just a pain to need to alter on there also.
slaskis stickupkid: with neash/nme?
[hx]Nicolas pimm > that's not Haxe philosophy : everything should be made ac
cessible if you know what you're doing
pimmhogeling Whereas if a class would have a composed field, or whate
ver, I know it could be accessed at some point
[hx]Nicolas waneck > you can define one single one ;)
pimmhogeling If you know what you're doing, yes
Pignoufou pimmhogeling: If you don't know what you're doing, you have no r
eason to go the hackish way
[hx]Nicolas that's not a feature you can use by accident ;)
waneck I didn't explain right... I actually do, but you always need to
make the changes on both places - the original and the frinedly type definition.
Huge slaskis - looks like may need to take native_midi out to 10.6.
That would be in one of the makefiles. I may upgrade to 10.6 soon
stickupkid yes -- although not tested for iphone, but works on 10.6
[hx]Nicolas yes, that's normal. Using "public" should fix this ;)
[hx]Nicolas let's move on libraries/tools questions maybe ?
Huge I like the package-level access the best
pimmhogeling The thing is, if I'm writing a class with a field that s
hould be accessible by the composer
pimmhogeling Then the developer using that class would have to use a
friend "pattern"
Dykam indeed
tonypolinelli hugh - have you made any more progress on the iphone tar
get?
Huge What is the status of haxelib?
pimmhogeling Which is, as you said, not something you use easily, unl
ess you really know what you're ding
pimmhogeling *doing
EdoRivai I agree
[hx]Nicolas some big changes have been proposed for haxelib
Huge tony - I've got NME 2 drawing an opengl rectangle without SDL. B
ack onto it next week
[hx]Nicolas very good work by blackdog btw
[hx]Nicolas I need to answer about it so we can move on
justinfront_ the compose thing seems to be encourage compose use... b
ut anyway we have moved on
Juan__ Huge > let us know when waxe is ready for some testing
[hx]Nicolas one issue we have to decide on protocols and file formats used b
y haxelib
pimmhogeling OK, let's move on then
[hx]Nicolas currently we have Haxe Remoting (protocol) + XML (haxelib.xml)
Huge Juan, YOu can create a frame and add some controls. There is a
*lot* of functions to implenent - any helpers?
[hx]Nicolas proposal changes by using JSON + HXP
[hx]Nicolas I agree with JSON if this can be wrapped with Haxe remoting
zjnue also, will there ba an official place for libs outside of public
/bsd/lgpl ? like extern ports for say puremvc. where should there go with new ha
xelib options?
tonypolinelli hugh - great, i really hope we can beat cs5 to the app s
tore ;P
Pignoufou [hx]Nicolas: I see no reason why it couldn't. It's just a string
[hx]Nicolas yes
[hx]Nicolas as for HXP, I'm still a bit reluctent to add a new format
[hx]Nicolas maybe we should have something like YAML (or plain JSON) instead
?
[hx]Nicolas haxelib should rely only on standard haxe library because you do
n't want a library to compile the library manager :)
Huge I think JSON is a good fit for haxe
Pignoufou YAML is pretty crappy, being based on tabulations
[hx]Nicolas Pingnoufou > that was an example ;)
[hx]Nicolas any thoughts on this ? are blackdog or John here ?
niel-grumpytoad +1 for JSON, not even sure haxe has a yaml parser..
Pignoufou [hx]Nicolas: How about moving JSON (en|de)coder to the std lib?
[hx]Nicolas Pignoufou > that's the idea
fponticelli YAML is good
Juan__ i like json support on std
[hx]Nicolas anyone to donate a fully tested implementation ? :)
pimmhogeling I am willing to work on that
pimmhogeling And I agree that it should be added to the language's co
re libraries
[hx]Nicolas JS implementation should be easy :)
pimmhogeling Providing that it will only be included in binaries that
use it, obviously
[hx]Nicolas the issue with JSON is that I think it does support Enums, right
?
fponticelli enums?
[hx]Nicolas that's why we came up with haxe.Serializer/Unserializer in the f
irst place
Huge Looking a yaml, seems json:haxe = yaml:python, and special w
hitespace is why I chose haxe over python
niel-grumpytoad well, most languages prefer a human readable-editable co
nfiguration format
[hx]Nicolas Huge > glad you did :)
[hx]Nicolas since neither John or blackdog are here I propose we talk about
it later on the lis
[hx]Nicolas -t
[hx]Nicolas other library/tools-related questions ?
justinfront_ Can I ask about android now?
tonypolinelli poll> who here uses the php target?
ciscoheat Almost to 100% :)
fponticelli not me ;)
Huge nor me
heinz sometimes
jpsecher php +1
stickupkid I'm interested in android, I would like to help where I can
jjdonald me
Huge (no smiley)
danielgrad me too
[hx]Nicolas who's working on android support ?
* zjnue uses php sometimes
pimmhogeling EdoRivai and I use the PHP target
[hx]Nicolas tp> I did run my blog in PHP for some time, but switched back to
Neko
Aduros I have a small site using haxe/php
justinfront_ Nicolas you can compile ocaml on it.
pimmhogeling Our customers often aren't able to setup a neko server
EdoRivai Yup
Huge Has anyone tries the fastcgi demo
Pignoufou I am working on Android [hx]Nicolas
[hx]Nicolas pimm > yes PHP really helps there
zjnue i have, great work Huge. CGI works on DreamHost as-is, but troub
le with FastCGI :(
fponticelli pimm > that was the reason I made it
stickupkid Huge > I couldn't get FastCGI working
[hx]Nicolas Pignoufou > can you tell which point you reached ? do you need h
elp from other people ?
niel-grumpytoad will give the community topic another 10 minutes, then w
e can switch to standards, code conventions and licensing clarifications
pimmhogeling And if so, what kind of help
stickupkid because you can't run bash as fastCGI
Pignoufou Ok so, to talk about Android
Pignoufou (since my time schedule is tight as I'm in course right now...)
stickupkid I'm on dreamhost - take it on the Mailing List?
Huge ok
Pignoufou I'm working with David Elahee from Mad Monkeys Studio on it
Pignoufou My goals were the following : allow full integration with the Ja
va frameworks
stickupkid k, is there any help other than ocaml needed?
Pignoufou Allow use on Android
pimmhogeling OCaml help is perfectly fine, too
[hx]Nicolas Are you compiling Haxe to Java source or bytecode ?
alijaya excuse me, may i suggest a feature? *is it a right time? *kinda
out of topic :p
Pignoufou At the moment, the compiler takes HaXe code and translates it to
Java code
Pignoufou (no Bytecode)
pimmhogeling A language feature, alijaya?
Huge What approach do you use to closures & dynamic?
alijaya i think so
Pignoufou Now, we are facing the "Dynamic problem"
alijaya it's minor, about import thingy
[hx]Nicolas is it written in Ocaml (a new code generator) or as a standalone
tool ?
Pignoufou As most of you certainly know, Java is very very static.
[hx]Nicolas supporting Dynamic on not Dynamic platforms is hard.
Pignoufou The compiler is completely based on the official Haxe compiler,
so it's just another "generator/backend"
[hx]Nicolas k
tonypolinelli hah.. just
Huge c++, everything inherits from the "dynamic interface". So you ca
n do basic ops on all "objects"
Pignoufou So, actually, we basically need people with strong ideas about J
avas inner working, so we can implement dynamics
pimmhogeling alijaya, we're passed language features to be honest
Huge I have some stong ideas on the java typing :)
Pignoufou Even if you don't know OCaml, if you have an idea and can tell u
s "here's an example, here's the code you have to output", we can always get our
hands dirty with OCaml ;) You don't need to
[hx]Nicolas I have some strong ideas about how java typing is bad :)
Pignoufou [hx]Nicolas: I now have the same opinion ;)
alijaya ok maybe later :p
Pignoufou Huge: Unfortunately, your approach won't allow integration with
native frameworks
Huge I think some hacking may be easier at the bytecode level
[hx]Nicolas I guess that you need then to do some wrapping/unwrapping
[hx]Nicolas a Dynamic class with a single Object data field
[hx]Nicolas and an API
Huge The main "work" in the c++ backend is garbage collection, so may
be easier on java
Pignoufou [hx]Nicolas: Yes, I do think so. Anyway, maybe we shouldn't go t
o deep in technical things here? Maybe we should discuss that later on IRC or Sk
ype with people who want to help?
[hx]Nicolas sure
Huge You can also use the native java reflectance to do most things I
think
ciscoheat After this topic I would be interested to know if there are some
thoughts on haxigniter, to know if I should dedicate time on something specific
.
pimmhogeling Pignoufou, if you need another pair of hands to get dirt
y with OCaml, consider mine
Pignoufou Thanks pimmhogeling, I'll be contacting you ASAP (that certainly
means tomorrow ;))
[hx]Nicolas David can come at MotionTwin too to discuss the details directly
;)
pimmhogeling Great, thanks
[hx]Nicolas you can come too but not sure you're located in Bordeaux ;)
justinfront_ Ok is that the only good approach to android?
fponticelli haxigniter is a great project that should deserve more attention
Pignoufou [hx]Nicolas: Would be good to have some days off by Bordeaux, wa
y more sunny than Caen :)
[hx]Nicolas is Andreas here ?
ciscoheat I'm here
fponticelli yeap
Pignoufou justinfront_: I do think so, since Android is a mobile platform,
you have to get "native" for the user experience to be good
[hx]Nicolas want to talk about haxigniter ?
pimmhogeling niel-grumpytoad, what's the scedule
[hx]Nicolas ok now that we have said all theses bad things about FlashDevelo
p, Philippe is joining us :)
ciscoheat I'm just looking for some thoughts, if you've tried it out and f
ound something strange, something that should *really* be documented... things l
ike that. :)
Huge I had a quick look - I will be using it if I ever do a website a
gain. A fastcgi front-end also looks trivial.
Juan__ what's the state of haxeigniter? is it being kept up-to-date wit
h the official branch?
niel-grumpytoad pimmhogeling: we're switching
Juan__ i used it time ago, but finally moved to cakephp
Juan__ i used the php version, i mean
Philippe_FD wow there's a lot of people there :)
ciscoheat It is very up-to-date, I push new features and changes almost ev
ery day.
niel-grumpytoad What Code Conventions are available or desirable for Hax
e ?
pimmhogeling Indeed, do we need code conventions?
niel-grumpytoad what standards are missing but wanted ?
pimmhogeling And if yes, which ones?
Juan__ ciscoheat > good to know
fponticelli when htemplate will become the official haxigniter template syst
em?
[hx]Nicolas andreas > nice
EdoRivai Conventions for class and interface names
niel-grumpytoad are there licensing issues - Haxe has different licenses
on different parts of the code
pimmhogeling Yes
EdoRivai should an interface be named ICar or just Car
Huge Naming convenetions are for java programmers :)
[hx]Nicolas Huge> +1
pimmhogeling Haxe has different licensed, which is not a problem
pimmhogeling Minus one
EdoRivai -1
pimmhogeling :)
jpsecher or CAR
EdoRivai (no smiley)
Skorps about haxigniter, will neko be kept as a target ?
niel-grumpytoad but in reality, dynamic linking between propriety code a
nd the linux kernel also exists, so ... IMO licensing is still a bit too grey in
practice
ciscoheat Skorps: Absolutely
pimmhogeling I disagree that it is too grey
pimmhogeling The thing is, it is not clear which license we want to u
se for libraries
niel-grumpytoad Skorps: I think it will
pimmhogeling [hx]Nicolas chose a license that obligates us all to add
attribution to our binaries, for instance, but I think that's not intentional
[hx]Nicolas should we switch all the standard library from BSD to Public Dom
ain ?
[hx]Nicolas anyone in the compiler team against it ?
Pignoufou pimmhogeling: Why is it so bad anyway? Lots of applications nowa
days depends on LGPL-ed programs for example
pimmhogeling Pignoufou, I'm not saying it's bad
fponticelli not me
Huge There is also the issue of static-linking (iphone) the LGPL code
from the standard libraries I copies from the neko base
pimmhogeling I'm just saying we should think about what we want
[hx]Nicolas Huge > we could relicense Neko under LGPL with linking exception
, that's what OCaml does for instance
pimmhogeling We're all writing libraries, I am, fponticelli is, Dykam
is
Huge Nicolas, I'm a big fan of public domain or mit for my stuff
pimmhogeling Huge, why?
jpsecher the 2-clause BSD licenses says that attribution should be in "in
the documentation and/or other materials provided with the distribution."
pimmhogeling I know
Huge I get satisfaction from when my code is used, not when I obligat
e smeone to do something
[hx]Nicolas Public domain or MIT for stdlib, any preference ?
tonypolinelli i'd say mit - as it is actually more open than public do
main as i understand
pimmhogeling I don't think we should focus on which license we have n
ow, I think we should focus on what we want to achieve
aho i like 0-clause BSD aka "i don't give a f-"
[hx]Nicolas I think we cannot define something for all the libraries on haxe
lib
pimmhogeling tonypolinelli, I disagree, PD is more "open"
Pignoufou Isn't Public Domain risky? Anyone could just ripoff your work an
d sell it without adding any plus-value to it
[hx]Nicolas but we can do a change once for the std library
pimmhogeling Pignoufou, PD can be risky, yes
Huge I have hear there are some issues with pd - IANAL. MIT is prett
y simple
pimmhogeling [hx]Nicolas, yes, and we can tell people that is the pre
fered license for all Haxe libraries
[hx]Nicolas we're talking about std library here, compiler will remain GPL
jpsecher ISC license is even simpler: http://en.wikipedia.org/wiki/ISC_li
cense
pimmhogeling Compiler should remain GPL in my opinion
Juan__ has everybody read the post by grant skinner about licenses?
Juan__ http://www.gskinner.com/blog/archives/2008/07/source_code_lic_1.
html
[hx]Nicolas there's not much use for the standard library without the compil
er ;)
[hx]Nicolas Juan__ > yes, it was quite narrow minded IMHO
Juan__ +1 for a GPL compiler
pimmhogeling There is, as the standard lib/core lib could be compiled
in another compiler
Juan__ it's a start : )
pimmhogeling Yes, keep the compiler GPL
Huge My vote: compiler GPL, libraries mit (you can choose mit from th
e pick box on code.google.com)
Pignoufou pimmhogeling: Not even necessarily in another one
pimmhogeling But the second thing is, the compiler does not get inclu
ded in our binaries
Pignoufou One could sell a package, arguing that he sells the std library,
not the compiler
[hx]Nicolas Huge have a good point :)
Juan__ i actually read this book, which i highly recommend
Juan__ http://oreilly.com/catalog/9780596517960
Juan__ about open source and intellectual property
Pignoufou I do agree that MIT is better
pimmhogeling Huge, you could also pick X11/MIT if the license you're
using is X11/MIT compatible
[hx]Nicolas let's not be paranoid, people can already sell haxe by just talk
ing about it in the documentation as per current license ;)
pimmhogeling Yup
[hx]Nicolas so, if we keep the compiler GPL and move the std library to MIT,
everyone is fine with that ?
stickupkid +1
Philippe_FD who would want to sell haxe?
pimmhogeling Changing to X11/MIT does not really change anything
pedromoraes +1
Huge excellent
pimmhogeling We would still have to attribute in our binaries
[hx]Nicolas Philippe > the bad guys with solar glasses
Philippe_FD +1 for GPL compiler and MIT libraries
pimmhogeling X11/MIT is almost the same as the two-clause BSD license
fponticelli +1
niel-grumpytoad Yes I like that idea
Huge nicolas - and neko too (or can I change the headers of the code
I "borrowed")
[hx]Nicolas really ?
pimmhogeling Yup, sunglassed guys
[hx]Nicolas wait a sec : does MIT requires code attribution as BSD does ?
Juan__ you guys mean Agent Smith is compiled in haxe? :P
pimmhogeling Yes, [hx]Nicolas
Philippe_FD I don't think MIT requires anything like that
niel-grumpytoad (everyone runs off to read MIT license)
pimmhogeling "The above copyright notice and this permission notice s
hall be included in all copies or substantial portions of the Software."
pimmhogeling Doesn't mention anything about binaries not being a port
ion of the software
pimmhogeling We could modify the X11/MIT license not to require it th
ough, if that's what we want
Philippe_FD yes you must keep the mentions if you take "substiantial portion
s" of the code
pimmhogeling That's my big question: what do we want?
Philippe_FD (in the code only)
pimmhogeling Philippe_FD, it doesn't say that, it says "included in a
ll copies"
pimmhogeling A binary is clearly a copy
jpsecher no it is not
Huge Yes, my intentions were keep the header in the code, do what you
want with binary
pimmhogeling Let's just say this: a judge could consider it a copy
[hx]Nicolas according to Wikipedia BSD-2 == MIT
pimmhogeling There's no reason for us to assume (s)he'll not
[hx]Nicolas now, we could do the following
pimmhogeling "Yes, my intentions were keep the header in the code, do
what you want with binary" is that what we want?
pimmhogeling As library writers?
jpsecher yes there is: a lot of projects, companies, etc has incorporated
DSB-like code and are not getting suid by anyone
jpsecher BSD
JohnDG Hi guys, sorry I'm late. Food poisoning.
pimmhogeling jpsecher, yes, most of the time because they comply with
the license
[hx]Nicolas make an official statement as std library writers and state that
will not enforce our rights on the BSD advertising
pimmhogeling If you buy an iPod, the license and copyright notices ar
e in the book you get along with it
Huge Yes, "BSD with attribution exception" would be fine
pimmhogeling [hx]Nicolas, we could also change the BSD/X11 (MIT) lice
nse a bit to say that
pimmhogeling So we're looking at BSD with no attribution in the binar
ies?
[hx]Nicolas I don't want to create another license, please :)
Huge also zlib seems good because they explicitly differentiate betwe
en the final product and ambiguous "software"
Juan__ yes, creating new licenses is usually highly un-recommended
[hx]Nicolas there's already way too much of these
pimmhogeling Adding an exception to a license is not considered writi
ng a new license in most cases
Huge http://www.gzip.org/zlib/zlib_license.html
pimmhogeling Especially if it's an additional permission
niel-grumpytoad JohnDG: I think there were some earlier questions about
haxelib.. it's status and future - maybe you can say something about it
justinfront_ Not really an exception just a clarification on what is
ment by copy? you can just add it underneath?
Huge still awake, tony?
tonypolinelli ...just....
pimmhogeling [hx]Nicolas, if we'd use the MIT/X11 license with an add
itional permission, it would be considered MIT/X11, not a new license
pimmhogeling justinfront_, yes, we could do that as well
pimmhogeling That's just details
[hx]Nicolas tony > licensing wars at 3am, you're not lucky :)
tonypolinelli good point hugh- night all 3:50 actually
pimmhogeling But let's ask this to anyone writing Haxe libraries
JohnDG niel: Blackdog, dg, and myself have devoted some effort to produ
cing a new version of haxelib. It takes haxelib into the direction of a package
manager. It makes it easy to submit libraries, supports custom repos, password r
eminders, clean file format, etc.
pimmhogeling Would such a license suit you guys?
fponticelli pimm, it is ok for me
slaskis JohnDG: awesome!
[hx]Nicolas I want something simple, without exceptions or other things of t
he kind if possible
[hx]Nicolas John > we talked a bit about it before
pimmhogeling I agree
pimmhogeling (with [hx]Nicolas)
pimmhogeling But if that means we'll obligate our users to do things
we don't want to obligate, then we should consider adding an additional permissi
on
pimmhogeling It's pretty common, I must say
pimmhogeling A lot of projects use licenses, but add additional permi
ssions
Huge So maybe zlib or PD
[hx]Nicolas I can understand, but that also makes things more complex than t
hey should. What about PublicDomain ?
niel-grumpytoad JohnDG: 2010-01-28 17:10:01 zjnue also, will there
ba an official place for libs outside of public/bsd/lgpl ? like extern ports fo
r say puremvc. where should there go with new haxelib options?
[hx]Nicolas I'm fine with ZLIB also
pimmhogeling Public Domain is risky, as mentioned
justinfront_ If you send it out as a maillist then any ports can repo
rt if they will have problems switching licences that the only issue maybe
jpsecher AFAIK, there is no such thing as PD in Europe
waneck what's the risk with PD if the libs can only be used along with
the Haxe compiler?
pimmhogeling But now that we know which direction we want to go, we c
ould discuss the details on the mailing list
JohnDG The new haxelib is called "Haxed" and will be released early nex
t week. But something the community needs to decide is whether to continue to us
e the current haxelib, or go with the new haxelib.
[hx]Nicolas right, especially in France since we cannot relinquish our write
rs rights
pimmhogeling Are there any other things on the scedule, niel-grumpyto
ad?
justinfront_ can the libs be automatically transfered.. and a cross o
ver period?
pimmhogeling [hx]Nicolas, right, here we can only release works under
PD in exceptional situations
slaskis JohnDG: i'd say push the new one with the next version of haxe (
2.0.6? or perhaps 2.1?)
niel-grumpytoad pimmhogeling: only final questions... I think we can wra
p things up otherwise
pimmhogeling Code conventions?
justinfront_ john> can the libs be automatically transfered.. and a c
ross over period?
Juan__ anybody fancies about the whole hippo / waxe / air / stuff?
[hx]Nicolas John > my stand was to add haxe Remoting JSON support so we can
keep typed client/server calls
Huge Writing a "SWF widget" from waxe should not be too hard - not su
re about all the platforms, though.
[hx]Nicolas John > and think which file format we can have as a standard for
HXP files
Juan__ huge, how can you think you can go without the player? would you
be able to cover the Flash API with neash + nme?
Juan__ how far, i mean
Huge you should be able to get a game going - not too sure about a fu
ll gui
[hx]Nicolas which are the main missing APIs ?
Huge I think the main issue is the flex mxml stufff e be implementeds
[hx]Nicolas (apart from flex)
Juan__ umm, but that would be out of scope, id say
fponticelli (once more we need a unified cross-platform GUI)
justinfront_ yep
Huge If the scope is keps small, neash should be able to do it
JohnDG Haxe remoting makes it difficult to make JavaScript clients; the
file format is very clean and supports many features the old haxelib does not.
That said, the project is open source and anyone can contribute patches, but by
next week, the project will be sufficiently developed that we think the communit
y will benefit from Haxed in its present form. Of course, people will add more f
eatures with time.
Huge I dont really have the time fore "100% coverage"
zjnue prelim flex support here, maybe code share: http://osflash.org/p
ipermail/osflash_osflash.org/2010-January/017078.html
niel-grumpytoad fponticelli: it's a bit difficult to get the performance
right across platforms
zjnue from lightspark open source flash player
[hx]Nicolas John > the idea is to use the Haxe remoting API with is standard
in Haxe but together with the JSON protocol
Juan__ zjnue > what happened to xinf?
fponticelli niel, it is ... nevertheless you can't ignore the wow factor
JohnDG fponticelli: Isn't there a Flash library emulator for most platf
orms? Stumbled across it on Google code I think.
zjnue don;t ask me, or well. i've been bad, but not skilled for non-fl
ash targets
niel-grumpytoad especially event handling is quite different on each pla
tform , and can cause much memory loss
Juan__ zjnue > just asking : )
zjnue ;)
justinfront_ hsl for events?
[hx]Nicolas Huge > could you maintain something like a compatibility list AP
I per API and method per method so people can have a look at what works very wel
l, partially or not at all ? Will be a good ground for submissions ;)
pimmhogeling HSL is awesome! :)
EdoRivai HSL FTW!!
JohnDG Nicolas, does Haxe remoting support JSON presently?
fponticelli emulator doesn't sound good ... and if you have flex on top of t
he flash api on top of an emulator I am not sure it is a good receipt ... withou
t taking into account that using flex in haxe is still far from being an easy fe
at
pimmhogeling Quick update, draft 3 of HSL is usable, and super stable
(and way better than draft 1 and draft 2)
Huge Nicolas, yes. I've started with flash.graphics mainly, and adde
d events and some other stuff "as needed"
zjnue how similar is HSL to robber penner's signals lib? http://github
.com/robertpenner/as3-signals
fponticelli Huge, such a list would help a lot
pimmhogeling zjnue, it is similar to it, but uses some Haxe-specific
features, and is easier to use
niel-grumpytoad we could try to make hxgui cross-platform using neash
[hx]Nicolas John > no but should be pretty easy to add, you just need to imp
lement haxe.remoting.Connection interface, see how it's done for other protocols
Huge The exact list is something I should work on - but so many ideas
, so little time
zjnue pimmhogeling > great :)
pimmhogeling zjnue, it is, I'm currently writing documentation and ex
amples
justinfront_ neash is inheritance based?
pimmhogeling BTW If anyone has any suggestions for documentation for
and examples of HSL, I'd love to hear it
Huge If you defined a whole new API, you would keep the work limites,
and I would be comfortable supporting the sub-set of calls.
justinfront_ have you put it in that code for me ;)
waneck HSL is really great... I really like the way that all events are
type-checked
[hx]Nicolas indeed, but someone need global vision in order to have a xplatf
orm API that works well
Huge justing - the neash inheritance follows the flash inheritance
heinz @hugh: maybe i can find the time to contribute in a few months,
but currently i'm totally out of time because of my study
justinfront_ yes that was the point I was making... maybe that is not
good for cross platform.
waneck the main problem is that is a little confusing at first - mainly
for the whole signaler part....
justinfront_ i mean neash is maybe nearly there but
waneck i still don't know what it does, jsut use the DirectSignaler all
the time
JohnDG Neash has no JS layer, does it?
pimmhogeling Signaler is an interface, DirectSignaler is an implement
ation of that interface (waneck)
justinfront_ canvas
Huge JS has not had a lot of love lately
Juan__ maybe keeping the as3 API is not the best for cross-platform-nes
s but makes supporting different platforms SO much easier (from a dev point of v
iew)
Philippe_FD ah about JS...
waneck i see, but I don't really see what good would it make to impleme
nt such interface... maybe it could be a little more direct
Huge I'm also re-writing NME to make neash a thinner layer for perfor
mance reasons
[hx]Nicolas good
fponticelli Nicolas, could it be possible to have using methods have precede
nce over private methods?
pimmhogeling waneck, the NullSignaler and the TranslatingSignaler
pimmhogeling I'll show you an example if you want to
waneck pimm > it would be great! I'll help in the documentation, then.
[hx]Nicolas franco > only private is not very easy
pimmhogeling waneck, I have your e-mail address, right?
JohnDG The way to make something like Neash is to first come up with a
low-level "primitives" routines, for example, something like JS Canvas + Sound +
Network. This should be the cross-platform low-level layer. The Flash API is th
en built on top the lower layer.
waneck right!
JohnDG Thus to add support for a new target, one needs only to implemen
t the low-level primitives.
fponticelli Nicolas, having that could make a nice trick with friend types
JohnDG Will cut down on the duplication across the neko/C++ targets.
[hx]Nicolas What about making an OS in Haxe :)
niel-grumpytoad no js-neash had some problematic performance problems, s
o it's unfortunately not being updated
fponticelli HaXeOS, sounds good ;)
pimmhogeling Minus one, lol
zjnue nicolas, oshx.org .com was available last time i checked ;)
Huge JohnDG, that was the separation between neash/NME
[hx]Nicolas franco > I see
Huge NME was a thin layer on SDL. Now I've sort of changed that :)
JohnDG Huge, looking at Neash source code base, I see lots of hard code
d preprocessor directives for the platform.
Huge Mainly because on flash, you don't need to do anything. JS was
a bit different.
JohnDG One would need a canvas abstraction, sound abstraction, network
abstraction, and perhaps timing abstraction, in order to build the Flash API.
JohnDG In fact, this strikes me as a good project for "haxe std lib 2.0
"
Huge THe problems with abstractions are performance and lowest-common
-denominators.
JohnDG Build all the above into the std lib in a cross-platform way. Th
en a community project could build Flash API on top of that.
justinfront_ graphics cross platform what I asked about last time :)
niel-grumpytoad the main problem with JS-NME/neash is the regular timer
that calls ENTER_FRAME - if the garbage collector kicks in you get a nice pause
Huge Seems JS performance has moved quite a bit in the last year
stickupkid V8 seems to be leading this at the moment
Huge and now you can "demo on the iphone" for the wow factor, it may
be worth another think
justinfront_ if you use php can it help the js?
JohnDG JS performance is on par with Flash
Huge The JS flash player that has been getting attention must have do
ne some kind of "jsash" project to get a similat API.
Huge So you could write an indepent flash API in JS, I've got one fo
r neko/c++/iphone and one comes for free in flash, so you would have you cross p
latform graphics api
waneck Yeah, but js works differently from flash... specially if we're
thinking on ie compatibility
Huge I guess I'm saying that the abstractions are already there - jus
t look in the flash doco. You do not need to think about "what messaging should
I use", the decision has already been made.
Philippe_FD IMHO I don't think a Flash API is JS is a great idea - each plat
form needs a specific treatment
waneck maybe we could pull up a html-kind of api (like adobe did with h
xml)...
waneck and like (correct me if i'm wrong) haxegui is making...
Huge well, I think that might be me for the night.....
pedromoraes Huge: if you mean Gordon it doesnt implement a AS interpreter in
javascript yet. the guy plans to do an AS2 vm soon.
justinfront_ flash player 2 or something
pedromoraes I think Philippe is right
Huge If you write in haxe, you don't need such code contortions :)
justinfront_ components structures should be cross platform, implemen
tation can vary
niel-grumpytoad Huge: the js-nme idea could work out a lot better if wor
ker-threads are implemented cross-browser
niel-grumpytoad that way, some of the more resource-intensive ops can be
moved into a separate thread
Huge I guess it's safari we are really thinking about
niel-grumpytoad not sure, need to read up on it a bit better
Huge bye
niel-grumpytoad later
Philippe_FD the discussion seems very *game oriented*
Philippe_FD I have nothing about games of course
iain Nicolas: are there any docs/usage example for format.swf?
[hx]Nicolas are there other questions ?
EdoRivai It was fun listening to you guys, but I'm going.
pimmhogeling I agree with EdoRivai, and I'll go too
[hx]Nicolas Check http://lib.haxe.org/d/format
iain there is no swf tag there?
EdoRivai Guys, check the meeting page on the haxe site. Pimm and I have t
aken notes.
iain i should be able to figure out most of it from the source.. all
i want to do is open an swf, insert some binary data and write it back
heinz are there any plans or ideas about using the same application-do
main (parent domains would be enough) for different haxe-swfs
[hx]Nicolas BTW everybody which have haxelib libraries can submit them with
an haxedoc.xml to have online documentation, see http://haxe.org/com/haxelib for
details
iain oh sorry, didn't click on the 'format' link
[hx]Nicolas heinz > we could add again the "trick" before 2.04
[hx]Nicolas but that's not a serious fix
niel-grumpytoad Think we can close the meeting unless there are some mor
e pressing comments
iain i done some investigation on the 16mb binary limit.. turns out o
caml binary Buffer only supports up to the maximum string length on the platform
(16mb on 32bit)
[hx]Nicolas the issue is that Flash consider two classes with the same name
to be the same :'(
waneck that would be nice!
justinfront_ could you document it and have it as a compiler option?
[hx]Nicolas iain > will try a fix somewhere
[hx]Nicolas justin> yes, fill an issue on googlecode for it
heinz is there also a problem if you create a new application domain w
ith the current as it's parent?
iain thanks, its not majorly important - i just need >16mb temporaril
y so i can try to work around with format.swf
[hx]Nicolas I think the issue is the same
[hx]Nicolas I don't remember exactly what is the difference between the two
[hx]Nicolas iain > swfmill should also be able to do it
heinz and where exactly lies the problem if the boot class gets a rand
om name?
[hx]Nicolas for instance, flash.Lib.current will not be correctly set
justinfront_ Nicholas did you see the cross e4x on the forum
iain i did try adding Rope.Buffer (which 'ties' strings together), an
d to my surprise it compiled with less than 16mb. however greater than 16mb it s
till needed to convert the Rope to a string which failed
waneck Nicolas, any reason why haxe.magic.Proxy got out off future haxe
features?
[hx]Nicolas and Type API will not work in loaded SWF
justinfront_ oleg suggested this for e4x http://code.google.com/p/e4x
u/source/browse/trunk/haxe/src/org/wvxvws/xml/W.hx
[hx]Nicolas waneck > can't remember, maybe na not so good idea ? or hard to
implement
[hx]Nicolas justin>love the package name :)
heinz ok, but if you use the loaded swf as an class-library, and you i
nstantiate these classes from the main-swf that shouldn't be a problem
iain i did use samhaxe to make a separate swf, but of course adding t
his swf to the main haxe-generated swf puts me back to square one (content swf i
s >16mb)
[hx]Nicolas heinz > yes, there are cases in which it works well enough ;)
justinfront_ Nicholas he shows its use here... http://haxe.org/forum/
thread/1144 but I wondered if it could be made cleaner and added to haxe probab
ly not?
waneck Nicolas > ok... I just have to say it was a great idea!!! Specia
lly if one could make a Proxy<T> extends T, implements Proxy, implements Generic
iain heinz: how do i add the content/main swf to the loader swf witho
ut going past 16mb?
heinz so i think the hack should be implemented ;)
[hx]Nicolas I guess you could do something like haxe.xml.Fast which allow fi
lters
Exey It's Oleg's nickname "wvxvws" :)
heinz i don't know, i would load it on runtime
[hx]Nicolas Proxy<T> extends T is not a planned feature ;)
zjnue Nicolas, would it be difficult to expose all Types info on a sta
tic var somewhere, even if it bloats binary size, as a compiler flag option ?
iain yea, looks like it'll have to be runtime
waneck Nicolas > actually in my case it would be for a query system for
a fully Haxe odms
waneck Nicolas > oh, that's what I thought!
waneck I think I'll use the prototype hack on platforms that allow it..
.
[hx]Nicolas anyway, have to go soon :)
justinfront_ Ok well I am going to add this full transcript to the fo
rum.
niel-grumpytoad thanks for making it, hope we can have you here again so
on...
[hx]Nicolas next meeting should be hosted on haxe.org with a nice Haxe-based
chat ;)
niel-grumpytoad or there
heinz thx for your time
waneck yes
waneck thanks!
jjdonald thanks Nicolas
iain thanks everyone
[hx]Nicolas and some tools for more clear questions/answers ;)
zjnue ty
heinz jep, thx everyone
[hx]Nicolas thanks to everyone
ConstNW thanx
Skorps thanks
JohnDG Bye guys, have fun haXing
ciscoheat thank you
mbaczynski thanks to all, very informative!
heinz bye
fponticelli thanks bye
Juan__ cheers people

Hope to see you next time!


version #7973, modified 2010-01-29 00:49:10 by Pimm