Comments on: Closures and asynchronous methods in Vala

By: Norman

Norman — Sat, 19 Sep 2009 13:52:47 +0000

I install monodevelop with vala plugin.

I create a new vala project and i can’t run a basic example of Gtk Window :
http://live.gnome.org/Vala/GTKSample

By: Amos Wenger

Amos Wenger — Fri, 18 Sep 2009 13:55:55 +0000

Hey Vala team, congrats here from the ooc team =)

What you implemented looks very much like what I had in mind for ooc closures. Mind to share some code/structs? (Or maybe you’re not comfortable with getting intimate that fast ^^) Interoperability might be nice, since we’re both C-generating languages.

Anyway, kudos, and keep up the great work =)

By: Jeff Waugh

Jeff Waugh — Fri, 18 Sep 2009 12:51:36 +0000

Yay, glad to see someone making excellent use of the syntax highlighting plugin on GNOME Blogs.

By: Tweets that mention Closures and asynchronous methods in Vala « Jürg Billeter’s blog -- Topsy.com

Fri, 18 Sep 2009 11:35:04 +0000

[...] This post was mentioned on Twitter by Programming Joy and Tech news. Tech news said: Reddit/p: Closures and asynchronous methods in Vala http://bit.ly/153FLR [...]

By: juergbi

juergbi — Fri, 18 Sep 2009 11:00:38 +0000

@Matt: As Neil correctly explains, the variable is kept alive as long as it’s referenced in a closure. The mentioned bug has been fixed now and your example will print 10 as expected.

@James: It is not based on generators as this would not allow the same level of integration with languages that are not as dynamic as, e.g., Python or JavaScript. However, support for generators is planned as well, and a large part of the implementation will be shared.

By: Cliff Brake

Cliff Brake — Fri, 18 Sep 2009 10:12:18 +0000

thanks for the examples

By: Nicolas Trangez

Nicolas Trangez — Fri, 18 Sep 2009 09:23:02 +0000

This is some insanely cool and impressive stuff. Well done Vala team!

@Yann: AFAIK it will not spin, the generated code is all async callback based, so no CPU is being used as long as no new data is available, and no threads involved (next to the main thread, obviously) either.

By: Neil Roberts

Neil Roberts — Fri, 18 Sep 2009 09:22:49 +0000

@Matt: I was going to post a similar comment. It looks like it stores the variables referenced in the closure to a ref-counted struct and each delegate variable now makes two C variables – one for the function pointer and one for the data pointer. So the local ‘i’ variable referenced in your closure should outlive the scope of ‘make_func’. However it looks like there’s some bugs there because the generated code does not take a reference to the struct when make_func returns so the data gets destroyed prematurely and your programs outputs “i is 0″!

By: James Henstridge

James Henstridge — Fri, 18 Sep 2009 09:16:35 +0000

@Yann: no. The idea of this kind of syntax is to let you write functions that look synchronous but run asynchronously. Each of the yield calls indicates a point where the function will pause, waiting for completion of that operation.

Internally, this would probably be implemented by storing the locals for the function inside a structure so that they can be maintained over multiple invocations, some glue to wait on the GAsyncResults and code to handle reentering the code path (possibly using Simon Tatham’s switch statement method).

By: Yann

Yann — Fri, 18 Sep 2009 06:43:21 +0000

I am just wondering:
does “while (true)” really iterates until we get a callback?
(and takes 100% of CPU)

or is it waiting for the callback? (and the thread is sleeping)