On Tue, 2009-09-15 at 22:50 +0200, Jan Hudec wrote:
> On Tue, Sep 15, 2009 at 20:33:03 +0200, Jan-Jaap van der Geer wrote:
> > On Tue, 2009-09-15 at 09:06 +0200, Jan Hudec wrote:
> > > On Tue, September 15, 2009 00:09, Jan-Jaap van der Geer wrote:
> > > wimp_menu_entry_new(int n, wimp_menu_entry[] entries);
> > > ...
> > > [Compact]
> > > public class Menu {
> > > ...
> > > public Menu([CCode(array_length_pos = 0.9)] MenuEntry[] entries);
> > > ...
> >
> > What does the array_length_pos = 0.9 mean?
> Notice, that I wrote the C function with two arguments, but the vala with
> only one. Arrays are normally passed as two arguments -- the pointer and the
> size. And this attribute specifies at which position the additional argument
> (size) goes.
> When generating the argument list, vala compiler assigns each argument
> a number. The declared arguments are number by whole numbers starting from
> one, invocant (this) in method gets 0 and implicit parameters get just
> a little bit higher number than the argument they accompany. Vala than emits
> the arguments in order of that number.
> So by specifying 0.9 I am saying the length of the array should go just
> before the array pointer (which is argument 1.0), not after.
Ah, right, I understand now. Thanks!
> ... which really changes it to the next option. It depends on how convenient
> you want to make the API in vala.
> If you only specify length to the constructor, you will have to assign each
> entry separately, so you will end up with a chain like
> menu.entries[0] = MenuEntry() { ... };
> menu.entries[1] = MenuEntry() { ... };
> No other option would work, because the other array operations (like
> assignment of whole array) depend on vala being able to reallocate the array
> and to know it's length, neither of which is allowed by the underlying data
> structure.
> On the other hand if you specify the whole content, you will be able to
> construct with single big constructor expression. Along the lines of:
> menu = new Menu(320, 240, new MenuEntry[] {
> { ... }, { ... }, ... });
> (I've added the dimensions here -- you would want to specify all member
> values as constructor arguments if you specify any for the API to make sense)
Yes, I see your point. I'll have to think about this one. The .vapi
files I have are all generated from files that describe the API using a
tool that generates assembly headers, C headers, C++ headers and
documentation files. I enhanced this system to include generation of
vapi files. So the above would mean I need to generate these
constructors and I am not really sure how practical that would be.
> > > > Related: Is there documentation about this? If so where can I find
> > > > that?
> > > Very scarce. Look through the wiki, but mostly you will have to
> > > look how similar things are done in the existing .vapi files.
Stupid question: What wiki are you talking about? I know about this
page, of course: http://live.gnome.org/Vala It looks somewhat wiki-ish,
but I'm not sure you're referring to that... Google didn't help me much.
BTW, on that page there is a link to "Hacker's Guide to Vala" which went
dead a few days ago.
> > I find it very difficult to look at other .vapi files as I have almost
> > no experience with GLIb and other libraries, coming from RISC OS where
> > these libraries are not generally used, or at the very least I am not
> > very familiar with them. So it is difficult for me to find relevant
> > comparable situations and their solutions in the .vapi files.
> If you know at least the standard POSIX API, you can focus on the posix.vapi.
> It does not use anything of glib. It is actually possible to use vala
> without glib altogether (using --profile=POSIX parameter), though many
> features are not available than.
Not even that is familiar at this point :-| But the --profile=POSIX
intrigues me. What kind of things would not be available using this
POSIX profile?
Cheers,
Jan-Jaap
_______________________________________________
Vala-list mailing list
[email protected]
http://mail.gnome.org/mailman/listinfo/vala-list
