Return values in Wayland XML specification

[Quentin Glidic]

On 5/15/17 9:46 AM, Pekka Paalanen wrote:
> On Mon, 15 May 2017 07:05:26 +0200
> Wojciech Kluczka <wojciech.kluczka at gmail.com> wrote:
> 
>> In Wayland array can contain only integers.
> 
> Hi,
> 
> not exactly. The data type inside an array is unspecified by the XML
> language. You can use literally anything, it's just a binary blob with
> a size, with some array-like helpers around.
> 
> The helpers are the ones that impose the most restrictions, which is
> that each element in the "array" must be of the same size. But if you
> never use wl_arrat_for_each(), even that is not strictly necessary.
> 
> Of course, one should use some common sense when defining (in
> documentation!) what array contents are. Getting too creative with
> array type arguments is probably not good.
> 
>> See how listing of outputs is done. Server creates one global `wl_output`
>> per output, client binds to the global, server sends events with output
>> parameters (list of modes among them) and then event `done` to indicate it
>> stopped sending info about this output. You probably don't want one global
>> per window. I'd go with one global `window_manager` which sends events (one
>> per window) after bind and after creation of window (or on request,
>> depending on your needs).
> 
> Yes, that is a good design suggestion.
> 
> If the events are always sent only as a response to a request, then
> there is no need for a 'done' event. The client can send the request
> followed by wl_display.sync, and once the sync callback arrives, all
> responses have been received.
> 
> But, if the server can spontaneously send a series of events, then you
> probably need a 'done' for the client to know when the series is
> complete again.
> 
> On Mon, 15 May 2017 04:05:43 +0100
> adlo <adloconwy at gmail.com> wrote:
> 
>> For example, getting a list of windows and returning it as an array.
>> How would this be done? How are arrays represented in Wayland XML?
> 
> The above is one design.
> 
> I would perhaps not use an array to list windows, because when you get
> to around a thousand simultaneous windows maybe, it's possible you
> cannot send the event anymore. There are limits to how big one message
> can be.
> 
> You also need to update the window list over time, which is much better
> done by sending events just for new and removed windows, rather than
> sending the whole list every time.
> 
> Also, if you need to transmit meta-data for each window, it might be
> useful to write a protocol interface for a window, and create a
> protocol object to represent each window. When the meta-data changes,
> the server can send an event on that protocol object rather than a
> tuple { window id, attribute }.

If you are writing/porting a window switcher, please consider using 
Wayland Wall window-switcher protocol[1]. If you are not, please don’t, 
as this protocol is very specific to switchers (it has "switch_to" and 
"close" requests).
If you need for more than this protocol offers, or something doesn’t fit 
your software, just fill an issue and we will discuss the possible designs.

Cheers,


[1] 
<https://github.com/wayland-wall/wayland-wall/tree/master/unstable/window-switcher>

-- 

Quentin “Sardem FF7” Glidic