[PD-dev] tclpd GUI API thought

[Mathieu Bouchard]

On Sun, 28 Oct 2007, Hans-Christoph Steiner wrote:

> The tclpd API/loader has gotten me thinking about making a pure Tcl API 
> for writing GUIs.  I was thinking a bit about how to handle the Tcl code 
> for the separate pd and pd-gui processes.  Is there a way in Tcl to get 
> the contents of a procedure and stick it into a variable?

You need to redefine proc proc in a way that calls the older proc 
proc and does something else with the arguments. e.g.:

rename proc oldproc
oldproc proc {name arglist body} {
   ...
   oldproc $name $arglist $body
}

> I was thinking that it would be nice if the person writing the GUI 
> object didn't have to think about the separate processes and sys_vgui.

I think it's good to not have to think about encoding and decoding values 
for the other side, which is why I introduced abstraction layers such as 
sys_mgui and pd_upload and proc netsend.

I think that inter-process synchronisation, doing things in the 
right order and at the right moment is such, is a lot harder to do and a 
lot harder to abstract. Basically, you never can fully abstract a remote 
procedure call in a non-blocking way, and you don't want to use the 
blocking way because this is a realtime app.

> Instead, the tclpd API would take the functions that run on pd-gui, and 
> handle sending them over to pd-gui via sys_vgui.  So in the API, it 
> would just look like one thing.

If you just want to upload code over the gui socket, have a wrapped 
sys_gui that you'd use as a block of code:

sys_gui {
   proc blah {} { ... }
   proc stuff {a b} { ... }
   ...
}

and it'll just do what it wants to do.

  _ _ __ ___ _____ ________ _____________ _____________________ ...
| Mathieu Bouchard - tél:+1.514.383.3801, Montréal QC Canada