www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - D, Unit_Threaded, and GtkD

reply Russel Winder <russel winder.org.uk> writes:
Has anyone got any D code using the Glib event loop, usually GtkD code I'd
guess, that is well tested using Unit_Threaded?

--=20
Russel.
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Dr Russel Winder      t: +44 20 7585 2200
41 Buckmaster Road    m: +44 7770 465 077
London SW11 1EN, UK   w: www.russel.org.uk
May 16 2020
parent reply Cogitri <oss cogitri.dev> writes:
On Saturday, 16 May 2020 at 10:51:07 UTC, Russel Winder wrote:

 Has anyone got any D code using the Glib event loop, usually 
 GtkD code I'd guess, that is well tested using Unit_Threaded?


I always had a hard time doing unittests for things with as many 
moving parts as glib based software, so I usually just do 
integration tests like so: 
https://gitlab.alpinelinux.org/Cogitri/apk-polkit/-/blob/1dfbe2b3d959e3c083fcb82419a0a0401c485937/tests/apkd_dbus_server/addAndDelete.d

Maybe I should look into Unit_Threaded for more fine grained 
tests, but I think the effort for all the mocking stuff that I'd 
have to implement even for a (relatively) simple GDBus 
application would be quite substantial.
May 16 2020
next sibling parent reply Russel Winder <russel winder.org.uk> writes:
On Sat, 2020-05-16 at 11:37 +0000, Cogitri via Digitalmars-d-learn wrote:

 On Saturday, 16 May 2020 at 10:51:07 UTC, Russel Winder wrote:

 Has anyone got any D code using the Glib event loop, usually=20
 GtkD code I'd guess, that is well tested using Unit_Threaded?


=20
 I always had a hard time doing unittests for things with as many=20
 moving parts as glib based software, so I usually just do=20
 integration tests like so:=20
 https://gitlab.alpinelinux.org/Cogitri/apk-polkit/-/blob/1dfbe2b3d959e3c0=


83fcb82419a0a0401c485937/tests/apkd_dbus_server/addAndDelete.d

=20
 Maybe I should look into Unit_Threaded for more fine grained=20
 tests, but I think the effort for all the mocking stuff that I'd=20
 have to implement even for a (relatively) simple GDBus=20
 application would be quite substantial.


I am experimenting with using manual control of the Glib event loop using t=
he
pending and iteration methods on the default MainContext within each unit-
threaded test. The alternative of running a GTK application and then puttin=
g
the tests in as an asynchronous sequence only works with Rust and Python si=
nce
D has no coroutines of any sort.

Of course now there is jin.go which is a synchronous multi-tasking approach
with channels rather than an asynchronous approach available in D.=20

--=20
Russel.
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Dr Russel Winder      t: +44 20 7585 2200
41 Buckmaster Road    m: +44 7770 465 077
London SW11 1EN, UK   w: www.russel.org.uk
May 17 2020
parent reply Luis <luis.panadero gmail.com> writes:
On Sunday, 17 May 2020 at 10:19:38 UTC, Russel Winder wrote:

 I am experimenting with using manual control of the Glib event 
 loop using the pending and iteration methods on the default 
 MainContext within each unit- threaded test. The alternative of 
 running a GTK application and then putting the tests in as an 
 asynchronous sequence only works with Rust and Python since D 
 has no coroutines of any sort.


A Fiber can't be used for this ? D Fibers don't have a scheduler. 
Yielding returns to the caller context.
https://dlang.org/phobos/core_thread_fiber.html
May 17 2020
parent Russel Winder <russel winder.org.uk> writes:
On Sun, 2020-05-17 at 20:32 +0000, Luis via Digitalmars-d-learn wrote:

 On Sunday, 17 May 2020 at 10:19:38 UTC, Russel Winder wrote:

 I am experimenting with using manual control of the Glib event=20
 loop using the pending and iteration methods on the default=20
 MainContext within each unit- threaded test. The alternative of=20
 running a GTK application and then putting the tests in as an=20
 asynchronous sequence only works with Rust and Python since D=20
 has no coroutines of any sort.
=20


=20
 A Fiber can't be used for this ? D Fibers don't have a scheduler.=20
 Yielding returns to the caller context.
 https://dlang.org/phobos/core_thread_fiber.html


Futures based co-routines with an executor could do it, but without some fo=
rm
of scheduling, manual control is needed. In fact I think spawning an OS thr=
ead
is probably a better route, the Glib event loop can then run as it wants be=
ing
controlled via events being placed on it's queue from another thread.

--=20
Russel.
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Dr Russel Winder      t: +44 20 7585 2200
41 Buckmaster Road    m: +44 7770 465 077
London SW11 1EN, UK   w: www.russel.org.uk
May 18 2020
prev sibling parent Russel Winder <russel winder.org.uk> writes:
On Sun, 2020-05-17 at 11:19 +0100, Russel Winder wrote:
[=E2=80=A6]

=20
 Of course now there is jin.go which is a synchronous multi-tasking approa=


ch

 with channels rather than an asynchronous approach available in D.=20


Had I checked I would have seen this was a four years ago package that has
been left fallow since. It seems like something that should be resurrected =
and
made a core package for concurrency and parallelism for D.

I suspect there is significant overlap of some task, thread, fibre, and
scheduling code with std.parallelism, but is it worth trying to share code =
ir
just go with separate code?

It might be worth extracting the futures code out of vibe.d so that D has a
futures package in the Dub repository so that people can create asynchronou=
s
coroutines on top of it.
 =20
--=20
Russel.
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
Dr Russel Winder      t: +44 20 7585 2200
41 Buckmaster Road    m: +44 7770 465 077
London SW11 1EN, UK   w: www.russel.org.uk
May 17 2020