02:29
jcea has quit [Ping timeout: 250 seconds]
06:11
<
fijal >
ctismer: hi Christian!
06:32
<
ctismer >
fijal: Good morning
06:34
<
ctismer >
cfbolz: yes, I was wondering how to do this at all. It is tightly integrated with CPython and refcounting.
06:37
<
cfbolz >
ctismer: we emulate refcounting quite well for c extensions
06:39
<
ctismer >
Currently, I break things like a Mandelbrot program into Gui and computation. The computation task is PyPy and accelerates by 10, very nice.
06:40
<
cfbolz >
ctismer: did you ever try to build pyside with PyPy and see what fails?
06:42
<
ctismer >
cfbolz: ok, then I should try it. - no, I didn’t. Hey why did I not try!
06:43
<
cfbolz >
There will likely be errors, but I am curious about the first one
06:47
<
ctismer >
Ok, will try to use PyPy for building PySide6 today and see what happens.
06:52
<
ctismer >
PySide with PyPy would really be the burner !
06:56
<
ctismer >
One good thing is that we use the Limited API, already. So the whole macro crap is already solved.
07:09
<
ctismer >
Probably I was blocked by this old article from Victor Stinner:
07:25
<
cfbolz >
ctismer: ah, great!
07:41
oberstet has joined #pypy
11:45
lritter has joined #pypy
11:50
<
ctismer >
cfbolz: After some hassles to let us find the library, it started to compile.
11:51
<
ctismer >
cfbolz: But can it be that PyPy does not support Py_LIMITED_API ?
11:51
<
ctismer >
I get problems with defining it and without.
11:52
<
cfbolz >
ctismer: I don't know, it's a question for mattip, who's on vacation
11:52
simpson has quit [Ping timeout: 258 seconds]
11:53
simpson has joined #pypy
11:54
<
ctismer >
cfbolz: OK, I will hack a bit and try to skip the macros that we both implement ;)
11:54
<
cfbolz >
ctismer: cool :-). maybe we can open an issue to track the problems a bit?
11:56
jcea has joined #pypy
11:56
<
ctismer >
cfbolz: Well, we have already an old one on PySide that I will revive, but probably redirecting to PyPy?
11:56
<
cfbolz >
sounds like a plan!
11:58
<
ctismer >
The limited API is then not THAT problem, since I implemented it for PySide. Good to know!
11:59
<
ctismer >
Are macro defs in PyPython compatible with the C versions?
12:01
<
cfbolz >
mostly they should be
12:03
<
ctismer >
ok then I'll use the limited api, but change the support to undef macros that should not exist.
12:03
<
cfbolz >
ctismer: sometimes we use functions instead of macros though
12:04
<
cfbolz >
so undef might not work
12:05
<
ctismer >
yes, we too! And when you use functions, then it is probably crucial to use them?
12:06
<
ctismer >
Well, I will see, after all they are not so many.
12:10
<
ctismer >
Is there a simple way to see if I'm PyPy? (did not load the sources, yet, but if it's promising...)
12:15
<
cfbolz >
ctismer: from C?
12:16
<
cfbolz >
PYPY_VERSION should be defined
12:17
<
cfbolz >
from python, we import
__pypy__ and check whether it exists
12:17
<
ctismer >
that would be great. No, some PyPy call from CMake would be ok, I already looked into sys.
12:17
<
ctismer >
thakk you, that saves time.
12:24
<
ctismer >
Funny: PyPy does not define PyObject_IS_GC
12:28
<
cfbolz >
ctismer: small stuff like that can just be added, I think (it would always return 1 on pypy?)
12:29
<
ctismer >
yep, I also do not need it any more, just a left-over from Qt 5.15
14:52
<
ctismer >
cfbolz: A few more functions are missing, but it begins to build!
14:52
<
cfbolz >
ctismer: cool, again, we can add them to pypy, too
14:52
<
cfbolz >
ctismer: thanks for working on it :-)
14:54
<
ctismer >
sure! Let me just try if I get far enough and it makes sense. I will open an issue when I know enough.
14:55
<
ctismer >
I would reeeeaaaaaaally love to have PyPy with PySide. That would benefit both projects :)
15:47
<
ctismer >
cfbolz: If it's really only about having everything generate correctly, it might work!
15:48
<
ctismer >
cfbolz: There is no Py_Initialize. Can I ignore this, or am I messing up?
15:48
<
cfbolz >
ctismer: I don't know 😅
15:49
<
cfbolz >
ctismer: If I understand correctly, embedding is not so well supported in PyPy
19:39
lritter has quit [Ping timeout: 265 seconds]
21:13
oberstet has quit [Quit: Leaving]
23:37
Daetalus has joined #pypy
23:43
<
Daetalus >
I had read a twitter said "unicode-utf8 just got merged! Unicode strings are now internally represented as utf-8 in PyPy, with an optional extra index data structure to make indexing O(1). We'll write a blog post about it eventually.
23:43
<
Daetalus >
I didn't find such a blog post. I wonder if there have any resources to describe how unicode-utf8 was implemented in PyPy besides the source code?
23:43
<
Daetalus >
Thank you!