12:48 <DuToit> Hey everyone, I checked the amount of Pillow functions using tuples as arguments, as these are not currently supported. There's only 11 of them, but some important methods like filtering and editing individual pixels use tuples

12:54 <steve_s> Do you mean that parse tuples using HPyArg_Parse? TBH it would seem reasonable to me to support that for use cases like when you want to "destruct" a tuple that you know/expect to contain a pair of integers. Doing that manually by parsing it as "O" and then extracting and converting the individual elements means quite some boiler-plate code and also if HPyArg_Parse was/could optionally be a context function, then lot more context calls

12:56 <steve_s> Relevant discussion https://github.com/hpyproject/hpy/pull/284

17:45 <mattip> here's another project that might be a candidate for an HPy version: Anthony Shaw's picologging

17:45 <mattip> https://github.com/microsoft/picologging

17:47 <mattip> it comes with benchmarks, and is "4x to 10x faster than the stdlib logging module", uses the C-API directly

19:46 <antocuni> mattip: it sounds like a perfect candidate, indeed

19:46 <antocuni> I did a quick grep to check which APIs it uses: https://paste.openstack.org/show/bjWax1QRXEJXyxm7ZMz1/

19:47 <mattip> I came across it because it doesn't work with PyPy since PyPy is missing at least PyObject_CallMethod_ONEARG

19:49 <antocuni> it's also a good benchmark for pypy

19:50 <antocuni> it would be interesting to compare pure python logging vs cpyext picologging vs hpy picologging