Denis Laxalde
Denis Laxalde
I have run [MagickStack's benchmark](https://github.com/MagicStack/pgbench) against 3.0.3 and this branch. The source code of the benchmark tool handling psycopg 3 is available at https://github.com/dlax/pgbench (branch `psycopg`). Running `./pgbench python-psycopg-async`, on...
> Why is your branch using internal functions instead of the public interface? You mean `psycopg.pq`? Yes I added benchmark runners implemented in this low-level API (benchmark names: `python-pq` and...
> Curious results, no? Do tests hit different paths, so that anyio is faster on the first two and slower on the other ones? Or am I reading something wrong?...
Reworked this to make the dependency on anyio optional. (Still needs some work on copy.)
`tests/types/test_net.py::test_lazy_load` fails in many cases in CI; not sure how it's related.
I see, makes sense; thanks. I'll see how to adjust this test then.
@Kludex, I'm afraid I don't quite get your point; would you mind reformulating or expanding a bit?
So I have spent a significant amount of time on this work of using anyio as an alternative to asyncio in psycopg (core only, not speaking about the pool). For...
I moved on with using anyio instead pytest-asyncio in the test suite, which indeed really helps for adding support for anyio in psycopg; this is done in the first commit...
I don't understand the test failure on Windows as it complains about the loop policy being incorrect (Proactor), *but only upon retry*.