Rearrange a codebase, metadata and make it abi3 compatible. (#6)

* Initial attempt to make some changes for abi3-compatibility.

* Massive rearrangement.
- Put source files in one package.
- Add an __init__.py and __init__.pyi to provide a target for type stubs.
 - Probably too eager; typeshed will probably provide these until 3.12 is EOL.
- Change setup configuration to
 a. Put as much as possible in pyproject.toml.
 b. Eliminate the need for setup.cfg.
 c. Don't break everything when attempting to build with `build` — the audioop header file just wasn't being seen.
- Most importantly, make it abi3. i.e., the resulting wheel should be compatible for >=3.12 environments.

* Account for some type errors and missing braces for an if clause.

* Slight bit of cleanup in pyproject.toml and setup.py.

* Revert the upgraded requirement to 3.12; it's true and makes it less annoying to publish before 3.13.0.

* Fix license name

Used the wrong one out of habit.

* Don't lie too much for the abi tag.
- More of a footgun than anything.

* change ci to work on PRs

* pin to the 3.13 rc for devcontainer, not a specific patch version

* update project meta, name and python version constraints

* Based on feedback, do restrict install to >=3.13.
- Add a note in setup.py about the limited API compatibility being more permissive but would lead to potential clobbering if requires-python matched it.

* Fix typo in comment and expand slightly.

* Add a py.typed file.

* change ci to work on PRs

* pin to the 3.13 rc for devcontainer, not a specific patch version

* update project meta, name and python version constraints

* Remove setup.cfg again and update project name.

* Ugly, but adding this back for a moment might fix the merge conflict?

* And now delete it again.

* Initial attempt to make some changes for abi3-compatibility.

* Massive rearrangement.
- Put source files in one package.
- Add an __init__.py and __init__.pyi to provide a target for type stubs.
 - Probably too eager; typeshed will probably provide these until 3.12 is EOL.
- Change setup configuration to
 a. Put as much as possible in pyproject.toml.
 b. Eliminate the need for setup.cfg.
 c. Don't break everything when attempting to build with `build` — the audioop header file just wasn't being seen.
- Most importantly, make it abi3. i.e., the resulting wheel should be compatible for >=3.12 environments.

* Account for some type errors and missing braces for an if clause.

* Slight bit of cleanup in pyproject.toml and setup.py.

* Revert the upgraded requirement to 3.12; it's true and makes it less annoying to publish before 3.13.0.

* Fix license name

Used the wrong one out of habit.

* Don't lie too much for the abi tag.
- More of a footgun than anything.

* Based on feedback, do restrict install to >=3.13.
- Add a note in setup.py about the limited API compatibility being more permissive but would lead to potential clobbering if requires-python matched it.

* Fix typo in comment and expand slightly.

* Add a py.typed file.

* Implement the limited api and abi3 tag in a less hacky way.

Source: https://github.com/pypa/wheel/issues/583

* Hardcode the limited api version on the wheel (both the macro and the tag) to be 3.13.
- This should be the final necessary measure to avoid clobbering.

* Remove a now unnecessary comment.

* Improve types beyond what typeshed has.
- Typeshed has argument types as `bytes` to substitute for buffers since buffers didn't have a representation in the type system for a long time. Now that `collections.abc.Buffer` exists, there's a better alternative.
 - This means that memoryview, bytarray, etc. won't show up as type-checker errors when given as inputs.
- Tests actually type-check as a result. No changes to logic were made, just type annotations.
- TODO: Consider upstreaming to typeshed?

Reference material:
- https://docs.python.org/3/library/typing.html#typing.ByteString
- https://docs.python.org/3/library/collections.abc.html#collections.abc.Buffer
- https://peps.python.org/pep-0688
 - "No special meaning for bytes" mentions why these were annotated with `bytes` before.
- https://typing.readthedocs.io/en/latest/source/modernizing.html#typing-bytestring

---------

Co-authored-by: Alex Nørgaard <umbra@abstractumbra.dev>

.gitignore

@@ -4,3 +4,4 @@
 build/
 dist/
 *.egg-info/
+__pycache__/

.vscode/c_cpp_properties.json

@@ -4,7 +4,8 @@
             "name": "Linux",
             "includePath": [
                 "${workspaceFolder}/**",
-                "/usr/local/include/python3.13"
+                "/usr/local/include/python3.13",
+                "/usr/include/python3.13"
             ],
             "defines": [],
             "compilerPath": "/usr/bin/gcc",

audioop/__init__.py

@@ -0,0 +1,59 @@
+from ._audioop import (
+    error,
+    add,
+    adpcm2lin,
+    alaw2lin,
+    avg,
+    avgpp,
+    bias,
+    byteswap,
+    cross,
+    findfactor,
+    findfit,
+    findmax,
+    getsample,
+    lin2adpcm,
+    lin2alaw,
+    lin2lin,
+    lin2ulaw,
+    max,
+    maxpp,
+    minmax,
+    mul,
+    ratecv,
+    reverse,
+    rms,
+    tomono,
+    tostereo,
+    ulaw2lin,
+)
+
+__all__ = (
+    "error",
+    "add",
+    "adpcm2lin",
+    "alaw2lin",
+    "avg",
+    "avgpp",
+    "bias",
+    "byteswap",
+    "cross",
+    "findfactor",
+    "findfit",
+    "findmax",
+    "getsample",
+    "lin2adpcm",
+    "lin2alaw",
+    "lin2lin",
+    "lin2ulaw",
+    "max",
+    "maxpp",
+    "minmax",
+    "mul",
+    "ratecv",
+    "reverse",
+    "rms",
+    "tomono",
+    "tostereo",
+    "ulaw2lin",
+)

audioop/__init__.pyi

@@ -0,0 +1,45 @@
+from collections.abc import Buffer
+from typing import TypeAlias
+
+_AdpcmState: TypeAlias = tuple[int, int]
+_RatecvState: TypeAlias = tuple[int, tuple[tuple[int, int], ...]]
+
+class error(Exception): ...
+
+def add(fragment1: Buffer, fragment2: Buffer, width: int, /) -> bytes: ...
+def adpcm2lin(fragment: Buffer, width: int, state: _AdpcmState | None, /) -> tuple[bytes, _AdpcmState]: ...
+def alaw2lin(fragment: Buffer, width: int, /) -> bytes: ...
+def avg(fragment: Buffer, width: int, /) -> int: ...
+def avgpp(fragment: Buffer, width: int, /) -> int: ...
+def bias(fragment: Buffer, width: int, bias: int, /) -> bytes: ...
+def byteswap(fragment: Buffer, width: int, /) -> bytes: ...
+def cross(fragment: Buffer, width: int, /) -> int: ...
+def findfactor(fragment: Buffer, reference: Buffer, /) -> float: ...
+def findfit(fragment: Buffer, reference: Buffer, /) -> tuple[int, float]: ...
+def findmax(fragment: Buffer, length: int, /) -> int: ...
+def getsample(fragment: Buffer, width: int, index: int, /) -> int: ...
+def lin2adpcm(fragment: Buffer, width: int, state: _AdpcmState | None, /) -> tuple[bytes, _AdpcmState]: ...
+def lin2alaw(fragment: Buffer, width: int, /) -> bytes: ...
+def lin2lin(fragment: Buffer, width: int, newwidth: int, /) -> bytes: ...
+def lin2ulaw(fragment: Buffer, width: int, /) -> bytes: ...
+def max(fragment: Buffer, width: int, /) -> int: ...
+def maxpp(fragment: Buffer, width: int, /) -> int: ...
+def minmax(fragment: Buffer, width: int, /) -> tuple[int, int]: ...
+def mul(fragment: Buffer, width: int, factor: float, /) -> bytes: ...
+def ratecv(
+    fragment: Buffer,
+    width: int,
+    nchannels: int,
+    inrate: int,
+    outrate: int,
+    state: _RatecvState | None,
+    weightA: int = 1,
+    weightB: int = 0,
+    /,
+) -> tuple[bytes, _RatecvState]: ...
+def reverse(fragment: Buffer, width: int, /) -> bytes: ...
+def rms(fragment: Buffer, width: int, /) -> int: ...
+def tomono(fragment: Buffer, width: int, lfactor: float, rfactor: float, /) -> bytes: ...
+def tostereo(fragment: Buffer, width: int, lfactor: float, rfactor: float, /) -> bytes: ...
+def ulaw2lin(fragment: Buffer, width: int, /) -> bytes: ...
+# pyright: ignore [reportShadowedImports]

audioop.c → audioop/_audioop.c

@@ -33,6 +33,13 @@
 
 #include "Python.h"
 
+#define _LTS_Py_SETREF(op, op2)                      \
+    do {                                        \
+        PyObject *_py_tmp = (PyObject *)(op);   \
+        (op) = (op2);                           \
+        Py_DECREF(_py_tmp);                     \
+    } while (0)
+
 static const int maxvals[] = {0, 0x7F, 0x7FFF, 0x7FFFFF, 0x7FFFFFFF};
 /* -1 trick is needed on Windows to support -0x80000000 without a warning */
 static const int minvals[] = {0, -0x80, -0x8000, -0x800000, -0x7FFFFFFF-1};
@@ -1469,7 +1476,7 @@ audioop_ratecv_impl(PyObject *module, Py_buffer *fragment, int width,
                 len = (Py_ssize_t)(ncp - PyBytes_AsString(str));
                 rv = PyBytes_FromStringAndSize
                     (PyBytes_AsString(str), len);
-                Py_SETREF(str, rv);
+                _LTS_Py_SETREF(str, rv);
                 if (str == NULL)
                     goto exit;
                 rv = Py_BuildValue("(O(iO))", str, d, samps);
@@ -1899,7 +1906,7 @@ audioop_adpcm2lin_impl(PyObject *module, Py_buffer *fragment, int width,
     return rv;
 }
 
-#include "clinic/audioop.c.h"
+#include "_audioop.c.h"
 
 static PyMethodDef audioop_methods[] = {
     AUDIOOP_MAX_METHODDEF
@@ -1990,7 +1997,7 @@ static struct PyModuleDef audioopmodule = {
 };
 
 PyMODINIT_FUNC
-PyInit_audioop(void)
+PyInit__audioop(void)
 {
     return PyModuleDef_Init(&audioopmodule);
 }