gh-104533: Add a dataclass-like decorator for ctypes structures

[ZeroIntensity]

• Issue: Annotation-based syntax for ctypes structs #104533

[read-the-docs-community]

Documentation build overview

📚 cpython-previews | 🛠️ Build #33605501 | 📁 Comparing bba7682 against main (b4662e8)

  🔍 Preview build  

3 files changed

± library/ctypes.html
± whatsnew/3.16.html
± whatsnew/changelog.html

[johnslavik]

This works for me just fine:

Suggested change

	# These fields don't apply correctly when set later.
	# As a workaround, we have this weird _StructData thing to set the attributes
	# in advance.
	class _StructData:
	pass
	if align is not None:
	_StructData._align_ = align
	if layout is not None:
	_StructData._layout_ = layout
	if pack is not None:
	_StructData._pack_ = pack
	for attr, value in klass.__dict__.items():
	if attr != "__dict__":
	setattr(_StructData, attr, value)
	class _Struct(_StructData, endian_class):
	_fields_ = fields
	_anonymous_ = anonymous
	class _Struct(endian_class):
	vars().update(vars(klass))
	if align is not None:
	_align_ = align
	if layout is not None:
	_layout_ = layout
	if pack is not None:
	_pack_ = pack
	_fields_ = fields
	_anonymous_ = anonymous

but maybe I'm missing sth

[johnslavik]

Consider using lazy-imported functools.wraps for that.
You'd decorate the class with @wraps(klass, updated=()) (updated=() because __dict__ is unwritable).

[johnslavik]

If you decide to apply functools above, you could return a partial() here idiomatically.

[johnslavik]

Curious why not

Suggested change

	if issubclass(klass, (Structure, LittleEndianStructure, BigEndianStructure)):
	if issubclass(klass, Structure):