Implement type stubs for ecodes

[Lillecarl]

Helps your $EDITOR find the constants from ecodes.c, making for a more pleasant developer experience.

I've got a "grand plan" to extend evdevremapkeys to support importing your own Python modules, and when building those modules it would be very nice to get autocomplete and type checking for ecodes.KEY_A and such.

I'm relatively new to Python and haven't messed with setuptools before at all, I've hacked together the install class to copy the pyi file with the wheel but I'm eager to know of a better way (MANIFEST.in doesn't seem to work because the file doesn't exist when setuptools starts).

[sezanzeb]

Hey, sorry I haven't spent time on this yet. I hope I'll manage to look into this soon.

[sezanzeb]

Sorry for the delay, thanks for waiting.

Hm, makes me wonder why setup.py is generating ecodes.c, and not ecodes.py.

Anyhow, your change works for me 👍.

A few questions/discussions, and a suggestion for a refactoring of genpyi.py.

Turns out that pycharm also generates stubs by the way, from /usr/lib/python3/dist-packages/evdev/_ecodes.cpython-312-x86_64-linux-gnu.so, putting them into ~/.cache/JetBrains/PyCharmCE2023.2/python_stubs/-32419094/evdev/_ecodes.py.

[sezanzeb]

where does the "src" environment variable come from? Is it something from when you were developing the change for debugging or something?

[sezanzeb]

installdir should probably be os.path.join(self.build_lib, 'evdev')

[sezanzeb]

I haven't tested my suggested code yet. Anyhow

• I think it's good practice to avoid code that gets executed when importing a file. So this should be warpped in if __name__ == '__main__':. (I think the same way about genecodes.py, but refactoring this now is not important)
• Also, I split this into a few separate specialized functions. I think that makes it a bit more readable.

I wouldn't want to bother you with somewhat opinionated review comments, so I just made the refactor myself:

Suggested change

	#! /usr/bin/env python3
	"""
	Generate a Python extension module with the constants defined in linux/input.h.
	"""
	from __future__ import print_function
	import os, sys, re
	# -----------------------------------------------------------------------------
	# The default header file locations to try.
	headers = [
	"/usr/include/linux/input.h",
	"/usr/include/linux/input-event-codes.h",
	"/usr/include/linux/uinput.h",
	]
	if sys.argv[1:]:
	headers = sys.argv[1:]
	uname = list(os.uname())
	del uname[1]
	uname = " ".join(uname)
	print(f"# used_linux_headers: {headers}")
	# -----------------------------------------------------------------------------
	macro_regex = r"#define +((?:KEY|ABS|REL|SW|MSC|LED|BTN|REP|SND|ID|EV|BUS|SYN|FF|UI_FF|INPUT_PROP)_\w+)"
	macro_regex = re.compile(macro_regex)
	# -----------------------------------------------------------------------------
	template = rf"""
	# Automatically generated by evdev.genecodes
	# Generated on {uname}
	# Headers: {headers}
	"""
	def parse_header(header):
	lines = ""
	for line in open(header):
	macro = macro_regex.search(line)
	if macro:
	#lines += f" {macro.group(1)}: int{os.linesep}"
	lines += f"{macro.group(1)}: int{os.linesep}"
	return lines
	for header in headers:
	try:
	fh = open(header)
	except (IOError, OSError):
	print(f"Unable to read header file: {header}")
	continue
	template += f"{parse_header(header)}"
	print(template)
	#! /usr/bin/env python3
	"""
	Generate a Python extension module with the constants defined in linux/input.h.
	"""
	from __future__ import print_function
	import os
	import sys
	import re
	from typing import List
	def get_headers() -> List[str]:
	# The default header file locations to try.
	header_files = [
	"/usr/include/linux/input.h",
	"/usr/include/linux/input-event-codes.h",
	"/usr/include/linux/uinput.h",
	]
	if sys.argv[1:]:
	header_files = sys.argv[1:]
	print(f"# used_linux_headers: {header_files}")
	return header_files
	def get_uname() -> str:
	uname = list(os.uname())
	del uname[1]
	return " ".join(uname)
	def parse_header(header_file: str) -> str:
	macro_regex = r"#define +((?:KEY|ABS|REL|SW|MSC|LED|BTN|REP|SND|ID|EV|BUS|SYN|FF|UI_FF|INPUT_PROP)_\w+)"
	macro_regex = re.compile(macro_regex)
	lines = ""
	for line in open(header_file):
	macro = macro_regex.search(line)
	if macro:
	# lines += f" {macro.group(1)}: int{os.linesep}"
	lines += f"{macro.group(1)}: int{os.linesep}"
	return lines
	def generate_template(uname: str, headers: List[str]) -> str:
	template = rf"""
	# Automatically generated by evdev.genecodes
	# Generated on {uname}
	# Headers: {headers}
	"""
	for header in headers:
	try:
	fh = open(header)
	except (IOError, OSError):
	print(f"Unable to read header file: {header}")
	continue
	template += f"{parse_header(header)}"
	return template
	if __name__ == '__main__':
	template = generate_template(
	get_uname(),
	get_headers()
	)
	print(template)

[sezanzeb]

I wonder why this was done using the subprocess module. Wouldn't just calling a method that does this be much cleaner?

For example in the case of the pyi stub, what would be a reason to not call generate_template and write the result into a file, without spawning a subprocess?

Anyhow, I'm fine with doing it as a subprocess since it has already been working like that in the past.

[sezanzeb]

The generation of ecodes.c doesn't seem to require such a step, why does ecodes.pyi?

[sezanzeb]

ecodes.pyi should be in the .gitignore file, just like ecodes.c

[Lillecarl]

@sezanzeb Don't worry, I've been on the lazy side "projectifying" and my thing as well. I could've refined this further during this time.

I'll get back to you 😁