Transition from using setup.py to pyproject.toml to specify project metadata

[R-Palazzo]

86azg1hek
Resolve #266

Hi few things to notice/discuss:

• Docker: I changed setup.py to pyproject.toml
• setup.cfg: I was not able to convert all the config inside pyproject.toml, mainly:
  • falke8 config
  • The bumpversion (is it ok because we will change it in Issue #267 no?)
  • aliases (don’t no if we’re still using them)

Let me know your thoughts on this, thanks

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 48.00%. Comparing base (581e1e7) to head (03c2d87).

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #269   +/-   ##
=======================================
  Coverage   48.00%   48.00%           
=======================================
  Files          19       19           
  Lines        1154     1154           
=======================================
  Hits          554      554           
  Misses        600      600           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[gsheni]

• Regarding the flake8 config, we can switch to ruff in different MR and move the config for that into pyproject.toml
• Regarding bumpversion, yes okay for now.
• yes, aliases are not used

[R-Palazzo]

Thanks, @gsheni, great advices! I addressed almost all the comments. I wanted to check with you and @amontanez24 about what to do in the update of task.py

[gsheni]

It might be better to use the Requirement and Specifier class from the packaging library, rather than using regexes. The code might be something like this:

Suggested change

	for line in dependencies:
	line = line.strip()
	if _validate_python_version(line):
	requirement = re.match(r'[^>]*', line).group(0)
	requirement = re.sub(r"""['",]""", '', requirement)
	version = re.search(r'>=?(\d\.?)+\w*', line).group(0)
	if version:
	version = re.sub(r'>=?', '==', version)
	version = re.sub(r"""['",]""", '', version)
	requirement += version
	versions.append(requirement)
	min_requirements = []
	for line in dependencies:
	min_package = find_min_requirement(line)
	min_requirements.append(str(min_package))
	min_requirements = "\n".join(min_requirements) + "\n"
	from packaging.requirements import Requirement
	from packaging.specifiers import Specifier
	def find_min_requirement(requirement):
	if is_requirement_path(requirement):
	# skip requirement paths
	# ex '-r core_requirements.txt'
	return
	requirement = remove_comment(requirement)
	if not verify_python_environment(requirement):
	return
	if ">=" in requirement:
	# minimum version specified (ex - 'package >= 0.0.4')
	package = Requirement(requirement)
	version = find_operator_version(package, ">=")
	minimum = create_strict_min(version)
	elif "==" in requirement:
	# version strictly specified
	package = Requirement(requirement)
	version = find_operator_version(package, "==")
	minimum = create_strict_min(version)
	else:
	# minimum version not specified (ex - 'package < 0.0.4')
	# version not specified (ex - 'package')
	raise ValueError(
	"Operator does not exist or is an invalid operator (such as '<'). Please specify the minimum version ('>=', '==')."
	)
	name = determine_package_name(package)
	min_requirement = Requirement(name + str(minimum))
	return min_requirement
	def is_requirement_path(requirement):
	if ".txt" in requirement and "-r" in requirement:
	return True
	return False
	def remove_comment(requirement):
	if "#" in requirement:
	# remove everything after comment character
	requirement = requirement.split("#")[0]
	return requirement
	def determine_package_name(package):
	name = package.name
	if len(package.extras) > 0:
	name = package.name + "[" + package.extras.pop() + "]"
	return name
	def verify_python_environment(requirement):
	package = Requirement(requirement)
	if not package.marker:
	# no python version specified in requirement
	return True
	elif package.marker and package.marker.evaluate():
	# evaluate --> evaluating the given marker against the current Python process environment
	return True
	return False

[amontanez24]

I think I agree with trying to remove regex if possible. One thing I'm wondering that doesn't really come into play in SDGym but might in other libraries is that sometimes we point to a branch directly to install from. If so, would the minimum tests crash there because that syntax doesn't use '<'? An example would be something like

dependencies = [
  "<packagename>@git+<url-to-repo>#egg=<branch or hash>",
]

[gsheni]

Looks good overall. Just 1 suggestion

[gsheni]

Wrong library. It is this one

Suggested change

	python -m pip install toml
	python -m pip install tomli

[gsheni]

Suggested change

	import toml
	import tomli

[amontanez24]