Skip to content

Instantly share code, notes, and snippets.

@sveetch
Created September 24, 2023 14:56
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save sveetch/deb606c31400fffb30d10a1ef5ed3364 to your computer and use it in GitHub Desktop.
Save sveetch/deb606c31400fffb30d10a1ef5ed3364 to your computer and use it in GitHub Desktop.
Build requirements file content from requirements collected from a package metadata with support of extras (optional requirements).
"""
This requires ``Python>=3.8`` and ``packaging>=23.1``.
"""
from importlib.metadata import requires
from pathlib import Path
from packaging.requirements import Requirement
class RequirementBuilder:
"""
Build requirements file content from requirements collected from a package
metadata with support of extras (optional requirements).
Package metadata format returned by importlib is not very simple to parse so this
implementation is naive and won't support everything that could exists in Python
packages:
* It lacks of support for 'python_version' marker. Requirement defined with this
will fail;
* CVS url or local path for package won't be properly defined in requirement file
with the right Pip option;
Usage of importlib metadata is compatible with almost any package instead of the
TOML way which is more efficient but only available with ``pyproject.toml``
configuration.
.. Hint::
* Requirements file is the standard from Pip, commonly seen in a
``requirements.txt`` file;
* Package metadata is the package configuration collected by setuptools during
package installation. This collection comes from either ``setup.cfg``,
``setup.py`` or ``pyproject.toml`` depending from what package has used to
define its configuration;
* Extras requirements are optional requirements defined in package option
``options.extras_require``;
Keyword Arguments:
blocking (boolean): If False, any exception from parsing will not raise and
instead print the error without blocking further script operations. If
True, any exception from parsing are raised. Default to True.
"""
def __init__(self, blocking=True):
self.blocking = blocking
def print_debug(self, source, r):
"""
Just print Requirement item debug informations.
Arguments:
source (string): Requirement item as collected from ``import.metadata``.
r (packaging.requirements.Requirement): ``Requirement`` object to get its
informations.
"""
print("-", source)
print(" 🎨 {name} - ver:{specifier} - [{extras}] - marker:{marker}".format(
name=r.name,
specifier=r.specifier,
extras=r.extras,
marker=r.marker
))
print(" 🚀 name:", r.name)
print(" 🚀 specifier:", r.specifier)
def parse_metadata_requirements(self, pkgname):
"""
Parse package requirement from importlib metadata on a package.
Arguments:
pkgname (string): The package name to inspect its requirements defined
from its metadata. Obviously the package must be installed in your
environment.
Returns:
dict: A dictionnary of package requirements gathered by their extras name.
For requirement unrelated to any extras, they will be stored in ``None``
item.
"""
requirements = requires(pkgname)
store = {}
extra_pattern = "extra == "
# python_version_pattern = "python_version == " # Unused for now
# Parse all requirement to dress a map index on extra
for item in requirements:
extra_rule = None
# Parse requirement item
r = Requirement(item)
# Normalize version specifier if any
specifier = str(r.specifier) if r.specifier else ""
# As tested with packaging==23.1, extras is always an empty set and extra
# name is located instead in 'marker'
if list(r.extras):
print(self.print_debug(item, r))
msg = (
"Item '{name}' extras is not empty, this is "
"unexpected: {extras_set}"
)
raise NotImplementedError(
msg.format(name=r.name, extras_set=list(r.extras))
)
elif r.marker and " and " in str(r.marker):
print(self.print_debug(item, r))
msg = (
"Item '{name}' have unimplemented marker content: {marker}"
)
raise NotImplementedError(
msg.format(name=r.name, marker=r.marker)
)
elif r.marker and not str(r.marker).startswith(extra_pattern):
print(self.print_debug(item, r))
msg = (
"Item '{name}' have unimplemented marker content: {marker}"
)
raise NotImplementedError(
msg.format(name=r.name, marker=r.marker)
)
# Normalize extra name if any
if r.marker:
extra_rule = str(r.marker)[len(extra_pattern):]
# Remove quotes around name
if extra_rule.startswith("'") or extra_rule.startswith('"'):
extra_rule = extra_rule[1:-1].strip()
if extra_rule not in store:
store[extra_rule] = []
store[extra_rule].append((r.name, specifier))
return store
def create_requirements_file(self, requirements, extras, destination=None):
"""
Create requirements file content and possibly file.
If a same requirement name (version specifier is ignored) exists in multiple
extras, only the first occurence is listed.
Arguments:
requirements (dict): Dict of requirements as returned from
``parse_metadata_requirements``.
extras (string or list): List of extras names to select requirements.
Use ``None`` value as name to get non optional requirements.
Instead of list, you can just give the string name ``all`` which will
collect every requirements, both non optional and optional requirements.
Keyword Arguments:
destination (pathlib.Path): File path where to write requirement file if
given.
Returns:
string: Requirement file content.
"""
lines = []
collected = set()
if extras == "all":
extras = [None] + list(requirements.keys())
for extra in extras:
if extra in requirements:
# Search in extra section
for item in requirements[extra]:
# Don't collect twice the same requirement depending its name.
if item[0] not in collected:
lines.append("".join(item))
collected.add(item[0])
content = "\n".join(lines)
if destination:
destination.write_text(content)
return content
def get_requirements(self, pkgname, extras, destination=None):
"""
Shortand to inspect, return requirements file content and possibly write it
on filesystem.
Arguments:
pkgname (string): The package name to inspect its requirements defined
from its metadata. Obviously the package must be installed in your
environment.
extras (string or list): List of extras names to select requirements.
Use ``None`` value as name to get non optional requirements.
Instead of list, you can just give the string name ``all`` which will
collect every requirements, both non optional and optional requirements.
Keyword Arguments:
destination (pathlib.Path): File path where to write requirement file if
given.
Returns:
string: Requirement file content if there was no parsing error else it
will return None.
"""
try:
pkg_requirements = self.parse_metadata_requirements(pkgname)
except NotImplementedError as e:
if not self.blocking:
created = None
print(e)
else:
raise e
else:
created = self.create_requirements_file(
pkg_requirements,
extras=extras,
destination=destination,
)
return created
if __name__ == "__main__":
builder = RequirementBuilder()
destination = Path("./docs/requirements.txt")
content = builder.get_requirements(
"py-css-styleguide",
["doc", "django"],
destination=destination,
)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment