From 968da4d331d0838ddbb7401592f35c276d4bb9fb Mon Sep 17 00:00:00 2001 From: Christian Clauss Date: Fri, 20 Oct 2023 15:44:20 +0200 Subject: [PATCH] chore (main): vendor in Python packaging for Py3.12 --- .github/workflows/node-gyp.yml | 6 +- .github/workflows/nodejs-windows.yml | 4 +- .github/workflows/python_tests.yml | 2 +- pylib/packaging/__init__.py | 15 + pylib/packaging/_elffile.py | 108 +++ pylib/packaging/_manylinux.py | 252 +++++++ pylib/packaging/_musllinux.py | 83 +++ pylib/packaging/_parser.py | 359 +++++++++ pylib/packaging/_structures.py | 61 ++ pylib/packaging/_tokenizer.py | 192 +++++ pylib/packaging/markers.py | 252 +++++++ pylib/packaging/metadata.py | 825 +++++++++++++++++++++ pylib/packaging/py.typed | 0 pylib/packaging/requirements.py | 90 +++ pylib/packaging/specifiers.py | 1030 ++++++++++++++++++++++++++ pylib/packaging/tags.py | 553 ++++++++++++++ pylib/packaging/utils.py | 172 +++++ pylib/packaging/version.py | 563 ++++++++++++++ 18 files changed, 4561 insertions(+), 6 deletions(-) create mode 100644 pylib/packaging/__init__.py create mode 100644 pylib/packaging/_elffile.py create mode 100644 pylib/packaging/_manylinux.py create mode 100644 pylib/packaging/_musllinux.py create mode 100644 pylib/packaging/_parser.py create mode 100644 pylib/packaging/_structures.py create mode 100644 pylib/packaging/_tokenizer.py create mode 100644 pylib/packaging/markers.py create mode 100644 pylib/packaging/metadata.py create mode 100644 pylib/packaging/py.typed create mode 100644 pylib/packaging/requirements.py create mode 100644 pylib/packaging/specifiers.py create mode 100644 pylib/packaging/tags.py create mode 100644 pylib/packaging/utils.py create mode 100644 pylib/packaging/version.py diff --git a/.github/workflows/node-gyp.yml b/.github/workflows/node-gyp.yml index edf68773..ebe74975 100644 --- a/.github/workflows/node-gyp.yml +++ b/.github/workflows/node-gyp.yml @@ -11,16 +11,16 @@ jobs: fail-fast: false matrix: os: [macos-latest, ubuntu-latest, windows-latest] - python: ["3.8", "3.x", "3.12"] + python: ["3.8", "3.10", "3.12"] runs-on: ${{ matrix.os }} steps: - name: Clone gyp-next - uses: actions/checkout@v3 + uses: actions/checkout@v4 with: path: gyp-next - name: Clone nodejs/node-gyp - uses: actions/checkout@v3 + uses: actions/checkout@v4 with: repository: nodejs/node-gyp path: node-gyp diff --git a/.github/workflows/nodejs-windows.yml b/.github/workflows/nodejs-windows.yml index bfbfb9f1..3f52ff9c 100644 --- a/.github/workflows/nodejs-windows.yml +++ b/.github/workflows/nodejs-windows.yml @@ -12,11 +12,11 @@ jobs: runs-on: windows-latest steps: - name: Clone gyp-next - uses: actions/checkout@v3 + uses: actions/checkout@v4 with: path: gyp-next - name: Clone nodejs/node - uses: actions/checkout@v3 + uses: actions/checkout@v4 with: repository: nodejs/node path: node diff --git a/.github/workflows/python_tests.yml b/.github/workflows/python_tests.yml index 87e74d73..049d5fe5 100644 --- a/.github/workflows/python_tests.yml +++ b/.github/workflows/python_tests.yml @@ -18,7 +18,7 @@ jobs: os: [macos-latest, ubuntu-latest] # , windows-latest] python-version: ["3.8", "3.9", "3.10", "3.11", "3.12"] steps: - - uses: actions/checkout@v3 + - uses: actions/checkout@v4 - name: Set up Python ${{ matrix.python-version }} uses: actions/setup-python@v4 with: diff --git a/pylib/packaging/__init__.py b/pylib/packaging/__init__.py new file mode 100644 index 00000000..5fd91838 --- /dev/null +++ b/pylib/packaging/__init__.py @@ -0,0 +1,15 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + +__title__ = "packaging" +__summary__ = "Core utilities for Python packages" +__uri__ = "https://github.com/pypa/packaging" + +__version__ = "23.3.dev0" + +__author__ = "Donald Stufft and individual contributors" +__email__ = "donald@stufft.io" + +__license__ = "BSD-2-Clause or Apache-2.0" +__copyright__ = "2014 %s" % __author__ diff --git a/pylib/packaging/_elffile.py b/pylib/packaging/_elffile.py new file mode 100644 index 00000000..6fb19b30 --- /dev/null +++ b/pylib/packaging/_elffile.py @@ -0,0 +1,108 @@ +""" +ELF file parser. + +This provides a class ``ELFFile`` that parses an ELF executable in a similar +interface to ``ZipFile``. Only the read interface is implemented. + +Based on: https://gist.github.com/lyssdod/f51579ae8d93c8657a5564aefc2ffbca +ELF header: https://refspecs.linuxfoundation.org/elf/gabi4+/ch4.eheader.html +""" + +import enum +import os +import struct +from typing import IO, Optional, Tuple + + +class ELFInvalid(ValueError): + pass + + +class EIClass(enum.IntEnum): + C32 = 1 + C64 = 2 + + +class EIData(enum.IntEnum): + Lsb = 1 + Msb = 2 + + +class EMachine(enum.IntEnum): + I386 = 3 + S390 = 22 + Arm = 40 + X8664 = 62 + AArc64 = 183 + + +class ELFFile: + """ + Representation of an ELF executable. + """ + + def __init__(self, f: IO[bytes]) -> None: + self._f = f + + try: + ident = self._read("16B") + except struct.error: + raise ELFInvalid("unable to parse identification") + magic = bytes(ident[:4]) + if magic != b"\x7fELF": + raise ELFInvalid(f"invalid magic: {magic!r}") + + self.capacity = ident[4] # Format for program header (bitness). + self.encoding = ident[5] # Data structure encoding (endianness). + + try: + # e_fmt: Format for program header. + # p_fmt: Format for section header. + # p_idx: Indexes to find p_type, p_offset, and p_filesz. + e_fmt, self._p_fmt, self._p_idx = { + (1, 1): ("HHIIIIIHHH", ">IIIIIIII", (0, 1, 4)), # 32-bit MSB. + (2, 1): ("HHIQQQIHHH", ">IIQQQQQQ", (0, 2, 5)), # 64-bit MSB. + }[(self.capacity, self.encoding)] + except KeyError: + raise ELFInvalid( + f"unrecognized capacity ({self.capacity}) or " + f"encoding ({self.encoding})" + ) + + try: + ( + _, + self.machine, # Architecture type. + _, + _, + self._e_phoff, # Offset of program header. + _, + self.flags, # Processor-specific flags. + _, + self._e_phentsize, # Size of section. + self._e_phnum, # Number of sections. + ) = self._read(e_fmt) + except struct.error as e: + raise ELFInvalid("unable to parse machine and section information") from e + + def _read(self, fmt: str) -> Tuple[int, ...]: + return struct.unpack(fmt, self._f.read(struct.calcsize(fmt))) + + @property + def interpreter(self) -> Optional[str]: + """ + The path recorded in the ``PT_INTERP`` section header. + """ + for index in range(self._e_phnum): + self._f.seek(self._e_phoff + self._e_phentsize * index) + try: + data = self._read(self._p_fmt) + except struct.error: + continue + if data[self._p_idx[0]] != 3: # Not PT_INTERP. + continue + self._f.seek(data[self._p_idx[1]]) + return os.fsdecode(self._f.read(data[self._p_idx[2]])).strip("\0") + return None diff --git a/pylib/packaging/_manylinux.py b/pylib/packaging/_manylinux.py new file mode 100644 index 00000000..3705d50d --- /dev/null +++ b/pylib/packaging/_manylinux.py @@ -0,0 +1,252 @@ +import collections +import contextlib +import functools +import os +import re +import sys +import warnings +from typing import Dict, Generator, Iterator, NamedTuple, Optional, Sequence, Tuple + +from ._elffile import EIClass, EIData, ELFFile, EMachine + +EF_ARM_ABIMASK = 0xFF000000 +EF_ARM_ABI_VER5 = 0x05000000 +EF_ARM_ABI_FLOAT_HARD = 0x00000400 + + +# `os.PathLike` not a generic type until Python 3.9, so sticking with `str` +# as the type for `path` until then. +@contextlib.contextmanager +def _parse_elf(path: str) -> Generator[Optional[ELFFile], None, None]: + try: + with open(path, "rb") as f: + yield ELFFile(f) + except (OSError, TypeError, ValueError): + yield None + + +def _is_linux_armhf(executable: str) -> bool: + # hard-float ABI can be detected from the ELF header of the running + # process + # https://static.docs.arm.com/ihi0044/g/aaelf32.pdf + with _parse_elf(executable) as f: + return ( + f is not None + and f.capacity == EIClass.C32 + and f.encoding == EIData.Lsb + and f.machine == EMachine.Arm + and f.flags & EF_ARM_ABIMASK == EF_ARM_ABI_VER5 + and f.flags & EF_ARM_ABI_FLOAT_HARD == EF_ARM_ABI_FLOAT_HARD + ) + + +def _is_linux_i686(executable: str) -> bool: + with _parse_elf(executable) as f: + return ( + f is not None + and f.capacity == EIClass.C32 + and f.encoding == EIData.Lsb + and f.machine == EMachine.I386 + ) + + +def _have_compatible_abi(executable: str, archs: Sequence[str]) -> bool: + if "armv7l" in archs: + return _is_linux_armhf(executable) + if "i686" in archs: + return _is_linux_i686(executable) + allowed_archs = {"x86_64", "aarch64", "ppc64", "ppc64le", "s390x", "loongarch64"} + return any(arch in allowed_archs for arch in archs) + + +# If glibc ever changes its major version, we need to know what the last +# minor version was, so we can build the complete list of all versions. +# For now, guess what the highest minor version might be, assume it will +# be 50 for testing. Once this actually happens, update the dictionary +# with the actual value. +_LAST_GLIBC_MINOR: Dict[int, int] = collections.defaultdict(lambda: 50) + + +class _GLibCVersion(NamedTuple): + major: int + minor: int + + +def _glibc_version_string_confstr() -> Optional[str]: + """ + Primary implementation of glibc_version_string using os.confstr. + """ + # os.confstr is quite a bit faster than ctypes.DLL. It's also less likely + # to be broken or missing. This strategy is used in the standard library + # platform module. + # https://github.com/python/cpython/blob/fcf1d003bf4f0100c/Lib/platform.py#L175-L183 + try: + # Should be a string like "glibc 2.17". + version_string: str = getattr(os, "confstr")("CS_GNU_LIBC_VERSION") + assert version_string is not None + _, version = version_string.rsplit() + except (AssertionError, AttributeError, OSError, ValueError): + # os.confstr() or CS_GNU_LIBC_VERSION not available (or a bad value)... + return None + return version + + +def _glibc_version_string_ctypes() -> Optional[str]: + """ + Fallback implementation of glibc_version_string using ctypes. + """ + try: + import ctypes + except ImportError: + return None + + # ctypes.CDLL(None) internally calls dlopen(NULL), and as the dlopen + # manpage says, "If filename is NULL, then the returned handle is for the + # main program". This way we can let the linker do the work to figure out + # which libc our process is actually using. + # + # We must also handle the special case where the executable is not a + # dynamically linked executable. This can occur when using musl libc, + # for example. In this situation, dlopen() will error, leading to an + # OSError. Interestingly, at least in the case of musl, there is no + # errno set on the OSError. The single string argument used to construct + # OSError comes from libc itself and is therefore not portable to + # hard code here. In any case, failure to call dlopen() means we + # can proceed, so we bail on our attempt. + try: + process_namespace = ctypes.CDLL(None) + except OSError: + return None + + try: + gnu_get_libc_version = process_namespace.gnu_get_libc_version + except AttributeError: + # Symbol doesn't exist -> therefore, we are not linked to + # glibc. + return None + + # Call gnu_get_libc_version, which returns a string like "2.5" + gnu_get_libc_version.restype = ctypes.c_char_p + version_str: str = gnu_get_libc_version() + # py2 / py3 compatibility: + if not isinstance(version_str, str): + version_str = version_str.decode("ascii") + + return version_str + + +def _glibc_version_string() -> Optional[str]: + """Returns glibc version string, or None if not using glibc.""" + return _glibc_version_string_confstr() or _glibc_version_string_ctypes() + + +def _parse_glibc_version(version_str: str) -> Tuple[int, int]: + """Parse glibc version. + + We use a regexp instead of str.split because we want to discard any + random junk that might come after the minor version -- this might happen + in patched/forked versions of glibc (e.g. Linaro's version of glibc + uses version strings like "2.20-2014.11"). See gh-3588. + """ + m = re.match(r"(?P[0-9]+)\.(?P[0-9]+)", version_str) + if not m: + warnings.warn( + f"Expected glibc version with 2 components major.minor," + f" got: {version_str}", + RuntimeWarning, + ) + return -1, -1 + return int(m.group("major")), int(m.group("minor")) + + +@functools.lru_cache() +def _get_glibc_version() -> Tuple[int, int]: + version_str = _glibc_version_string() + if version_str is None: + return (-1, -1) + return _parse_glibc_version(version_str) + + +# From PEP 513, PEP 600 +def _is_compatible(arch: str, version: _GLibCVersion) -> bool: + sys_glibc = _get_glibc_version() + if sys_glibc < version: + return False + # Check for presence of _manylinux module. + try: + import _manylinux # noqa + except ImportError: + return True + if hasattr(_manylinux, "manylinux_compatible"): + result = _manylinux.manylinux_compatible(version[0], version[1], arch) + if result is not None: + return bool(result) + return True + if version == _GLibCVersion(2, 5): + if hasattr(_manylinux, "manylinux1_compatible"): + return bool(_manylinux.manylinux1_compatible) + if version == _GLibCVersion(2, 12): + if hasattr(_manylinux, "manylinux2010_compatible"): + return bool(_manylinux.manylinux2010_compatible) + if version == _GLibCVersion(2, 17): + if hasattr(_manylinux, "manylinux2014_compatible"): + return bool(_manylinux.manylinux2014_compatible) + return True + + +_LEGACY_MANYLINUX_MAP = { + # CentOS 7 w/ glibc 2.17 (PEP 599) + (2, 17): "manylinux2014", + # CentOS 6 w/ glibc 2.12 (PEP 571) + (2, 12): "manylinux2010", + # CentOS 5 w/ glibc 2.5 (PEP 513) + (2, 5): "manylinux1", +} + + +def platform_tags(archs: Sequence[str]) -> Iterator[str]: + """Generate manylinux tags compatible to the current platform. + + :param archs: Sequence of compatible architectures. + The first one shall be the closest to the actual architecture and be the part of + platform tag after the ``linux_`` prefix, e.g. ``x86_64``. + The ``linux_`` prefix is assumed as a prerequisite for the current platform to + be manylinux-compatible. + + :returns: An iterator of compatible manylinux tags. + """ + if not _have_compatible_abi(sys.executable, archs): + return + # Oldest glibc to be supported regardless of architecture is (2, 17). + too_old_glibc2 = _GLibCVersion(2, 16) + if set(archs) & {"x86_64", "i686"}: + # On x86/i686 also oldest glibc to be supported is (2, 5). + too_old_glibc2 = _GLibCVersion(2, 4) + current_glibc = _GLibCVersion(*_get_glibc_version()) + glibc_max_list = [current_glibc] + # We can assume compatibility across glibc major versions. + # https://sourceware.org/bugzilla/show_bug.cgi?id=24636 + # + # Build a list of maximum glibc versions so that we can + # output the canonical list of all glibc from current_glibc + # down to too_old_glibc2, including all intermediary versions. + for glibc_major in range(current_glibc.major - 1, 1, -1): + glibc_minor = _LAST_GLIBC_MINOR[glibc_major] + glibc_max_list.append(_GLibCVersion(glibc_major, glibc_minor)) + for arch in archs: + for glibc_max in glibc_max_list: + if glibc_max.major == too_old_glibc2.major: + min_minor = too_old_glibc2.minor + else: + # For other glibc major versions oldest supported is (x, 0). + min_minor = -1 + for glibc_minor in range(glibc_max.minor, min_minor, -1): + glibc_version = _GLibCVersion(glibc_max.major, glibc_minor) + tag = "manylinux_{}_{}".format(*glibc_version) + if _is_compatible(arch, glibc_version): + yield f"{tag}_{arch}" + # Handle the legacy manylinux1, manylinux2010, manylinux2014 tags. + if glibc_version in _LEGACY_MANYLINUX_MAP: + legacy_tag = _LEGACY_MANYLINUX_MAP[glibc_version] + if _is_compatible(arch, glibc_version): + yield f"{legacy_tag}_{arch}" diff --git a/pylib/packaging/_musllinux.py b/pylib/packaging/_musllinux.py new file mode 100644 index 00000000..86419df9 --- /dev/null +++ b/pylib/packaging/_musllinux.py @@ -0,0 +1,83 @@ +"""PEP 656 support. + +This module implements logic to detect if the currently running Python is +linked against musl, and what musl version is used. +""" + +import functools +import re +import subprocess +import sys +from typing import Iterator, NamedTuple, Optional, Sequence + +from ._elffile import ELFFile + + +class _MuslVersion(NamedTuple): + major: int + minor: int + + +def _parse_musl_version(output: str) -> Optional[_MuslVersion]: + lines = [n for n in (n.strip() for n in output.splitlines()) if n] + if len(lines) < 2 or lines[0][:4] != "musl": + return None + m = re.match(r"Version (\d+)\.(\d+)", lines[1]) + if not m: + return None + return _MuslVersion(major=int(m.group(1)), minor=int(m.group(2))) + + +@functools.lru_cache() +def _get_musl_version(executable: str) -> Optional[_MuslVersion]: + """Detect currently-running musl runtime version. + + This is done by checking the specified executable's dynamic linking + information, and invoking the loader to parse its output for a version + string. If the loader is musl, the output would be something like:: + + musl libc (x86_64) + Version 1.2.2 + Dynamic Program Loader + """ + try: + with open(executable, "rb") as f: + ld = ELFFile(f).interpreter + except (OSError, TypeError, ValueError): + return None + if ld is None or "musl" not in ld: + return None + proc = subprocess.run([ld], stderr=subprocess.PIPE, text=True) + return _parse_musl_version(proc.stderr) + + +def platform_tags(archs: Sequence[str]) -> Iterator[str]: + """Generate musllinux tags compatible to the current platform. + + :param archs: Sequence of compatible architectures. + The first one shall be the closest to the actual architecture and be the part of + platform tag after the ``linux_`` prefix, e.g. ``x86_64``. + The ``linux_`` prefix is assumed as a prerequisite for the current platform to + be musllinux-compatible. + + :returns: An iterator of compatible musllinux tags. + """ + sys_musl = _get_musl_version(sys.executable) + if sys_musl is None: # Python not dynamically linked against musl. + return + for arch in archs: + for minor in range(sys_musl.minor, -1, -1): + yield f"musllinux_{sys_musl.major}_{minor}_{arch}" + + +if __name__ == "__main__": # pragma: no cover + import sysconfig + + plat = sysconfig.get_platform() + assert plat.startswith("linux-"), "not linux" + + print("plat:", plat) + print("musl:", _get_musl_version(sys.executable)) + print("tags:", end=" ") + for t in platform_tags(re.sub(r"[.-]", "_", plat.split("-", 1)[-1])): + print(t, end="\n ") diff --git a/pylib/packaging/_parser.py b/pylib/packaging/_parser.py new file mode 100644 index 00000000..4576981c --- /dev/null +++ b/pylib/packaging/_parser.py @@ -0,0 +1,359 @@ +"""Handwritten parser of dependency specifiers. + +The docstring for each __parse_* function contains ENBF-inspired grammar representing +the implementation. +""" + +import ast +from typing import Any, List, NamedTuple, Optional, Tuple, Union + +from ._tokenizer import DEFAULT_RULES, Tokenizer + + +class Node: + def __init__(self, value: str) -> None: + self.value = value + + def __str__(self) -> str: + return self.value + + def __repr__(self) -> str: + return f"<{self.__class__.__name__}('{self}')>" + + def serialize(self) -> str: + raise NotImplementedError + + +class Variable(Node): + def serialize(self) -> str: + return str(self) + + +class Value(Node): + def serialize(self) -> str: + return f'"{self}"' + + +class Op(Node): + def serialize(self) -> str: + return str(self) + + +MarkerVar = Union[Variable, Value] +MarkerItem = Tuple[MarkerVar, Op, MarkerVar] +# MarkerAtom = Union[MarkerItem, List["MarkerAtom"]] +# MarkerList = List[Union["MarkerList", MarkerAtom, str]] +# mypy does not support recursive type definition +# https://github.com/python/mypy/issues/731 +MarkerAtom = Any +MarkerList = List[Any] + + +class ParsedRequirement(NamedTuple): + name: str + url: str + extras: List[str] + specifier: str + marker: Optional[MarkerList] + + +# -------------------------------------------------------------------------------------- +# Recursive descent parser for dependency specifier +# -------------------------------------------------------------------------------------- +def parse_requirement(source: str) -> ParsedRequirement: + return _parse_requirement(Tokenizer(source, rules=DEFAULT_RULES)) + + +def _parse_requirement(tokenizer: Tokenizer) -> ParsedRequirement: + """ + requirement = WS? IDENTIFIER WS? extras WS? requirement_details + """ + tokenizer.consume("WS") + + name_token = tokenizer.expect( + "IDENTIFIER", expected="package name at the start of dependency specifier" + ) + name = name_token.text + tokenizer.consume("WS") + + extras = _parse_extras(tokenizer) + tokenizer.consume("WS") + + url, specifier, marker = _parse_requirement_details(tokenizer) + tokenizer.expect("END", expected="end of dependency specifier") + + return ParsedRequirement(name, url, extras, specifier, marker) + + +def _parse_requirement_details( + tokenizer: Tokenizer, +) -> Tuple[str, str, Optional[MarkerList]]: + """ + requirement_details = AT URL (WS requirement_marker?)? + | specifier WS? (requirement_marker)? + """ + + specifier = "" + url = "" + marker = None + + if tokenizer.check("AT"): + tokenizer.read() + tokenizer.consume("WS") + + url_start = tokenizer.position + url = tokenizer.expect("URL", expected="URL after @").text + if tokenizer.check("END", peek=True): + return (url, specifier, marker) + + tokenizer.expect("WS", expected="whitespace after URL") + + # The input might end after whitespace. + if tokenizer.check("END", peek=True): + return (url, specifier, marker) + + marker = _parse_requirement_marker( + tokenizer, span_start=url_start, after="URL and whitespace" + ) + else: + specifier_start = tokenizer.position + specifier = _parse_specifier(tokenizer) + tokenizer.consume("WS") + + if tokenizer.check("END", peek=True): + return (url, specifier, marker) + + marker = _parse_requirement_marker( + tokenizer, + span_start=specifier_start, + after=( + "version specifier" + if specifier + else "name and no valid version specifier" + ), + ) + + return (url, specifier, marker) + + +def _parse_requirement_marker( + tokenizer: Tokenizer, *, span_start: int, after: str +) -> MarkerList: + """ + requirement_marker = SEMICOLON marker WS? + """ + + if not tokenizer.check("SEMICOLON"): + tokenizer.raise_syntax_error( + f"Expected end or semicolon (after {after})", + span_start=span_start, + ) + tokenizer.read() + + marker = _parse_marker(tokenizer) + tokenizer.consume("WS") + + return marker + + +def _parse_extras(tokenizer: Tokenizer) -> List[str]: + """ + extras = (LEFT_BRACKET wsp* extras_list? wsp* RIGHT_BRACKET)? + """ + if not tokenizer.check("LEFT_BRACKET", peek=True): + return [] + + with tokenizer.enclosing_tokens( + "LEFT_BRACKET", + "RIGHT_BRACKET", + around="extras", + ): + tokenizer.consume("WS") + extras = _parse_extras_list(tokenizer) + tokenizer.consume("WS") + + return extras + + +def _parse_extras_list(tokenizer: Tokenizer) -> List[str]: + """ + extras_list = identifier (wsp* ',' wsp* identifier)* + """ + extras: List[str] = [] + + if not tokenizer.check("IDENTIFIER"): + return extras + + extras.append(tokenizer.read().text) + + while True: + tokenizer.consume("WS") + if tokenizer.check("IDENTIFIER", peek=True): + tokenizer.raise_syntax_error("Expected comma between extra names") + elif not tokenizer.check("COMMA"): + break + + tokenizer.read() + tokenizer.consume("WS") + + extra_token = tokenizer.expect("IDENTIFIER", expected="extra name after comma") + extras.append(extra_token.text) + + return extras + + +def _parse_specifier(tokenizer: Tokenizer) -> str: + """ + specifier = LEFT_PARENTHESIS WS? version_many WS? RIGHT_PARENTHESIS + | WS? version_many WS? + """ + with tokenizer.enclosing_tokens( + "LEFT_PARENTHESIS", + "RIGHT_PARENTHESIS", + around="version specifier", + ): + tokenizer.consume("WS") + parsed_specifiers = _parse_version_many(tokenizer) + tokenizer.consume("WS") + + return parsed_specifiers + + +def _parse_version_many(tokenizer: Tokenizer) -> str: + """ + version_many = (SPECIFIER (WS? COMMA WS? SPECIFIER)*)? + """ + parsed_specifiers = "" + while tokenizer.check("SPECIFIER"): + span_start = tokenizer.position + parsed_specifiers += tokenizer.read().text + if tokenizer.check("VERSION_PREFIX_TRAIL", peek=True): + tokenizer.raise_syntax_error( + ".* suffix can only be used with `==` or `!=` operators", + span_start=span_start, + span_end=tokenizer.position + 1, + ) + if tokenizer.check("VERSION_LOCAL_LABEL_TRAIL", peek=True): + tokenizer.raise_syntax_error( + "Local version label can only be used with `==` or `!=` operators", + span_start=span_start, + span_end=tokenizer.position, + ) + tokenizer.consume("WS") + if not tokenizer.check("COMMA"): + break + parsed_specifiers += tokenizer.read().text + tokenizer.consume("WS") + + return parsed_specifiers + + +# -------------------------------------------------------------------------------------- +# Recursive descent parser for marker expression +# -------------------------------------------------------------------------------------- +def parse_marker(source: str) -> MarkerList: + return _parse_full_marker(Tokenizer(source, rules=DEFAULT_RULES)) + + +def _parse_full_marker(tokenizer: Tokenizer) -> MarkerList: + retval = _parse_marker(tokenizer) + tokenizer.expect("END", expected="end of marker expression") + return retval + + +def _parse_marker(tokenizer: Tokenizer) -> MarkerList: + """ + marker = marker_atom (BOOLOP marker_atom)+ + """ + expression = [_parse_marker_atom(tokenizer)] + while tokenizer.check("BOOLOP"): + token = tokenizer.read() + expr_right = _parse_marker_atom(tokenizer) + expression.extend((token.text, expr_right)) + return expression + + +def _parse_marker_atom(tokenizer: Tokenizer) -> MarkerAtom: + """ + marker_atom = WS? LEFT_PARENTHESIS WS? marker WS? RIGHT_PARENTHESIS WS? + | WS? marker_item WS? + """ + + tokenizer.consume("WS") + if tokenizer.check("LEFT_PARENTHESIS", peek=True): + with tokenizer.enclosing_tokens( + "LEFT_PARENTHESIS", + "RIGHT_PARENTHESIS", + around="marker expression", + ): + tokenizer.consume("WS") + marker: MarkerAtom = _parse_marker(tokenizer) + tokenizer.consume("WS") + else: + marker = _parse_marker_item(tokenizer) + tokenizer.consume("WS") + return marker + + +def _parse_marker_item(tokenizer: Tokenizer) -> MarkerItem: + """ + marker_item = WS? marker_var WS? marker_op WS? marker_var WS? + """ + tokenizer.consume("WS") + marker_var_left = _parse_marker_var(tokenizer) + tokenizer.consume("WS") + marker_op = _parse_marker_op(tokenizer) + tokenizer.consume("WS") + marker_var_right = _parse_marker_var(tokenizer) + tokenizer.consume("WS") + return (marker_var_left, marker_op, marker_var_right) + + +def _parse_marker_var(tokenizer: Tokenizer) -> MarkerVar: + """ + marker_var = VARIABLE | QUOTED_STRING + """ + if tokenizer.check("VARIABLE"): + return process_env_var(tokenizer.read().text.replace(".", "_")) + elif tokenizer.check("QUOTED_STRING"): + return process_python_str(tokenizer.read().text) + else: + tokenizer.raise_syntax_error( + message="Expected a marker variable or quoted string" + ) + + +def process_env_var(env_var: str) -> Variable: + if ( + env_var == "platform_python_implementation" + or env_var == "python_implementation" + ): + return Variable("platform_python_implementation") + else: + return Variable(env_var) + + +def process_python_str(python_str: str) -> Value: + value = ast.literal_eval(python_str) + return Value(str(value)) + + +def _parse_marker_op(tokenizer: Tokenizer) -> Op: + """ + marker_op = IN | NOT IN | OP + """ + if tokenizer.check("IN"): + tokenizer.read() + return Op("in") + elif tokenizer.check("NOT"): + tokenizer.read() + tokenizer.expect("WS", expected="whitespace after 'not'") + tokenizer.expect("IN", expected="'in' after 'not'") + return Op("not in") + elif tokenizer.check("OP"): + return Op(tokenizer.read().text) + else: + return tokenizer.raise_syntax_error( + "Expected marker operator, one of " + "<=, <, !=, ==, >=, >, ~=, ===, in, not in" + ) diff --git a/pylib/packaging/_structures.py b/pylib/packaging/_structures.py new file mode 100644 index 00000000..90a6465f --- /dev/null +++ b/pylib/packaging/_structures.py @@ -0,0 +1,61 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + + +class InfinityType: + def __repr__(self) -> str: + return "Infinity" + + def __hash__(self) -> int: + return hash(repr(self)) + + def __lt__(self, other: object) -> bool: + return False + + def __le__(self, other: object) -> bool: + return False + + def __eq__(self, other: object) -> bool: + return isinstance(other, self.__class__) + + def __gt__(self, other: object) -> bool: + return True + + def __ge__(self, other: object) -> bool: + return True + + def __neg__(self: object) -> "NegativeInfinityType": + return NegativeInfinity + + +Infinity = InfinityType() + + +class NegativeInfinityType: + def __repr__(self) -> str: + return "-Infinity" + + def __hash__(self) -> int: + return hash(repr(self)) + + def __lt__(self, other: object) -> bool: + return True + + def __le__(self, other: object) -> bool: + return True + + def __eq__(self, other: object) -> bool: + return isinstance(other, self.__class__) + + def __gt__(self, other: object) -> bool: + return False + + def __ge__(self, other: object) -> bool: + return False + + def __neg__(self: object) -> InfinityType: + return Infinity + + +NegativeInfinity = NegativeInfinityType() diff --git a/pylib/packaging/_tokenizer.py b/pylib/packaging/_tokenizer.py new file mode 100644 index 00000000..dd0d648d --- /dev/null +++ b/pylib/packaging/_tokenizer.py @@ -0,0 +1,192 @@ +import contextlib +import re +from dataclasses import dataclass +from typing import Dict, Iterator, NoReturn, Optional, Tuple, Union + +from .specifiers import Specifier + + +@dataclass +class Token: + name: str + text: str + position: int + + +class ParserSyntaxError(Exception): + """The provided source text could not be parsed correctly.""" + + def __init__( + self, + message: str, + *, + source: str, + span: Tuple[int, int], + ) -> None: + self.span = span + self.message = message + self.source = source + + super().__init__() + + def __str__(self) -> str: + marker = " " * self.span[0] + "~" * (self.span[1] - self.span[0]) + "^" + return "\n ".join([self.message, self.source, marker]) + + +DEFAULT_RULES: "Dict[str, Union[str, re.Pattern[str]]]" = { + "LEFT_PARENTHESIS": r"\(", + "RIGHT_PARENTHESIS": r"\)", + "LEFT_BRACKET": r"\[", + "RIGHT_BRACKET": r"\]", + "SEMICOLON": r";", + "COMMA": r",", + "QUOTED_STRING": re.compile( + r""" + ( + ('[^']*') + | + ("[^"]*") + ) + """, + re.VERBOSE, + ), + "OP": r"(===|==|~=|!=|<=|>=|<|>)", + "BOOLOP": r"\b(or|and)\b", + "IN": r"\bin\b", + "NOT": r"\bnot\b", + "VARIABLE": re.compile( + r""" + \b( + python_version + |python_full_version + |os[._]name + |sys[._]platform + |platform_(release|system) + |platform[._](version|machine|python_implementation) + |python_implementation + |implementation_(name|version) + |extra + )\b + """, + re.VERBOSE, + ), + "SPECIFIER": re.compile( + Specifier._operator_regex_str + Specifier._version_regex_str, + re.VERBOSE | re.IGNORECASE, + ), + "AT": r"\@", + "URL": r"[^ \t]+", + "IDENTIFIER": r"\b[a-zA-Z0-9][a-zA-Z0-9._-]*\b", + "VERSION_PREFIX_TRAIL": r"\.\*", + "VERSION_LOCAL_LABEL_TRAIL": r"\+[a-z0-9]+(?:[-_\.][a-z0-9]+)*", + "WS": r"[ \t]+", + "END": r"$", +} + + +class Tokenizer: + """Context-sensitive token parsing. + + Provides methods to examine the input stream to check whether the next token + matches. + """ + + def __init__( + self, + source: str, + *, + rules: "Dict[str, Union[str, re.Pattern[str]]]", + ) -> None: + self.source = source + self.rules: Dict[str, re.Pattern[str]] = { + name: re.compile(pattern) for name, pattern in rules.items() + } + self.next_token: Optional[Token] = None + self.position = 0 + + def consume(self, name: str) -> None: + """Move beyond provided token name, if at current position.""" + if self.check(name): + self.read() + + def check(self, name: str, *, peek: bool = False) -> bool: + """Check whether the next token has the provided name. + + By default, if the check succeeds, the token *must* be read before + another check. If `peek` is set to `True`, the token is not loaded and + would need to be checked again. + """ + assert ( + self.next_token is None + ), f"Cannot check for {name!r}, already have {self.next_token!r}" + assert name in self.rules, f"Unknown token name: {name!r}" + + expression = self.rules[name] + + match = expression.match(self.source, self.position) + if match is None: + return False + if not peek: + self.next_token = Token(name, match[0], self.position) + return True + + def expect(self, name: str, *, expected: str) -> Token: + """Expect a certain token name next, failing with a syntax error otherwise. + + The token is *not* read. + """ + if not self.check(name): + raise self.raise_syntax_error(f"Expected {expected}") + return self.read() + + def read(self) -> Token: + """Consume the next token and return it.""" + token = self.next_token + assert token is not None + + self.position += len(token.text) + self.next_token = None + + return token + + def raise_syntax_error( + self, + message: str, + *, + span_start: Optional[int] = None, + span_end: Optional[int] = None, + ) -> NoReturn: + """Raise ParserSyntaxError at the given position.""" + span = ( + self.position if span_start is None else span_start, + self.position if span_end is None else span_end, + ) + raise ParserSyntaxError( + message, + source=self.source, + span=span, + ) + + @contextlib.contextmanager + def enclosing_tokens( + self, open_token: str, close_token: str, *, around: str + ) -> Iterator[None]: + if self.check(open_token): + open_position = self.position + self.read() + else: + open_position = None + + yield + + if open_position is None: + return + + if not self.check(close_token): + self.raise_syntax_error( + f"Expected matching {close_token} for {open_token}, after {around}", + span_start=open_position, + ) + + self.read() diff --git a/pylib/packaging/markers.py b/pylib/packaging/markers.py new file mode 100644 index 00000000..8b98fca7 --- /dev/null +++ b/pylib/packaging/markers.py @@ -0,0 +1,252 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + +import operator +import os +import platform +import sys +from typing import Any, Callable, Dict, List, Optional, Tuple, Union + +from ._parser import ( + MarkerAtom, + MarkerList, + Op, + Value, + Variable, + parse_marker as _parse_marker, +) +from ._tokenizer import ParserSyntaxError +from .specifiers import InvalidSpecifier, Specifier +from .utils import canonicalize_name + +__all__ = [ + "InvalidMarker", + "UndefinedComparison", + "UndefinedEnvironmentName", + "Marker", + "default_environment", +] + +Operator = Callable[[str, str], bool] + + +class InvalidMarker(ValueError): + """ + An invalid marker was found, users should refer to PEP 508. + """ + + +class UndefinedComparison(ValueError): + """ + An invalid operation was attempted on a value that doesn't support it. + """ + + +class UndefinedEnvironmentName(ValueError): + """ + A name was attempted to be used that does not exist inside of the + environment. + """ + + +def _normalize_extra_values(results: Any) -> Any: + """ + Normalize extra values. + """ + if isinstance(results[0], tuple): + lhs, op, rhs = results[0] + if isinstance(lhs, Variable) and lhs.value == "extra": + normalized_extra = canonicalize_name(rhs.value) + rhs = Value(normalized_extra) + elif isinstance(rhs, Variable) and rhs.value == "extra": + normalized_extra = canonicalize_name(lhs.value) + lhs = Value(normalized_extra) + results[0] = lhs, op, rhs + return results + + +def _format_marker( + marker: Union[List[str], MarkerAtom, str], first: Optional[bool] = True +) -> str: + + assert isinstance(marker, (list, tuple, str)) + + # Sometimes we have a structure like [[...]] which is a single item list + # where the single item is itself it's own list. In that case we want skip + # the rest of this function so that we don't get extraneous () on the + # outside. + if ( + isinstance(marker, list) + and len(marker) == 1 + and isinstance(marker[0], (list, tuple)) + ): + return _format_marker(marker[0]) + + if isinstance(marker, list): + inner = (_format_marker(m, first=False) for m in marker) + if first: + return " ".join(inner) + else: + return "(" + " ".join(inner) + ")" + elif isinstance(marker, tuple): + return " ".join([m.serialize() for m in marker]) + else: + return marker + + +_operators: Dict[str, Operator] = { + "in": lambda lhs, rhs: lhs in rhs, + "not in": lambda lhs, rhs: lhs not in rhs, + "<": operator.lt, + "<=": operator.le, + "==": operator.eq, + "!=": operator.ne, + ">=": operator.ge, + ">": operator.gt, +} + + +def _eval_op(lhs: str, op: Op, rhs: str) -> bool: + try: + spec = Specifier("".join([op.serialize(), rhs])) + except InvalidSpecifier: + pass + else: + return spec.contains(lhs, prereleases=True) + + oper: Optional[Operator] = _operators.get(op.serialize()) + if oper is None: + raise UndefinedComparison(f"Undefined {op!r} on {lhs!r} and {rhs!r}.") + + return oper(lhs, rhs) + + +def _normalize(*values: str, key: str) -> Tuple[str, ...]: + # PEP 685 – Comparison of extra names for optional distribution dependencies + # https://peps.python.org/pep-0685/ + # > When comparing extra names, tools MUST normalize the names being + # > compared using the semantics outlined in PEP 503 for names + if key == "extra": + return tuple(canonicalize_name(v) for v in values) + + # other environment markers don't have such standards + return values + + +def _evaluate_markers(markers: MarkerList, environment: Dict[str, str]) -> bool: + groups: List[List[bool]] = [[]] + + for marker in markers: + assert isinstance(marker, (list, tuple, str)) + + if isinstance(marker, list): + groups[-1].append(_evaluate_markers(marker, environment)) + elif isinstance(marker, tuple): + lhs, op, rhs = marker + + if isinstance(lhs, Variable): + environment_key = lhs.value + lhs_value = environment[environment_key] + rhs_value = rhs.value + else: + lhs_value = lhs.value + environment_key = rhs.value + rhs_value = environment[environment_key] + + lhs_value, rhs_value = _normalize(lhs_value, rhs_value, key=environment_key) + groups[-1].append(_eval_op(lhs_value, op, rhs_value)) + else: + assert marker in ["and", "or"] + if marker == "or": + groups.append([]) + + return any(all(item) for item in groups) + + +def format_full_version(info: "sys._version_info") -> str: + version = "{0.major}.{0.minor}.{0.micro}".format(info) + kind = info.releaselevel + if kind != "final": + version += kind[0] + str(info.serial) + return version + + +def default_environment() -> Dict[str, str]: + iver = format_full_version(sys.implementation.version) + implementation_name = sys.implementation.name + return { + "implementation_name": implementation_name, + "implementation_version": iver, + "os_name": os.name, + "platform_machine": platform.machine(), + "platform_release": platform.release(), + "platform_system": platform.system(), + "platform_version": platform.version(), + "python_full_version": platform.python_version(), + "platform_python_implementation": platform.python_implementation(), + "python_version": ".".join(platform.python_version_tuple()[:2]), + "sys_platform": sys.platform, + } + + +class Marker: + def __init__(self, marker: str) -> None: + # Note: We create a Marker object without calling this constructor in + # packaging.requirements.Requirement. If any additional logic is + # added here, make sure to mirror/adapt Requirement. + try: + self._markers = _normalize_extra_values(_parse_marker(marker)) + # The attribute `_markers` can be described in terms of a recursive type: + # MarkerList = List[Union[Tuple[Node, ...], str, MarkerList]] + # + # For example, the following expression: + # python_version > "3.6" or (python_version == "3.6" and os_name == "unix") + # + # is parsed into: + # [ + # (, ')>, ), + # 'and', + # [ + # (, , ), + # 'or', + # (, , ) + # ] + # ] + except ParserSyntaxError as e: + raise InvalidMarker(str(e)) from e + + def __str__(self) -> str: + return _format_marker(self._markers) + + def __repr__(self) -> str: + return f"" + + def __hash__(self) -> int: + return hash((self.__class__.__name__, str(self))) + + def __eq__(self, other: Any) -> bool: + if not isinstance(other, Marker): + return NotImplemented + + return str(self) == str(other) + + def evaluate(self, environment: Optional[Dict[str, str]] = None) -> bool: + """Evaluate a marker. + + Return the boolean from evaluating the given marker against the + environment. environment is an optional argument to override all or + part of the determined environment. + + The environment is determined from the current Python process. + """ + current_environment = default_environment() + current_environment["extra"] = "" + if environment is not None: + current_environment.update(environment) + # The API used to allow setting extra to None. We need to handle this + # case for backwards compatibility. + if current_environment["extra"] is None: + current_environment["extra"] = "" + + return _evaluate_markers(self._markers, current_environment) diff --git a/pylib/packaging/metadata.py b/pylib/packaging/metadata.py new file mode 100644 index 00000000..fb274930 --- /dev/null +++ b/pylib/packaging/metadata.py @@ -0,0 +1,825 @@ +import email.feedparser +import email.header +import email.message +import email.parser +import email.policy +import sys +import typing +from typing import ( + Any, + Callable, + Dict, + Generic, + List, + Optional, + Tuple, + Type, + Union, + cast, +) + +from . import requirements, specifiers, utils, version as version_module + +T = typing.TypeVar("T") +if sys.version_info[:2] >= (3, 8): # pragma: no cover + from typing import Literal, TypedDict +else: # pragma: no cover + if typing.TYPE_CHECKING: + from typing_extensions import Literal, TypedDict + else: + try: + from typing_extensions import Literal, TypedDict + except ImportError: + + class Literal: + def __init_subclass__(*_args, **_kwargs): + pass + + class TypedDict: + def __init_subclass__(*_args, **_kwargs): + pass + + +try: + ExceptionGroup +except NameError: # pragma: no cover + + class ExceptionGroup(Exception): # noqa: N818 + """A minimal implementation of :external:exc:`ExceptionGroup` from Python 3.11. + + If :external:exc:`ExceptionGroup` is already defined by Python itself, + that version is used instead. + """ + + message: str + exceptions: List[Exception] + + def __init__(self, message: str, exceptions: List[Exception]) -> None: + self.message = message + self.exceptions = exceptions + + def __repr__(self) -> str: + return f"{self.__class__.__name__}({self.message!r}, {self.exceptions!r})" + +else: # pragma: no cover + ExceptionGroup = ExceptionGroup + + +class InvalidMetadata(ValueError): + """A metadata field contains invalid data.""" + + field: str + """The name of the field that contains invalid data.""" + + def __init__(self, field: str, message: str) -> None: + self.field = field + super().__init__(message) + + +# The RawMetadata class attempts to make as few assumptions about the underlying +# serialization formats as possible. The idea is that as long as a serialization +# formats offer some very basic primitives in *some* way then we can support +# serializing to and from that format. +class RawMetadata(TypedDict, total=False): + """A dictionary of raw core metadata. + + Each field in core metadata maps to a key of this dictionary (when data is + provided). The key is lower-case and underscores are used instead of dashes + compared to the equivalent core metadata field. Any core metadata field that + can be specified multiple times or can hold multiple values in a single + field have a key with a plural name. See :class:`Metadata` whose attributes + match the keys of this dictionary. + + Core metadata fields that can be specified multiple times are stored as a + list or dict depending on which is appropriate for the field. Any fields + which hold multiple values in a single field are stored as a list. + + """ + + # Metadata 1.0 - PEP 241 + metadata_version: str + name: str + version: str + platforms: List[str] + summary: str + description: str + keywords: List[str] + home_page: str + author: str + author_email: str + license: str + + # Metadata 1.1 - PEP 314 + supported_platforms: List[str] + download_url: str + classifiers: List[str] + requires: List[str] + provides: List[str] + obsoletes: List[str] + + # Metadata 1.2 - PEP 345 + maintainer: str + maintainer_email: str + requires_dist: List[str] + provides_dist: List[str] + obsoletes_dist: List[str] + requires_python: str + requires_external: List[str] + project_urls: Dict[str, str] + + # Metadata 2.0 + # PEP 426 attempted to completely revamp the metadata format + # but got stuck without ever being able to build consensus on + # it and ultimately ended up withdrawn. + # + # However, a number of tools had started emitting METADATA with + # `2.0` Metadata-Version, so for historical reasons, this version + # was skipped. + + # Metadata 2.1 - PEP 566 + description_content_type: str + provides_extra: List[str] + + # Metadata 2.2 - PEP 643 + dynamic: List[str] + + # Metadata 2.3 - PEP 685 + # No new fields were added in PEP 685, just some edge case were + # tightened up to provide better interoptability. + + +_STRING_FIELDS = { + "author", + "author_email", + "description", + "description_content_type", + "download_url", + "home_page", + "license", + "maintainer", + "maintainer_email", + "metadata_version", + "name", + "requires_python", + "summary", + "version", +} + +_LIST_FIELDS = { + "classifiers", + "dynamic", + "obsoletes", + "obsoletes_dist", + "platforms", + "provides", + "provides_dist", + "provides_extra", + "requires", + "requires_dist", + "requires_external", + "supported_platforms", +} + +_DICT_FIELDS = { + "project_urls", +} + + +def _parse_keywords(data: str) -> List[str]: + """Split a string of comma-separate keyboards into a list of keywords.""" + return [k.strip() for k in data.split(",")] + + +def _parse_project_urls(data: List[str]) -> Dict[str, str]: + """Parse a list of label/URL string pairings separated by a comma.""" + urls = {} + for pair in data: + # Our logic is slightly tricky here as we want to try and do + # *something* reasonable with malformed data. + # + # The main thing that we have to worry about, is data that does + # not have a ',' at all to split the label from the Value. There + # isn't a singular right answer here, and we will fail validation + # later on (if the caller is validating) so it doesn't *really* + # matter, but since the missing value has to be an empty str + # and our return value is dict[str, str], if we let the key + # be the missing value, then they'd have multiple '' values that + # overwrite each other in a accumulating dict. + # + # The other potentional issue is that it's possible to have the + # same label multiple times in the metadata, with no solid "right" + # answer with what to do in that case. As such, we'll do the only + # thing we can, which is treat the field as unparseable and add it + # to our list of unparsed fields. + parts = [p.strip() for p in pair.split(",", 1)] + parts.extend([""] * (max(0, 2 - len(parts)))) # Ensure 2 items + + # TODO: The spec doesn't say anything about if the keys should be + # considered case sensitive or not... logically they should + # be case-preserving and case-insensitive, but doing that + # would open up more cases where we might have duplicate + # entries. + label, url = parts + if label in urls: + # The label already exists in our set of urls, so this field + # is unparseable, and we can just add the whole thing to our + # unparseable data and stop processing it. + raise KeyError("duplicate labels in project urls") + urls[label] = url + + return urls + + +def _get_payload(msg: email.message.Message, source: Union[bytes, str]) -> str: + """Get the body of the message.""" + # If our source is a str, then our caller has managed encodings for us, + # and we don't need to deal with it. + if isinstance(source, str): + payload: str = msg.get_payload() + return payload + # If our source is a bytes, then we're managing the encoding and we need + # to deal with it. + else: + bpayload: bytes = msg.get_payload(decode=True) + try: + return bpayload.decode("utf8", "strict") + except UnicodeDecodeError: + raise ValueError("payload in an invalid encoding") + + +# The various parse_FORMAT functions here are intended to be as lenient as +# possible in their parsing, while still returning a correctly typed +# RawMetadata. +# +# To aid in this, we also generally want to do as little touching of the +# data as possible, except where there are possibly some historic holdovers +# that make valid data awkward to work with. +# +# While this is a lower level, intermediate format than our ``Metadata`` +# class, some light touch ups can make a massive difference in usability. + +# Map METADATA fields to RawMetadata. +_EMAIL_TO_RAW_MAPPING = { + "author": "author", + "author-email": "author_email", + "classifier": "classifiers", + "description": "description", + "description-content-type": "description_content_type", + "download-url": "download_url", + "dynamic": "dynamic", + "home-page": "home_page", + "keywords": "keywords", + "license": "license", + "maintainer": "maintainer", + "maintainer-email": "maintainer_email", + "metadata-version": "metadata_version", + "name": "name", + "obsoletes": "obsoletes", + "obsoletes-dist": "obsoletes_dist", + "platform": "platforms", + "project-url": "project_urls", + "provides": "provides", + "provides-dist": "provides_dist", + "provides-extra": "provides_extra", + "requires": "requires", + "requires-dist": "requires_dist", + "requires-external": "requires_external", + "requires-python": "requires_python", + "summary": "summary", + "supported-platform": "supported_platforms", + "version": "version", +} +_RAW_TO_EMAIL_MAPPING = {raw: email for email, raw in _EMAIL_TO_RAW_MAPPING.items()} + + +def parse_email(data: Union[bytes, str]) -> Tuple[RawMetadata, Dict[str, List[str]]]: + """Parse a distribution's metadata stored as email headers (e.g. from ``METADATA``). + + This function returns a two-item tuple of dicts. The first dict is of + recognized fields from the core metadata specification. Fields that can be + parsed and translated into Python's built-in types are converted + appropriately. All other fields are left as-is. Fields that are allowed to + appear multiple times are stored as lists. + + The second dict contains all other fields from the metadata. This includes + any unrecognized fields. It also includes any fields which are expected to + be parsed into a built-in type but were not formatted appropriately. Finally, + any fields that are expected to appear only once but are repeated are + included in this dict. + + """ + raw: Dict[str, Union[str, List[str], Dict[str, str]]] = {} + unparsed: Dict[str, List[str]] = {} + + if isinstance(data, str): + parsed = email.parser.Parser(policy=email.policy.compat32).parsestr(data) + else: + parsed = email.parser.BytesParser(policy=email.policy.compat32).parsebytes(data) + + # We have to wrap parsed.keys() in a set, because in the case of multiple + # values for a key (a list), the key will appear multiple times in the + # list of keys, but we're avoiding that by using get_all(). + for name in frozenset(parsed.keys()): + # Header names in RFC are case insensitive, so we'll normalize to all + # lower case to make comparisons easier. + name = name.lower() + + # We use get_all() here, even for fields that aren't multiple use, + # because otherwise someone could have e.g. two Name fields, and we + # would just silently ignore it rather than doing something about it. + headers = parsed.get_all(name) or [] + + # The way the email module works when parsing bytes is that it + # unconditionally decodes the bytes as ascii using the surrogateescape + # handler. When you pull that data back out (such as with get_all() ), + # it looks to see if the str has any surrogate escapes, and if it does + # it wraps it in a Header object instead of returning the string. + # + # As such, we'll look for those Header objects, and fix up the encoding. + value = [] + # Flag if we have run into any issues processing the headers, thus + # signalling that the data belongs in 'unparsed'. + valid_encoding = True + for h in headers: + # It's unclear if this can return more types than just a Header or + # a str, so we'll just assert here to make sure. + assert isinstance(h, (email.header.Header, str)) + + # If it's a header object, we need to do our little dance to get + # the real data out of it. In cases where there is invalid data + # we're going to end up with mojibake, but there's no obvious, good + # way around that without reimplementing parts of the Header object + # ourselves. + # + # That should be fine since, if mojibacked happens, this key is + # going into the unparsed dict anyways. + if isinstance(h, email.header.Header): + # The Header object stores it's data as chunks, and each chunk + # can be independently encoded, so we'll need to check each + # of them. + chunks: List[Tuple[bytes, Optional[str]]] = [] + for bin, encoding in email.header.decode_header(h): + try: + bin.decode("utf8", "strict") + except UnicodeDecodeError: + # Enable mojibake. + encoding = "latin1" + valid_encoding = False + else: + encoding = "utf8" + chunks.append((bin, encoding)) + + # Turn our chunks back into a Header object, then let that + # Header object do the right thing to turn them into a + # string for us. + value.append(str(email.header.make_header(chunks))) + # This is already a string, so just add it. + else: + value.append(h) + + # We've processed all of our values to get them into a list of str, + # but we may have mojibake data, in which case this is an unparsed + # field. + if not valid_encoding: + unparsed[name] = value + continue + + raw_name = _EMAIL_TO_RAW_MAPPING.get(name) + if raw_name is None: + # This is a bit of a weird situation, we've encountered a key that + # we don't know what it means, so we don't know whether it's meant + # to be a list or not. + # + # Since we can't really tell one way or another, we'll just leave it + # as a list, even though it may be a single item list, because that's + # what makes the most sense for email headers. + unparsed[name] = value + continue + + # If this is one of our string fields, then we'll check to see if our + # value is a list of a single item. If it is then we'll assume that + # it was emitted as a single string, and unwrap the str from inside + # the list. + # + # If it's any other kind of data, then we haven't the faintest clue + # what we should parse it as, and we have to just add it to our list + # of unparsed stuff. + if raw_name in _STRING_FIELDS and len(value) == 1: + raw[raw_name] = value[0] + # If this is one of our list of string fields, then we can just assign + # the value, since email *only* has strings, and our get_all() call + # above ensures that this is a list. + elif raw_name in _LIST_FIELDS: + raw[raw_name] = value + # Special Case: Keywords + # The keywords field is implemented in the metadata spec as a str, + # but it conceptually is a list of strings, and is serialized using + # ", ".join(keywords), so we'll do some light data massaging to turn + # this into what it logically is. + elif raw_name == "keywords" and len(value) == 1: + raw[raw_name] = _parse_keywords(value[0]) + # Special Case: Project-URL + # The project urls is implemented in the metadata spec as a list of + # specially-formatted strings that represent a key and a value, which + # is fundamentally a mapping, however the email format doesn't support + # mappings in a sane way, so it was crammed into a list of strings + # instead. + # + # We will do a little light data massaging to turn this into a map as + # it logically should be. + elif raw_name == "project_urls": + try: + raw[raw_name] = _parse_project_urls(value) + except KeyError: + unparsed[name] = value + # Nothing that we've done has managed to parse this, so it'll just + # throw it in our unparseable data and move on. + else: + unparsed[name] = value + + # We need to support getting the Description from the message payload in + # addition to getting it from the the headers. This does mean, though, there + # is the possibility of it being set both ways, in which case we put both + # in 'unparsed' since we don't know which is right. + try: + payload = _get_payload(parsed, data) + except ValueError: + unparsed.setdefault("description", []).append( + parsed.get_payload(decode=isinstance(data, bytes)) + ) + else: + if payload: + # Check to see if we've already got a description, if so then both + # it, and this body move to unparseable. + if "description" in raw: + description_header = cast(str, raw.pop("description")) + unparsed.setdefault("description", []).extend( + [description_header, payload] + ) + elif "description" in unparsed: + unparsed["description"].append(payload) + else: + raw["description"] = payload + + # We need to cast our `raw` to a metadata, because a TypedDict only support + # literal key names, but we're computing our key names on purpose, but the + # way this function is implemented, our `TypedDict` can only have valid key + # names. + return cast(RawMetadata, raw), unparsed + + +_NOT_FOUND = object() + + +# Keep the two values in sync. +_VALID_METADATA_VERSIONS = ["1.0", "1.1", "1.2", "2.1", "2.2", "2.3"] +_MetadataVersion = Literal["1.0", "1.1", "1.2", "2.1", "2.2", "2.3"] + +_REQUIRED_ATTRS = frozenset(["metadata_version", "name", "version"]) + + +class _Validator(Generic[T]): + """Validate a metadata field. + + All _process_*() methods correspond to a core metadata field. The method is + called with the field's raw value. If the raw value is valid it is returned + in its "enriched" form (e.g. ``version.Version`` for the ``Version`` field). + If the raw value is invalid, :exc:`InvalidMetadata` is raised (with a cause + as appropriate). + """ + + name: str + raw_name: str + added: _MetadataVersion + + def __init__( + self, + *, + added: _MetadataVersion = "1.0", + ) -> None: + self.added = added + + def __set_name__(self, _owner: "Metadata", name: str) -> None: + self.name = name + self.raw_name = _RAW_TO_EMAIL_MAPPING[name] + + def __get__(self, instance: "Metadata", _owner: Type["Metadata"]) -> T: + # With Python 3.8, the caching can be replaced with functools.cached_property(). + # No need to check the cache as attribute lookup will resolve into the + # instance's __dict__ before __get__ is called. + cache = instance.__dict__ + value = instance._raw.get(self.name) + + # To make the _process_* methods easier, we'll check if the value is None + # and if this field is NOT a required attribute, and if both of those + # things are true, we'll skip the the converter. This will mean that the + # converters never have to deal with the None union. + if self.name in _REQUIRED_ATTRS or value is not None: + try: + converter: Callable[[Any], T] = getattr(self, f"_process_{self.name}") + except AttributeError: + pass + else: + value = converter(value) + + cache[self.name] = value + try: + del instance._raw[self.name] # type: ignore[misc] + except KeyError: + pass + + return cast(T, value) + + def _invalid_metadata( + self, msg: str, cause: Optional[Exception] = None + ) -> InvalidMetadata: + exc = InvalidMetadata( + self.raw_name, msg.format_map({"field": repr(self.raw_name)}) + ) + exc.__cause__ = cause + return exc + + def _process_metadata_version(self, value: str) -> _MetadataVersion: + # Implicitly makes Metadata-Version required. + if value not in _VALID_METADATA_VERSIONS: + raise self._invalid_metadata(f"{value!r} is not a valid metadata version") + return cast(_MetadataVersion, value) + + def _process_name(self, value: str) -> str: + if not value: + raise self._invalid_metadata("{field} is a required field") + # Validate the name as a side-effect. + try: + utils.canonicalize_name(value, validate=True) + except utils.InvalidName as exc: + raise self._invalid_metadata( + f"{value!r} is invalid for {{field}}", cause=exc + ) + else: + return value + + def _process_version(self, value: str) -> version_module.Version: + if not value: + raise self._invalid_metadata("{field} is a required field") + try: + return version_module.parse(value) + except version_module.InvalidVersion as exc: + raise self._invalid_metadata( + f"{value!r} is invalid for {{field}}", cause=exc + ) + + def _process_summary(self, value: str) -> str: + """Check the field contains no newlines.""" + if "\n" in value: + raise self._invalid_metadata("{field} must be a single line") + return value + + def _process_description_content_type(self, value: str) -> str: + content_types = {"text/plain", "text/x-rst", "text/markdown"} + message = email.message.EmailMessage() + message["content-type"] = value + + content_type, parameters = ( + # Defaults to `text/plain` if parsing failed. + message.get_content_type().lower(), + message["content-type"].params, + ) + # Check if content-type is valid or defaulted to `text/plain` and thus was + # not parseable. + if content_type not in content_types or content_type not in value.lower(): + raise self._invalid_metadata( + f"{{field}} must be one of {list(content_types)}, not {value!r}" + ) + + charset = parameters.get("charset", "UTF-8") + if charset != "UTF-8": + raise self._invalid_metadata( + f"{{field}} can only specify the UTF-8 charset, not {list(charset)}" + ) + + markdown_variants = {"GFM", "CommonMark"} + variant = parameters.get("variant", "GFM") # Use an acceptable default. + if content_type == "text/markdown" and variant not in markdown_variants: + raise self._invalid_metadata( + f"valid Markdown variants for {{field}} are {list(markdown_variants)}, " + f"not {variant!r}", + ) + return value + + def _process_dynamic(self, value: List[str]) -> List[str]: + for dynamic_field in map(str.lower, value): + if dynamic_field in {"name", "version", "metadata-version"}: + raise self._invalid_metadata( + f"{value!r} is not allowed as a dynamic field" + ) + elif dynamic_field not in _EMAIL_TO_RAW_MAPPING: + raise self._invalid_metadata(f"{value!r} is not a valid dynamic field") + return list(map(str.lower, value)) + + def _process_provides_extra( + self, + value: List[str], + ) -> List[utils.NormalizedName]: + normalized_names = [] + try: + for name in value: + normalized_names.append(utils.canonicalize_name(name, validate=True)) + except utils.InvalidName as exc: + raise self._invalid_metadata( + f"{name!r} is invalid for {{field}}", cause=exc + ) + else: + return normalized_names + + def _process_requires_python(self, value: str) -> specifiers.SpecifierSet: + try: + return specifiers.SpecifierSet(value) + except specifiers.InvalidSpecifier as exc: + raise self._invalid_metadata( + f"{value!r} is invalid for {{field}}", cause=exc + ) + + def _process_requires_dist( + self, + value: List[str], + ) -> List[requirements.Requirement]: + reqs = [] + try: + for req in value: + reqs.append(requirements.Requirement(req)) + except requirements.InvalidRequirement as exc: + raise self._invalid_metadata(f"{req!r} is invalid for {{field}}", cause=exc) + else: + return reqs + + +class Metadata: + """Representation of distribution metadata. + + Compared to :class:`RawMetadata`, this class provides objects representing + metadata fields instead of only using built-in types. Any invalid metadata + will cause :exc:`InvalidMetadata` to be raised (with a + :py:attr:`~BaseException.__cause__` attribute as appropriate). + """ + + _raw: RawMetadata + + @classmethod + def from_raw(cls, data: RawMetadata, *, validate: bool = True) -> "Metadata": + """Create an instance from :class:`RawMetadata`. + + If *validate* is true, all metadata will be validated. All exceptions + related to validation will be gathered and raised as an :class:`ExceptionGroup`. + """ + ins = cls() + ins._raw = data.copy() # Mutations occur due to caching enriched values. + + if validate: + exceptions: List[Exception] = [] + try: + metadata_version = ins.metadata_version + metadata_age = _VALID_METADATA_VERSIONS.index(metadata_version) + except InvalidMetadata as metadata_version_exc: + exceptions.append(metadata_version_exc) + metadata_version = None + + # Make sure to check for the fields that are present, the required + # fields (so their absence can be reported). + fields_to_check = frozenset(ins._raw) | _REQUIRED_ATTRS + # Remove fields that have already been checked. + fields_to_check -= {"metadata_version"} + + for key in fields_to_check: + try: + if metadata_version: + # Can't use getattr() as that triggers descriptor protocol which + # will fail due to no value for the instance argument. + try: + field_metadata_version = cls.__dict__[key].added + except KeyError: + exc = InvalidMetadata(key, f"unrecognized field: {key!r}") + exceptions.append(exc) + continue + field_age = _VALID_METADATA_VERSIONS.index( + field_metadata_version + ) + if field_age > metadata_age: + field = _RAW_TO_EMAIL_MAPPING[key] + exc = InvalidMetadata( + field, + "{field} introduced in metadata version " + "{field_metadata_version}, not {metadata_version}", + ) + exceptions.append(exc) + continue + getattr(ins, key) + except InvalidMetadata as exc: + exceptions.append(exc) + + if exceptions: + raise ExceptionGroup("invalid metadata", exceptions) + + return ins + + @classmethod + def from_email( + cls, data: Union[bytes, str], *, validate: bool = True + ) -> "Metadata": + """Parse metadata from email headers. + + If *validate* is true, the metadata will be validated. All exceptions + related to validation will be gathered and raised as an :class:`ExceptionGroup`. + """ + raw, unparsed = parse_email(data) + + if validate: + exceptions: list[Exception] = [] + for unparsed_key in unparsed: + if unparsed_key in _EMAIL_TO_RAW_MAPPING: + message = f"{unparsed_key!r} has invalid data" + else: + message = f"unrecognized field: {unparsed_key!r}" + exceptions.append(InvalidMetadata(unparsed_key, message)) + + if exceptions: + raise ExceptionGroup("unparsed", exceptions) + + try: + return cls.from_raw(raw, validate=validate) + except ExceptionGroup as exc_group: + raise ExceptionGroup( + "invalid or unparsed metadata", exc_group.exceptions + ) from None + + metadata_version: _Validator[_MetadataVersion] = _Validator() + """:external:ref:`core-metadata-metadata-version` + (required; validated to be a valid metadata version)""" + name: _Validator[str] = _Validator() + """:external:ref:`core-metadata-name` + (required; validated using :func:`~packaging.utils.canonicalize_name` and its + *validate* parameter)""" + version: _Validator[version_module.Version] = _Validator() + """:external:ref:`core-metadata-version` (required)""" + dynamic: _Validator[Optional[List[str]]] = _Validator( + added="2.2", + ) + """:external:ref:`core-metadata-dynamic` + (validated against core metadata field names and lowercased)""" + platforms: _Validator[Optional[List[str]]] = _Validator() + """:external:ref:`core-metadata-platform`""" + supported_platforms: _Validator[Optional[List[str]]] = _Validator(added="1.1") + """:external:ref:`core-metadata-supported-platform`""" + summary: _Validator[Optional[str]] = _Validator() + """:external:ref:`core-metadata-summary` (validated to contain no newlines)""" + description: _Validator[Optional[str]] = _Validator() # TODO 2.1: can be in body + """:external:ref:`core-metadata-description`""" + description_content_type: _Validator[Optional[str]] = _Validator(added="2.1") + """:external:ref:`core-metadata-description-content-type` (validated)""" + keywords: _Validator[Optional[List[str]]] = _Validator() + """:external:ref:`core-metadata-keywords`""" + home_page: _Validator[Optional[str]] = _Validator() + """:external:ref:`core-metadata-home-page`""" + download_url: _Validator[Optional[str]] = _Validator(added="1.1") + """:external:ref:`core-metadata-download-url`""" + author: _Validator[Optional[str]] = _Validator() + """:external:ref:`core-metadata-author`""" + author_email: _Validator[Optional[str]] = _Validator() + """:external:ref:`core-metadata-author-email`""" + maintainer: _Validator[Optional[str]] = _Validator(added="1.2") + """:external:ref:`core-metadata-maintainer`""" + maintainer_email: _Validator[Optional[str]] = _Validator(added="1.2") + """:external:ref:`core-metadata-maintainer-email`""" + license: _Validator[Optional[str]] = _Validator() + """:external:ref:`core-metadata-license`""" + classifiers: _Validator[Optional[List[str]]] = _Validator(added="1.1") + """:external:ref:`core-metadata-classifier`""" + requires_dist: _Validator[Optional[List[requirements.Requirement]]] = _Validator( + added="1.2" + ) + """:external:ref:`core-metadata-requires-dist`""" + requires_python: _Validator[Optional[specifiers.SpecifierSet]] = _Validator( + added="1.2" + ) + """:external:ref:`core-metadata-requires-python`""" + # Because `Requires-External` allows for non-PEP 440 version specifiers, we + # don't do any processing on the values. + requires_external: _Validator[Optional[List[str]]] = _Validator(added="1.2") + """:external:ref:`core-metadata-requires-external`""" + project_urls: _Validator[Optional[Dict[str, str]]] = _Validator(added="1.2") + """:external:ref:`core-metadata-project-url`""" + # PEP 685 lets us raise an error if an extra doesn't pass `Name` validation + # regardless of metadata version. + provides_extra: _Validator[Optional[List[utils.NormalizedName]]] = _Validator( + added="2.1", + ) + """:external:ref:`core-metadata-provides-extra`""" + provides_dist: _Validator[Optional[List[str]]] = _Validator(added="1.2") + """:external:ref:`core-metadata-provides-dist`""" + obsoletes_dist: _Validator[Optional[List[str]]] = _Validator(added="1.2") + """:external:ref:`core-metadata-obsoletes-dist`""" + requires: _Validator[Optional[List[str]]] = _Validator(added="1.1") + """``Requires`` (deprecated)""" + provides: _Validator[Optional[List[str]]] = _Validator(added="1.1") + """``Provides`` (deprecated)""" + obsoletes: _Validator[Optional[List[str]]] = _Validator(added="1.1") + """``Obsoletes`` (deprecated)""" diff --git a/pylib/packaging/py.typed b/pylib/packaging/py.typed new file mode 100644 index 00000000..e69de29b diff --git a/pylib/packaging/requirements.py b/pylib/packaging/requirements.py new file mode 100644 index 00000000..0c00eba3 --- /dev/null +++ b/pylib/packaging/requirements.py @@ -0,0 +1,90 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + +from typing import Any, Iterator, Optional, Set + +from ._parser import parse_requirement as _parse_requirement +from ._tokenizer import ParserSyntaxError +from .markers import Marker, _normalize_extra_values +from .specifiers import SpecifierSet +from .utils import canonicalize_name + + +class InvalidRequirement(ValueError): + """ + An invalid requirement was found, users should refer to PEP 508. + """ + + +class Requirement: + """Parse a requirement. + + Parse a given requirement string into its parts, such as name, specifier, + URL, and extras. Raises InvalidRequirement on a badly-formed requirement + string. + """ + + # TODO: Can we test whether something is contained within a requirement? + # If so how do we do that? Do we need to test against the _name_ of + # the thing as well as the version? What about the markers? + # TODO: Can we normalize the name and extra name? + + def __init__(self, requirement_string: str) -> None: + try: + parsed = _parse_requirement(requirement_string) + except ParserSyntaxError as e: + raise InvalidRequirement(str(e)) from e + + self.name: str = parsed.name + self.url: Optional[str] = parsed.url or None + self.extras: Set[str] = set(parsed.extras if parsed.extras else []) + self.specifier: SpecifierSet = SpecifierSet(parsed.specifier) + self.marker: Optional[Marker] = None + if parsed.marker is not None: + self.marker = Marker.__new__(Marker) + self.marker._markers = _normalize_extra_values(parsed.marker) + + def _iter_parts(self, name: str) -> Iterator[str]: + yield name + + if self.extras: + formatted_extras = ",".join(sorted(self.extras)) + yield f"[{formatted_extras}]" + + if self.specifier: + yield str(self.specifier) + + if self.url: + yield f"@ {self.url}" + if self.marker: + yield " " + + if self.marker: + yield f"; {self.marker}" + + def __str__(self) -> str: + return "".join(self._iter_parts(self.name)) + + def __repr__(self) -> str: + return f"" + + def __hash__(self) -> int: + return hash( + ( + self.__class__.__name__, + *self._iter_parts(canonicalize_name(self.name)), + ) + ) + + def __eq__(self, other: Any) -> bool: + if not isinstance(other, Requirement): + return NotImplemented + + return ( + canonicalize_name(self.name) == canonicalize_name(other.name) + and self.extras == other.extras + and self.specifier == other.specifier + and self.url == other.url + and self.marker == other.marker + ) diff --git a/pylib/packaging/specifiers.py b/pylib/packaging/specifiers.py new file mode 100644 index 00000000..94448327 --- /dev/null +++ b/pylib/packaging/specifiers.py @@ -0,0 +1,1030 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. +""" +.. testsetup:: + + from packaging.specifiers import Specifier, SpecifierSet, InvalidSpecifier + from packaging.version import Version +""" + +import abc +import itertools +import re +from typing import ( + Callable, + Iterable, + Iterator, + List, + Optional, + Set, + Tuple, + TypeVar, + Union, +) + +from .utils import canonicalize_version +from .version import Version + +UnparsedVersion = Union[Version, str] +UnparsedVersionVar = TypeVar("UnparsedVersionVar", bound=UnparsedVersion) +CallableOperator = Callable[[Version, str], bool] + + +def _coerce_version(version: UnparsedVersion) -> Version: + if not isinstance(version, Version): + version = Version(version) + return version + + +class InvalidSpecifier(ValueError): + """ + Raised when attempting to create a :class:`Specifier` with a specifier + string that is invalid. + + >>> Specifier("lolwat") + Traceback (most recent call last): + ... + packaging.specifiers.InvalidSpecifier: Invalid specifier: 'lolwat' + """ + + +class BaseSpecifier(metaclass=abc.ABCMeta): + @abc.abstractmethod + def __str__(self) -> str: + """ + Returns the str representation of this Specifier-like object. This + should be representative of the Specifier itself. + """ + + @abc.abstractmethod + def __hash__(self) -> int: + """ + Returns a hash value for this Specifier-like object. + """ + + @abc.abstractmethod + def __eq__(self, other: object) -> bool: + """ + Returns a boolean representing whether or not the two Specifier-like + objects are equal. + + :param other: The other object to check against. + """ + + @property + @abc.abstractmethod + def prereleases(self) -> Optional[bool]: + """Whether or not pre-releases as a whole are allowed. + + This can be set to either ``True`` or ``False`` to explicitly enable or disable + prereleases or it can be set to ``None`` (the default) to use default semantics. + """ + + @prereleases.setter + def prereleases(self, value: bool) -> None: + """Setter for :attr:`prereleases`. + + :param value: The value to set. + """ + + @abc.abstractmethod + def contains(self, item: str, prereleases: Optional[bool] = None) -> bool: + """ + Determines if the given item is contained within this specifier. + """ + + @abc.abstractmethod + def filter( + self, iterable: Iterable[UnparsedVersionVar], prereleases: Optional[bool] = None + ) -> Iterator[UnparsedVersionVar]: + """ + Takes an iterable of items and filters them so that only items which + are contained within this specifier are allowed in it. + """ + + +class Specifier(BaseSpecifier): + """This class abstracts handling of version specifiers. + + .. tip:: + + It is generally not required to instantiate this manually. You should instead + prefer to work with :class:`SpecifierSet` instead, which can parse + comma-separated version specifiers (which is what package metadata contains). + """ + + _operator_regex_str = r""" + (?P(~=|==|!=|<=|>=|<|>|===)) + """ + _version_regex_str = r""" + (?P + (?: + # The identity operators allow for an escape hatch that will + # do an exact string match of the version you wish to install. + # This will not be parsed by PEP 440 and we cannot determine + # any semantic meaning from it. This operator is discouraged + # but included entirely as an escape hatch. + (?<====) # Only match for the identity operator + \s* + [^\s;)]* # The arbitrary version can be just about anything, + # we match everything except for whitespace, a + # semi-colon for marker support, and a closing paren + # since versions can be enclosed in them. + ) + | + (?: + # The (non)equality operators allow for wild card and local + # versions to be specified so we have to define these two + # operators separately to enable that. + (?<===|!=) # Only match for equals and not equals + + \s* + v? + (?:[0-9]+!)? # epoch + [0-9]+(?:\.[0-9]+)* # release + + # You cannot use a wild card and a pre-release, post-release, a dev or + # local version together so group them with a | and make them optional. + (?: + \.\* # Wild card syntax of .* + | + (?: # pre release + [-_\.]? + (alpha|beta|preview|pre|a|b|c|rc) + [-_\.]? + [0-9]* + )? + (?: # post release + (?:-[0-9]+)|(?:[-_\.]?(post|rev|r)[-_\.]?[0-9]*) + )? + (?:[-_\.]?dev[-_\.]?[0-9]*)? # dev release + (?:\+[a-z0-9]+(?:[-_\.][a-z0-9]+)*)? # local + )? + ) + | + (?: + # The compatible operator requires at least two digits in the + # release segment. + (?<=~=) # Only match for the compatible operator + + \s* + v? + (?:[0-9]+!)? # epoch + [0-9]+(?:\.[0-9]+)+ # release (We have a + instead of a *) + (?: # pre release + [-_\.]? + (alpha|beta|preview|pre|a|b|c|rc) + [-_\.]? + [0-9]* + )? + (?: # post release + (?:-[0-9]+)|(?:[-_\.]?(post|rev|r)[-_\.]?[0-9]*) + )? + (?:[-_\.]?dev[-_\.]?[0-9]*)? # dev release + ) + | + (?: + # All other operators only allow a sub set of what the + # (non)equality operators do. Specifically they do not allow + # local versions to be specified nor do they allow the prefix + # matching wild cards. + (?=": "greater_than_equal", + "<": "less_than", + ">": "greater_than", + "===": "arbitrary", + } + + def __init__(self, spec: str = "", prereleases: Optional[bool] = None) -> None: + """Initialize a Specifier instance. + + :param spec: + The string representation of a specifier which will be parsed and + normalized before use. + :param prereleases: + This tells the specifier if it should accept prerelease versions if + applicable or not. The default of ``None`` will autodetect it from the + given specifiers. + :raises InvalidSpecifier: + If the given specifier is invalid (i.e. bad syntax). + """ + match = self._regex.search(spec) + if not match: + raise InvalidSpecifier(f"Invalid specifier: '{spec}'") + + self._spec: Tuple[str, str] = ( + match.group("operator").strip(), + match.group("version").strip(), + ) + + # Store whether or not this Specifier should accept prereleases + self._prereleases = prereleases + + # https://github.com/python/mypy/pull/13475#pullrequestreview-1079784515 + @property # type: ignore[override] + def prereleases(self) -> bool: + # If there is an explicit prereleases set for this, then we'll just + # blindly use that. + if self._prereleases is not None: + return self._prereleases + + # Look at all of our specifiers and determine if they are inclusive + # operators, and if they are if they are including an explicit + # prerelease. + operator, version = self._spec + if operator in ["==", ">=", "<=", "~=", "==="]: + # The == specifier can include a trailing .*, if it does we + # want to remove before parsing. + if operator == "==" and version.endswith(".*"): + version = version[:-2] + + # Parse the version, and if it is a pre-release than this + # specifier allows pre-releases. + if Version(version).is_prerelease: + return True + + return False + + @prereleases.setter + def prereleases(self, value: bool) -> None: + self._prereleases = value + + @property + def operator(self) -> str: + """The operator of this specifier. + + >>> Specifier("==1.2.3").operator + '==' + """ + return self._spec[0] + + @property + def version(self) -> str: + """The version of this specifier. + + >>> Specifier("==1.2.3").version + '1.2.3' + """ + return self._spec[1] + + def __repr__(self) -> str: + """A representation of the Specifier that shows all internal state. + + >>> Specifier('>=1.0.0') + =1.0.0')> + >>> Specifier('>=1.0.0', prereleases=False) + =1.0.0', prereleases=False)> + >>> Specifier('>=1.0.0', prereleases=True) + =1.0.0', prereleases=True)> + """ + pre = ( + f", prereleases={self.prereleases!r}" + if self._prereleases is not None + else "" + ) + + return f"<{self.__class__.__name__}({str(self)!r}{pre})>" + + def __str__(self) -> str: + """A string representation of the Specifier that can be round-tripped. + + >>> str(Specifier('>=1.0.0')) + '>=1.0.0' + >>> str(Specifier('>=1.0.0', prereleases=False)) + '>=1.0.0' + """ + return "{}{}".format(*self._spec) + + @property + def _canonical_spec(self) -> Tuple[str, str]: + canonical_version = canonicalize_version( + self._spec[1], + strip_trailing_zero=(self._spec[0] != "~="), + ) + return self._spec[0], canonical_version + + def __hash__(self) -> int: + return hash(self._canonical_spec) + + def __eq__(self, other: object) -> bool: + """Whether or not the two Specifier-like objects are equal. + + :param other: The other object to check against. + + The value of :attr:`prereleases` is ignored. + + >>> Specifier("==1.2.3") == Specifier("== 1.2.3.0") + True + >>> (Specifier("==1.2.3", prereleases=False) == + ... Specifier("==1.2.3", prereleases=True)) + True + >>> Specifier("==1.2.3") == "==1.2.3" + True + >>> Specifier("==1.2.3") == Specifier("==1.2.4") + False + >>> Specifier("==1.2.3") == Specifier("~=1.2.3") + False + """ + if isinstance(other, str): + try: + other = self.__class__(str(other)) + except InvalidSpecifier: + return NotImplemented + elif not isinstance(other, self.__class__): + return NotImplemented + + return self._canonical_spec == other._canonical_spec + + def _get_operator(self, op: str) -> CallableOperator: + operator_callable: CallableOperator = getattr( + self, f"_compare_{self._operators[op]}" + ) + return operator_callable + + def _compare_compatible(self, prospective: Version, spec: str) -> bool: + + # Compatible releases have an equivalent combination of >= and ==. That + # is that ~=2.2 is equivalent to >=2.2,==2.*. This allows us to + # implement this in terms of the other specifiers instead of + # implementing it ourselves. The only thing we need to do is construct + # the other specifiers. + + # We want everything but the last item in the version, but we want to + # ignore suffix segments. + prefix = _version_join( + list(itertools.takewhile(_is_not_suffix, _version_split(spec)))[:-1] + ) + + # Add the prefix notation to the end of our string + prefix += ".*" + + return self._get_operator(">=")(prospective, spec) and self._get_operator("==")( + prospective, prefix + ) + + def _compare_equal(self, prospective: Version, spec: str) -> bool: + + # We need special logic to handle prefix matching + if spec.endswith(".*"): + # In the case of prefix matching we want to ignore local segment. + normalized_prospective = canonicalize_version( + prospective.public, strip_trailing_zero=False + ) + # Get the normalized version string ignoring the trailing .* + normalized_spec = canonicalize_version(spec[:-2], strip_trailing_zero=False) + # Split the spec out by bangs and dots, and pretend that there is + # an implicit dot in between a release segment and a pre-release segment. + split_spec = _version_split(normalized_spec) + + # Split the prospective version out by bangs and dots, and pretend + # that there is an implicit dot in between a release segment and + # a pre-release segment. + split_prospective = _version_split(normalized_prospective) + + # 0-pad the prospective version before shortening it to get the correct + # shortened version. + padded_prospective, _ = _pad_version(split_prospective, split_spec) + + # Shorten the prospective version to be the same length as the spec + # so that we can determine if the specifier is a prefix of the + # prospective version or not. + shortened_prospective = padded_prospective[: len(split_spec)] + + return shortened_prospective == split_spec + else: + # Convert our spec string into a Version + spec_version = Version(spec) + + # If the specifier does not have a local segment, then we want to + # act as if the prospective version also does not have a local + # segment. + if not spec_version.local: + prospective = Version(prospective.public) + + return prospective == spec_version + + def _compare_not_equal(self, prospective: Version, spec: str) -> bool: + return not self._compare_equal(prospective, spec) + + def _compare_less_than_equal(self, prospective: Version, spec: str) -> bool: + + # NB: Local version identifiers are NOT permitted in the version + # specifier, so local version labels can be universally removed from + # the prospective version. + return Version(prospective.public) <= Version(spec) + + def _compare_greater_than_equal(self, prospective: Version, spec: str) -> bool: + + # NB: Local version identifiers are NOT permitted in the version + # specifier, so local version labels can be universally removed from + # the prospective version. + return Version(prospective.public) >= Version(spec) + + def _compare_less_than(self, prospective: Version, spec_str: str) -> bool: + + # Convert our spec to a Version instance, since we'll want to work with + # it as a version. + spec = Version(spec_str) + + # Check to see if the prospective version is less than the spec + # version. If it's not we can short circuit and just return False now + # instead of doing extra unneeded work. + if not prospective < spec: + return False + + # This special case is here so that, unless the specifier itself + # includes is a pre-release version, that we do not accept pre-release + # versions for the version mentioned in the specifier (e.g. <3.1 should + # not match 3.1.dev0, but should match 3.0.dev0). + if not spec.is_prerelease and prospective.is_prerelease: + if Version(prospective.base_version) == Version(spec.base_version): + return False + + # If we've gotten to here, it means that prospective version is both + # less than the spec version *and* it's not a pre-release of the same + # version in the spec. + return True + + def _compare_greater_than(self, prospective: Version, spec_str: str) -> bool: + + # Convert our spec to a Version instance, since we'll want to work with + # it as a version. + spec = Version(spec_str) + + # Check to see if the prospective version is greater than the spec + # version. If it's not we can short circuit and just return False now + # instead of doing extra unneeded work. + if not prospective > spec: + return False + + # This special case is here so that, unless the specifier itself + # includes is a post-release version, that we do not accept + # post-release versions for the version mentioned in the specifier + # (e.g. >3.1 should not match 3.0.post0, but should match 3.2.post0). + if not spec.is_postrelease and prospective.is_postrelease: + if Version(prospective.base_version) == Version(spec.base_version): + return False + + # Ensure that we do not allow a local version of the version mentioned + # in the specifier, which is technically greater than, to match. + if prospective.local is not None: + if Version(prospective.base_version) == Version(spec.base_version): + return False + + # If we've gotten to here, it means that prospective version is both + # greater than the spec version *and* it's not a pre-release of the + # same version in the spec. + return True + + def _compare_arbitrary(self, prospective: Version, spec: str) -> bool: + return str(prospective).lower() == str(spec).lower() + + def __contains__(self, item: Union[str, Version]) -> bool: + """Return whether or not the item is contained in this specifier. + + :param item: The item to check for. + + This is used for the ``in`` operator and behaves the same as + :meth:`contains` with no ``prereleases`` argument passed. + + >>> "1.2.3" in Specifier(">=1.2.3") + True + >>> Version("1.2.3") in Specifier(">=1.2.3") + True + >>> "1.0.0" in Specifier(">=1.2.3") + False + >>> "1.3.0a1" in Specifier(">=1.2.3") + False + >>> "1.3.0a1" in Specifier(">=1.2.3", prereleases=True) + True + """ + return self.contains(item) + + def contains( + self, item: UnparsedVersion, prereleases: Optional[bool] = None + ) -> bool: + """Return whether or not the item is contained in this specifier. + + :param item: + The item to check for, which can be a version string or a + :class:`Version` instance. + :param prereleases: + Whether or not to match prereleases with this Specifier. If set to + ``None`` (the default), it uses :attr:`prereleases` to determine + whether or not prereleases are allowed. + + >>> Specifier(">=1.2.3").contains("1.2.3") + True + >>> Specifier(">=1.2.3").contains(Version("1.2.3")) + True + >>> Specifier(">=1.2.3").contains("1.0.0") + False + >>> Specifier(">=1.2.3").contains("1.3.0a1") + False + >>> Specifier(">=1.2.3", prereleases=True).contains("1.3.0a1") + True + >>> Specifier(">=1.2.3").contains("1.3.0a1", prereleases=True) + True + """ + + # Determine if prereleases are to be allowed or not. + if prereleases is None: + prereleases = self.prereleases + + # Normalize item to a Version, this allows us to have a shortcut for + # "2.0" in Specifier(">=2") + normalized_item = _coerce_version(item) + + # Determine if we should be supporting prereleases in this specifier + # or not, if we do not support prereleases than we can short circuit + # logic if this version is a prereleases. + if normalized_item.is_prerelease and not prereleases: + return False + + # Actually do the comparison to determine if this item is contained + # within this Specifier or not. + operator_callable: CallableOperator = self._get_operator(self.operator) + return operator_callable(normalized_item, self.version) + + def filter( + self, iterable: Iterable[UnparsedVersionVar], prereleases: Optional[bool] = None + ) -> Iterator[UnparsedVersionVar]: + """Filter items in the given iterable, that match the specifier. + + :param iterable: + An iterable that can contain version strings and :class:`Version` instances. + The items in the iterable will be filtered according to the specifier. + :param prereleases: + Whether or not to allow prereleases in the returned iterator. If set to + ``None`` (the default), it will be intelligently decide whether to allow + prereleases or not (based on the :attr:`prereleases` attribute, and + whether the only versions matching are prereleases). + + This method is smarter than just ``filter(Specifier().contains, [...])`` + because it implements the rule from :pep:`440` that a prerelease item + SHOULD be accepted if no other versions match the given specifier. + + >>> list(Specifier(">=1.2.3").filter(["1.2", "1.3", "1.5a1"])) + ['1.3'] + >>> list(Specifier(">=1.2.3").filter(["1.2", "1.2.3", "1.3", Version("1.4")])) + ['1.2.3', '1.3', ] + >>> list(Specifier(">=1.2.3").filter(["1.2", "1.5a1"])) + ['1.5a1'] + >>> list(Specifier(">=1.2.3").filter(["1.3", "1.5a1"], prereleases=True)) + ['1.3', '1.5a1'] + >>> list(Specifier(">=1.2.3", prereleases=True).filter(["1.3", "1.5a1"])) + ['1.3', '1.5a1'] + """ + + yielded = False + found_prereleases = [] + + kw = {"prereleases": prereleases if prereleases is not None else True} + + # Attempt to iterate over all the values in the iterable and if any of + # them match, yield them. + for version in iterable: + parsed_version = _coerce_version(version) + + if self.contains(parsed_version, **kw): + # If our version is a prerelease, and we were not set to allow + # prereleases, then we'll store it for later in case nothing + # else matches this specifier. + if parsed_version.is_prerelease and not ( + prereleases or self.prereleases + ): + found_prereleases.append(version) + # Either this is not a prerelease, or we should have been + # accepting prereleases from the beginning. + else: + yielded = True + yield version + + # Now that we've iterated over everything, determine if we've yielded + # any values, and if we have not and we have any prereleases stored up + # then we will go ahead and yield the prereleases. + if not yielded and found_prereleases: + for version in found_prereleases: + yield version + + +_prefix_regex = re.compile(r"^([0-9]+)((?:a|b|c|rc)[0-9]+)$") + + +def _version_split(version: str) -> List[str]: + """Split version into components. + + The split components are intended for version comparison. The logic does + not attempt to retain the original version string, so joining the + components back with :func:`_version_join` may not produce the original + version string. + """ + result: List[str] = [] + + epoch, _, rest = version.rpartition("!") + result.append(epoch or "0") + + for item in rest.split("."): + match = _prefix_regex.search(item) + if match: + result.extend(match.groups()) + else: + result.append(item) + return result + + +def _version_join(components: List[str]) -> str: + """Join split version components into a version string. + + This function assumes the input came from :func:`_version_split`, where the + first component must be the epoch (either empty or numeric), and all other + components numeric. + """ + epoch, *rest = components + return f"{epoch}!{'.'.join(rest)}" + + +def _is_not_suffix(segment: str) -> bool: + return not any( + segment.startswith(prefix) for prefix in ("dev", "a", "b", "rc", "post") + ) + + +def _pad_version(left: List[str], right: List[str]) -> Tuple[List[str], List[str]]: + left_split, right_split = [], [] + + # Get the release segment of our versions + left_split.append(list(itertools.takewhile(lambda x: x.isdigit(), left))) + right_split.append(list(itertools.takewhile(lambda x: x.isdigit(), right))) + + # Get the rest of our versions + left_split.append(left[len(left_split[0]) :]) + right_split.append(right[len(right_split[0]) :]) + + # Insert our padding + left_split.insert(1, ["0"] * max(0, len(right_split[0]) - len(left_split[0]))) + right_split.insert(1, ["0"] * max(0, len(left_split[0]) - len(right_split[0]))) + + return (list(itertools.chain(*left_split)), list(itertools.chain(*right_split))) + + +class SpecifierSet(BaseSpecifier): + """This class abstracts handling of a set of version specifiers. + + It can be passed a single specifier (``>=3.0``), a comma-separated list of + specifiers (``>=3.0,!=3.1``), or no specifier at all. + """ + + def __init__( + self, specifiers: str = "", prereleases: Optional[bool] = None + ) -> None: + """Initialize a SpecifierSet instance. + + :param specifiers: + The string representation of a specifier or a comma-separated list of + specifiers which will be parsed and normalized before use. + :param prereleases: + This tells the SpecifierSet if it should accept prerelease versions if + applicable or not. The default of ``None`` will autodetect it from the + given specifiers. + + :raises InvalidSpecifier: + If the given ``specifiers`` are not parseable than this exception will be + raised. + """ + + # Split on `,` to break each individual specifier into it's own item, and + # strip each item to remove leading/trailing whitespace. + split_specifiers = [s.strip() for s in specifiers.split(",") if s.strip()] + + # Parsed each individual specifier, attempting first to make it a + # Specifier. + parsed: Set[Specifier] = set() + for specifier in split_specifiers: + parsed.add(Specifier(specifier)) + + # Turn our parsed specifiers into a frozen set and save them for later. + self._specs = frozenset(parsed) + + # Store our prereleases value so we can use it later to determine if + # we accept prereleases or not. + self._prereleases = prereleases + + @property + def prereleases(self) -> Optional[bool]: + # If we have been given an explicit prerelease modifier, then we'll + # pass that through here. + if self._prereleases is not None: + return self._prereleases + + # If we don't have any specifiers, and we don't have a forced value, + # then we'll just return None since we don't know if this should have + # pre-releases or not. + if not self._specs: + return None + + # Otherwise we'll see if any of the given specifiers accept + # prereleases, if any of them do we'll return True, otherwise False. + return any(s.prereleases for s in self._specs) + + @prereleases.setter + def prereleases(self, value: bool) -> None: + self._prereleases = value + + def __repr__(self) -> str: + """A representation of the specifier set that shows all internal state. + + Note that the ordering of the individual specifiers within the set may not + match the input string. + + >>> SpecifierSet('>=1.0.0,!=2.0.0') + =1.0.0')> + >>> SpecifierSet('>=1.0.0,!=2.0.0', prereleases=False) + =1.0.0', prereleases=False)> + >>> SpecifierSet('>=1.0.0,!=2.0.0', prereleases=True) + =1.0.0', prereleases=True)> + """ + pre = ( + f", prereleases={self.prereleases!r}" + if self._prereleases is not None + else "" + ) + + return f"" + + def __str__(self) -> str: + """A string representation of the specifier set that can be round-tripped. + + Note that the ordering of the individual specifiers within the set may not + match the input string. + + >>> str(SpecifierSet(">=1.0.0,!=1.0.1")) + '!=1.0.1,>=1.0.0' + >>> str(SpecifierSet(">=1.0.0,!=1.0.1", prereleases=False)) + '!=1.0.1,>=1.0.0' + """ + return ",".join(sorted(str(s) for s in self._specs)) + + def __hash__(self) -> int: + return hash(self._specs) + + def __and__(self, other: Union["SpecifierSet", str]) -> "SpecifierSet": + """Return a SpecifierSet which is a combination of the two sets. + + :param other: The other object to combine with. + + >>> SpecifierSet(">=1.0.0,!=1.0.1") & '<=2.0.0,!=2.0.1' + =1.0.0')> + >>> SpecifierSet(">=1.0.0,!=1.0.1") & SpecifierSet('<=2.0.0,!=2.0.1') + =1.0.0')> + """ + if isinstance(other, str): + other = SpecifierSet(other) + elif not isinstance(other, SpecifierSet): + return NotImplemented + + specifier = SpecifierSet() + specifier._specs = frozenset(self._specs | other._specs) + + if self._prereleases is None and other._prereleases is not None: + specifier._prereleases = other._prereleases + elif self._prereleases is not None and other._prereleases is None: + specifier._prereleases = self._prereleases + elif self._prereleases == other._prereleases: + specifier._prereleases = self._prereleases + else: + raise ValueError( + "Cannot combine SpecifierSets with True and False prerelease " + "overrides." + ) + + return specifier + + def __eq__(self, other: object) -> bool: + """Whether or not the two SpecifierSet-like objects are equal. + + :param other: The other object to check against. + + The value of :attr:`prereleases` is ignored. + + >>> SpecifierSet(">=1.0.0,!=1.0.1") == SpecifierSet(">=1.0.0,!=1.0.1") + True + >>> (SpecifierSet(">=1.0.0,!=1.0.1", prereleases=False) == + ... SpecifierSet(">=1.0.0,!=1.0.1", prereleases=True)) + True + >>> SpecifierSet(">=1.0.0,!=1.0.1") == ">=1.0.0,!=1.0.1" + True + >>> SpecifierSet(">=1.0.0,!=1.0.1") == SpecifierSet(">=1.0.0") + False + >>> SpecifierSet(">=1.0.0,!=1.0.1") == SpecifierSet(">=1.0.0,!=1.0.2") + False + """ + if isinstance(other, (str, Specifier)): + other = SpecifierSet(str(other)) + elif not isinstance(other, SpecifierSet): + return NotImplemented + + return self._specs == other._specs + + def __len__(self) -> int: + """Returns the number of specifiers in this specifier set.""" + return len(self._specs) + + def __iter__(self) -> Iterator[Specifier]: + """ + Returns an iterator over all the underlying :class:`Specifier` instances + in this specifier set. + + >>> sorted(SpecifierSet(">=1.0.0,!=1.0.1"), key=str) + [, =1.0.0')>] + """ + return iter(self._specs) + + def __contains__(self, item: UnparsedVersion) -> bool: + """Return whether or not the item is contained in this specifier. + + :param item: The item to check for. + + This is used for the ``in`` operator and behaves the same as + :meth:`contains` with no ``prereleases`` argument passed. + + >>> "1.2.3" in SpecifierSet(">=1.0.0,!=1.0.1") + True + >>> Version("1.2.3") in SpecifierSet(">=1.0.0,!=1.0.1") + True + >>> "1.0.1" in SpecifierSet(">=1.0.0,!=1.0.1") + False + >>> "1.3.0a1" in SpecifierSet(">=1.0.0,!=1.0.1") + False + >>> "1.3.0a1" in SpecifierSet(">=1.0.0,!=1.0.1", prereleases=True) + True + """ + return self.contains(item) + + def contains( + self, + item: UnparsedVersion, + prereleases: Optional[bool] = None, + installed: Optional[bool] = None, + ) -> bool: + """Return whether or not the item is contained in this SpecifierSet. + + :param item: + The item to check for, which can be a version string or a + :class:`Version` instance. + :param prereleases: + Whether or not to match prereleases with this SpecifierSet. If set to + ``None`` (the default), it uses :attr:`prereleases` to determine + whether or not prereleases are allowed. + + >>> SpecifierSet(">=1.0.0,!=1.0.1").contains("1.2.3") + True + >>> SpecifierSet(">=1.0.0,!=1.0.1").contains(Version("1.2.3")) + True + >>> SpecifierSet(">=1.0.0,!=1.0.1").contains("1.0.1") + False + >>> SpecifierSet(">=1.0.0,!=1.0.1").contains("1.3.0a1") + False + >>> SpecifierSet(">=1.0.0,!=1.0.1", prereleases=True).contains("1.3.0a1") + True + >>> SpecifierSet(">=1.0.0,!=1.0.1").contains("1.3.0a1", prereleases=True) + True + """ + # Ensure that our item is a Version instance. + if not isinstance(item, Version): + item = Version(item) + + # Determine if we're forcing a prerelease or not, if we're not forcing + # one for this particular filter call, then we'll use whatever the + # SpecifierSet thinks for whether or not we should support prereleases. + if prereleases is None: + prereleases = self.prereleases + + # We can determine if we're going to allow pre-releases by looking to + # see if any of the underlying items supports them. If none of them do + # and this item is a pre-release then we do not allow it and we can + # short circuit that here. + # Note: This means that 1.0.dev1 would not be contained in something + # like >=1.0.devabc however it would be in >=1.0.debabc,>0.0.dev0 + if not prereleases and item.is_prerelease: + return False + + if installed and item.is_prerelease: + item = Version(item.base_version) + + # We simply dispatch to the underlying specs here to make sure that the + # given version is contained within all of them. + # Note: This use of all() here means that an empty set of specifiers + # will always return True, this is an explicit design decision. + return all(s.contains(item, prereleases=prereleases) for s in self._specs) + + def filter( + self, iterable: Iterable[UnparsedVersionVar], prereleases: Optional[bool] = None + ) -> Iterator[UnparsedVersionVar]: + """Filter items in the given iterable, that match the specifiers in this set. + + :param iterable: + An iterable that can contain version strings and :class:`Version` instances. + The items in the iterable will be filtered according to the specifier. + :param prereleases: + Whether or not to allow prereleases in the returned iterator. If set to + ``None`` (the default), it will be intelligently decide whether to allow + prereleases or not (based on the :attr:`prereleases` attribute, and + whether the only versions matching are prereleases). + + This method is smarter than just ``filter(SpecifierSet(...).contains, [...])`` + because it implements the rule from :pep:`440` that a prerelease item + SHOULD be accepted if no other versions match the given specifier. + + >>> list(SpecifierSet(">=1.2.3").filter(["1.2", "1.3", "1.5a1"])) + ['1.3'] + >>> list(SpecifierSet(">=1.2.3").filter(["1.2", "1.3", Version("1.4")])) + ['1.3', ] + >>> list(SpecifierSet(">=1.2.3").filter(["1.2", "1.5a1"])) + [] + >>> list(SpecifierSet(">=1.2.3").filter(["1.3", "1.5a1"], prereleases=True)) + ['1.3', '1.5a1'] + >>> list(SpecifierSet(">=1.2.3", prereleases=True).filter(["1.3", "1.5a1"])) + ['1.3', '1.5a1'] + + An "empty" SpecifierSet will filter items based on the presence of prerelease + versions in the set. + + >>> list(SpecifierSet("").filter(["1.3", "1.5a1"])) + ['1.3'] + >>> list(SpecifierSet("").filter(["1.5a1"])) + ['1.5a1'] + >>> list(SpecifierSet("", prereleases=True).filter(["1.3", "1.5a1"])) + ['1.3', '1.5a1'] + >>> list(SpecifierSet("").filter(["1.3", "1.5a1"], prereleases=True)) + ['1.3', '1.5a1'] + """ + # Determine if we're forcing a prerelease or not, if we're not forcing + # one for this particular filter call, then we'll use whatever the + # SpecifierSet thinks for whether or not we should support prereleases. + if prereleases is None: + prereleases = self.prereleases + + # If we have any specifiers, then we want to wrap our iterable in the + # filter method for each one, this will act as a logical AND amongst + # each specifier. + if self._specs: + for spec in self._specs: + iterable = spec.filter(iterable, prereleases=bool(prereleases)) + return iter(iterable) + # If we do not have any specifiers, then we need to have a rough filter + # which will filter out any pre-releases, unless there are no final + # releases. + else: + filtered: List[UnparsedVersionVar] = [] + found_prereleases: List[UnparsedVersionVar] = [] + + for item in iterable: + parsed_version = _coerce_version(item) + + # Store any item which is a pre-release for later unless we've + # already found a final version or we are accepting prereleases + if parsed_version.is_prerelease and not prereleases: + if not filtered: + found_prereleases.append(item) + else: + filtered.append(item) + + # If we've found no items except for pre-releases, then we'll go + # ahead and use the pre-releases + if not filtered and found_prereleases and prereleases is None: + return iter(found_prereleases) + + return iter(filtered) diff --git a/pylib/packaging/tags.py b/pylib/packaging/tags.py new file mode 100644 index 00000000..37f33b1e --- /dev/null +++ b/pylib/packaging/tags.py @@ -0,0 +1,553 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + +import logging +import platform +import struct +import subprocess +import sys +import sysconfig +from importlib.machinery import EXTENSION_SUFFIXES +from typing import ( + Dict, + FrozenSet, + Iterable, + Iterator, + List, + Optional, + Sequence, + Tuple, + Union, + cast, +) + +from . import _manylinux, _musllinux + +logger = logging.getLogger(__name__) + +PythonVersion = Sequence[int] +MacVersion = Tuple[int, int] + +INTERPRETER_SHORT_NAMES: Dict[str, str] = { + "python": "py", # Generic. + "cpython": "cp", + "pypy": "pp", + "ironpython": "ip", + "jython": "jy", +} + + +_32_BIT_INTERPRETER = struct.calcsize("P") == 4 + + +class Tag: + """ + A representation of the tag triple for a wheel. + + Instances are considered immutable and thus are hashable. Equality checking + is also supported. + """ + + __slots__ = ["_interpreter", "_abi", "_platform", "_hash"] + + def __init__(self, interpreter: str, abi: str, platform: str) -> None: + self._interpreter = interpreter.lower() + self._abi = abi.lower() + self._platform = platform.lower() + # The __hash__ of every single element in a Set[Tag] will be evaluated each time + # that a set calls its `.disjoint()` method, which may be called hundreds of + # times when scanning a page of links for packages with tags matching that + # Set[Tag]. Pre-computing the value here produces significant speedups for + # downstream consumers. + self._hash = hash((self._interpreter, self._abi, self._platform)) + + @property + def interpreter(self) -> str: + return self._interpreter + + @property + def abi(self) -> str: + return self._abi + + @property + def platform(self) -> str: + return self._platform + + def __eq__(self, other: object) -> bool: + if not isinstance(other, Tag): + return NotImplemented + + return ( + (self._hash == other._hash) # Short-circuit ASAP for perf reasons. + and (self._platform == other._platform) + and (self._abi == other._abi) + and (self._interpreter == other._interpreter) + ) + + def __hash__(self) -> int: + return self._hash + + def __str__(self) -> str: + return f"{self._interpreter}-{self._abi}-{self._platform}" + + def __repr__(self) -> str: + return f"<{self} @ {id(self)}>" + + +def parse_tag(tag: str) -> FrozenSet[Tag]: + """ + Parses the provided tag (e.g. `py3-none-any`) into a frozenset of Tag instances. + + Returning a set is required due to the possibility that the tag is a + compressed tag set. + """ + tags = set() + interpreters, abis, platforms = tag.split("-") + for interpreter in interpreters.split("."): + for abi in abis.split("."): + for platform_ in platforms.split("."): + tags.add(Tag(interpreter, abi, platform_)) + return frozenset(tags) + + +def _get_config_var(name: str, warn: bool = False) -> Union[int, str, None]: + value: Union[int, str, None] = sysconfig.get_config_var(name) + if value is None and warn: + logger.debug( + "Config variable '%s' is unset, Python ABI tag may be incorrect", name + ) + return value + + +def _normalize_string(string: str) -> str: + return string.replace(".", "_").replace("-", "_").replace(" ", "_") + + +def _abi3_applies(python_version: PythonVersion) -> bool: + """ + Determine if the Python version supports abi3. + + PEP 384 was first implemented in Python 3.2. + """ + return len(python_version) > 1 and tuple(python_version) >= (3, 2) + + +def _cpython_abis(py_version: PythonVersion, warn: bool = False) -> List[str]: + py_version = tuple(py_version) # To allow for version comparison. + abis = [] + version = _version_nodot(py_version[:2]) + debug = pymalloc = ucs4 = "" + with_debug = _get_config_var("Py_DEBUG", warn) + has_refcount = hasattr(sys, "gettotalrefcount") + # Windows doesn't set Py_DEBUG, so checking for support of debug-compiled + # extension modules is the best option. + # https://github.com/pypa/pip/issues/3383#issuecomment-173267692 + has_ext = "_d.pyd" in EXTENSION_SUFFIXES + if with_debug or (with_debug is None and (has_refcount or has_ext)): + debug = "d" + if py_version < (3, 8): + with_pymalloc = _get_config_var("WITH_PYMALLOC", warn) + if with_pymalloc or with_pymalloc is None: + pymalloc = "m" + if py_version < (3, 3): + unicode_size = _get_config_var("Py_UNICODE_SIZE", warn) + if unicode_size == 4 or ( + unicode_size is None and sys.maxunicode == 0x10FFFF + ): + ucs4 = "u" + elif debug: + # Debug builds can also load "normal" extension modules. + # We can also assume no UCS-4 or pymalloc requirement. + abis.append(f"cp{version}") + abis.insert( + 0, + "cp{version}{debug}{pymalloc}{ucs4}".format( + version=version, debug=debug, pymalloc=pymalloc, ucs4=ucs4 + ), + ) + return abis + + +def cpython_tags( + python_version: Optional[PythonVersion] = None, + abis: Optional[Iterable[str]] = None, + platforms: Optional[Iterable[str]] = None, + *, + warn: bool = False, +) -> Iterator[Tag]: + """ + Yields the tags for a CPython interpreter. + + The tags consist of: + - cp-- + - cp-abi3- + - cp-none- + - cp-abi3- # Older Python versions down to 3.2. + + If python_version only specifies a major version then user-provided ABIs and + the 'none' ABItag will be used. + + If 'abi3' or 'none' are specified in 'abis' then they will be yielded at + their normal position and not at the beginning. + """ + if not python_version: + python_version = sys.version_info[:2] + + interpreter = f"cp{_version_nodot(python_version[:2])}" + + if abis is None: + if len(python_version) > 1: + abis = _cpython_abis(python_version, warn) + else: + abis = [] + abis = list(abis) + # 'abi3' and 'none' are explicitly handled later. + for explicit_abi in ("abi3", "none"): + try: + abis.remove(explicit_abi) + except ValueError: + pass + + platforms = list(platforms or platform_tags()) + for abi in abis: + for platform_ in platforms: + yield Tag(interpreter, abi, platform_) + if _abi3_applies(python_version): + yield from (Tag(interpreter, "abi3", platform_) for platform_ in platforms) + yield from (Tag(interpreter, "none", platform_) for platform_ in platforms) + + if _abi3_applies(python_version): + for minor_version in range(python_version[1] - 1, 1, -1): + for platform_ in platforms: + interpreter = "cp{version}".format( + version=_version_nodot((python_version[0], minor_version)) + ) + yield Tag(interpreter, "abi3", platform_) + + +def _generic_abi() -> List[str]: + """ + Return the ABI tag based on EXT_SUFFIX. + """ + # The following are examples of `EXT_SUFFIX`. + # We want to keep the parts which are related to the ABI and remove the + # parts which are related to the platform: + # - linux: '.cpython-310-x86_64-linux-gnu.so' => cp310 + # - mac: '.cpython-310-darwin.so' => cp310 + # - win: '.cp310-win_amd64.pyd' => cp310 + # - win: '.pyd' => cp37 (uses _cpython_abis()) + # - pypy: '.pypy38-pp73-x86_64-linux-gnu.so' => pypy38_pp73 + # - graalpy: '.graalpy-38-native-x86_64-darwin.dylib' + # => graalpy_38_native + + ext_suffix = _get_config_var("EXT_SUFFIX", warn=True) + if not isinstance(ext_suffix, str) or ext_suffix[0] != ".": + raise SystemError("invalid sysconfig.get_config_var('EXT_SUFFIX')") + parts = ext_suffix.split(".") + if len(parts) < 3: + # CPython3.7 and earlier uses ".pyd" on Windows. + return _cpython_abis(sys.version_info[:2]) + soabi = parts[1] + if soabi.startswith("cpython"): + # non-windows + abi = "cp" + soabi.split("-")[1] + elif soabi.startswith("cp"): + # windows + abi = soabi.split("-")[0] + elif soabi.startswith("pypy"): + abi = "-".join(soabi.split("-")[:2]) + elif soabi.startswith("graalpy"): + abi = "-".join(soabi.split("-")[:3]) + elif soabi: + # pyston, ironpython, others? + abi = soabi + else: + return [] + return [_normalize_string(abi)] + + +def generic_tags( + interpreter: Optional[str] = None, + abis: Optional[Iterable[str]] = None, + platforms: Optional[Iterable[str]] = None, + *, + warn: bool = False, +) -> Iterator[Tag]: + """ + Yields the tags for a generic interpreter. + + The tags consist of: + - -- + + The "none" ABI will be added if it was not explicitly provided. + """ + if not interpreter: + interp_name = interpreter_name() + interp_version = interpreter_version(warn=warn) + interpreter = "".join([interp_name, interp_version]) + if abis is None: + abis = _generic_abi() + else: + abis = list(abis) + platforms = list(platforms or platform_tags()) + if "none" not in abis: + abis.append("none") + for abi in abis: + for platform_ in platforms: + yield Tag(interpreter, abi, platform_) + + +def _py_interpreter_range(py_version: PythonVersion) -> Iterator[str]: + """ + Yields Python versions in descending order. + + After the latest version, the major-only version will be yielded, and then + all previous versions of that major version. + """ + if len(py_version) > 1: + yield f"py{_version_nodot(py_version[:2])}" + yield f"py{py_version[0]}" + if len(py_version) > 1: + for minor in range(py_version[1] - 1, -1, -1): + yield f"py{_version_nodot((py_version[0], minor))}" + + +def compatible_tags( + python_version: Optional[PythonVersion] = None, + interpreter: Optional[str] = None, + platforms: Optional[Iterable[str]] = None, +) -> Iterator[Tag]: + """ + Yields the sequence of tags that are compatible with a specific version of Python. + + The tags consist of: + - py*-none- + - -none-any # ... if `interpreter` is provided. + - py*-none-any + """ + if not python_version: + python_version = sys.version_info[:2] + platforms = list(platforms or platform_tags()) + for version in _py_interpreter_range(python_version): + for platform_ in platforms: + yield Tag(version, "none", platform_) + if interpreter: + yield Tag(interpreter, "none", "any") + for version in _py_interpreter_range(python_version): + yield Tag(version, "none", "any") + + +def _mac_arch(arch: str, is_32bit: bool = _32_BIT_INTERPRETER) -> str: + if not is_32bit: + return arch + + if arch.startswith("ppc"): + return "ppc" + + return "i386" + + +def _mac_binary_formats(version: MacVersion, cpu_arch: str) -> List[str]: + formats = [cpu_arch] + if cpu_arch == "x86_64": + if version < (10, 4): + return [] + formats.extend(["intel", "fat64", "fat32"]) + + elif cpu_arch == "i386": + if version < (10, 4): + return [] + formats.extend(["intel", "fat32", "fat"]) + + elif cpu_arch == "ppc64": + # TODO: Need to care about 32-bit PPC for ppc64 through 10.2? + if version > (10, 5) or version < (10, 4): + return [] + formats.append("fat64") + + elif cpu_arch == "ppc": + if version > (10, 6): + return [] + formats.extend(["fat32", "fat"]) + + if cpu_arch in {"arm64", "x86_64"}: + formats.append("universal2") + + if cpu_arch in {"x86_64", "i386", "ppc64", "ppc", "intel"}: + formats.append("universal") + + return formats + + +def mac_platforms( + version: Optional[MacVersion] = None, arch: Optional[str] = None +) -> Iterator[str]: + """ + Yields the platform tags for a macOS system. + + The `version` parameter is a two-item tuple specifying the macOS version to + generate platform tags for. The `arch` parameter is the CPU architecture to + generate platform tags for. Both parameters default to the appropriate value + for the current system. + """ + version_str, _, cpu_arch = platform.mac_ver() + if version is None: + version = cast("MacVersion", tuple(map(int, version_str.split(".")[:2]))) + if version == (10, 16): + # When built against an older macOS SDK, Python will report macOS 10.16 + # instead of the real version. + version_str = subprocess.run( + [ + sys.executable, + "-sS", + "-c", + "import platform; print(platform.mac_ver()[0])", + ], + check=True, + env={"SYSTEM_VERSION_COMPAT": "0"}, + stdout=subprocess.PIPE, + text=True, + ).stdout + version = cast("MacVersion", tuple(map(int, version_str.split(".")[:2]))) + else: + version = version + if arch is None: + arch = _mac_arch(cpu_arch) + else: + arch = arch + + if (10, 0) <= version and version < (11, 0): + # Prior to Mac OS 11, each yearly release of Mac OS bumped the + # "minor" version number. The major version was always 10. + for minor_version in range(version[1], -1, -1): + compat_version = 10, minor_version + binary_formats = _mac_binary_formats(compat_version, arch) + for binary_format in binary_formats: + yield "macosx_{major}_{minor}_{binary_format}".format( + major=10, minor=minor_version, binary_format=binary_format + ) + + if version >= (11, 0): + # Starting with Mac OS 11, each yearly release bumps the major version + # number. The minor versions are now the midyear updates. + for major_version in range(version[0], 10, -1): + compat_version = major_version, 0 + binary_formats = _mac_binary_formats(compat_version, arch) + for binary_format in binary_formats: + yield "macosx_{major}_{minor}_{binary_format}".format( + major=major_version, minor=0, binary_format=binary_format + ) + + if version >= (11, 0): + # Mac OS 11 on x86_64 is compatible with binaries from previous releases. + # Arm64 support was introduced in 11.0, so no Arm binaries from previous + # releases exist. + # + # However, the "universal2" binary format can have a + # macOS version earlier than 11.0 when the x86_64 part of the binary supports + # that version of macOS. + if arch == "x86_64": + for minor_version in range(16, 3, -1): + compat_version = 10, minor_version + binary_formats = _mac_binary_formats(compat_version, arch) + for binary_format in binary_formats: + yield "macosx_{major}_{minor}_{binary_format}".format( + major=compat_version[0], + minor=compat_version[1], + binary_format=binary_format, + ) + else: + for minor_version in range(16, 3, -1): + compat_version = 10, minor_version + binary_format = "universal2" + yield "macosx_{major}_{minor}_{binary_format}".format( + major=compat_version[0], + minor=compat_version[1], + binary_format=binary_format, + ) + + +def _linux_platforms(is_32bit: bool = _32_BIT_INTERPRETER) -> Iterator[str]: + linux = _normalize_string(sysconfig.get_platform()) + if not linux.startswith("linux_"): + # we should never be here, just yield the sysconfig one and return + yield linux + return + if is_32bit: + if linux == "linux_x86_64": + linux = "linux_i686" + elif linux == "linux_aarch64": + linux = "linux_armv8l" + _, arch = linux.split("_", 1) + archs = {"armv8l": ["armv8l", "armv7l"]}.get(arch, [arch]) + yield from _manylinux.platform_tags(archs) + yield from _musllinux.platform_tags(archs) + for arch in archs: + yield f"linux_{arch}" + + +def _generic_platforms() -> Iterator[str]: + yield _normalize_string(sysconfig.get_platform()) + + +def platform_tags() -> Iterator[str]: + """ + Provides the platform tags for this installation. + """ + if platform.system() == "Darwin": + return mac_platforms() + elif platform.system() == "Linux": + return _linux_platforms() + else: + return _generic_platforms() + + +def interpreter_name() -> str: + """ + Returns the name of the running interpreter. + + Some implementations have a reserved, two-letter abbreviation which will + be returned when appropriate. + """ + name = sys.implementation.name + return INTERPRETER_SHORT_NAMES.get(name) or name + + +def interpreter_version(*, warn: bool = False) -> str: + """ + Returns the version of the running interpreter. + """ + version = _get_config_var("py_version_nodot", warn=warn) + if version: + version = str(version) + else: + version = _version_nodot(sys.version_info[:2]) + return version + + +def _version_nodot(version: PythonVersion) -> str: + return "".join(map(str, version)) + + +def sys_tags(*, warn: bool = False) -> Iterator[Tag]: + """ + Returns the sequence of tag triples for the running interpreter. + + The order of the sequence corresponds to priority order for the + interpreter, from most to least important. + """ + + interp_name = interpreter_name() + if interp_name == "cp": + yield from cpython_tags(warn=warn) + else: + yield from generic_tags() + + if interp_name == "pp": + interp = "pp3" + elif interp_name == "cp": + interp = "cp" + interpreter_version(warn=warn) + else: + interp = None + yield from compatible_tags(interpreter=interp) diff --git a/pylib/packaging/utils.py b/pylib/packaging/utils.py new file mode 100644 index 00000000..c2c2f75a --- /dev/null +++ b/pylib/packaging/utils.py @@ -0,0 +1,172 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. + +import re +from typing import FrozenSet, NewType, Tuple, Union, cast + +from .tags import Tag, parse_tag +from .version import InvalidVersion, Version + +BuildTag = Union[Tuple[()], Tuple[int, str]] +NormalizedName = NewType("NormalizedName", str) + + +class InvalidName(ValueError): + """ + An invalid distribution name; users should refer to the packaging user guide. + """ + + +class InvalidWheelFilename(ValueError): + """ + An invalid wheel filename was found, users should refer to PEP 427. + """ + + +class InvalidSdistFilename(ValueError): + """ + An invalid sdist filename was found, users should refer to the packaging user guide. + """ + + +# Core metadata spec for `Name` +_validate_regex = re.compile( + r"^([A-Z0-9]|[A-Z0-9][A-Z0-9._-]*[A-Z0-9])$", re.IGNORECASE +) +_canonicalize_regex = re.compile(r"[-_.]+") +_normalized_regex = re.compile(r"^([a-z0-9]|[a-z0-9]([a-z0-9-](?!--))*[a-z0-9])$") +# PEP 427: The build number must start with a digit. +_build_tag_regex = re.compile(r"(\d+)(.*)") + + +def canonicalize_name(name: str, *, validate: bool = False) -> NormalizedName: + if validate and not _validate_regex.match(name): + raise InvalidName(f"name is invalid: {name!r}") + # This is taken from PEP 503. + value = _canonicalize_regex.sub("-", name).lower() + return cast(NormalizedName, value) + + +def is_normalized_name(name: str) -> bool: + return _normalized_regex.match(name) is not None + + +def canonicalize_version( + version: Union[Version, str], *, strip_trailing_zero: bool = True +) -> str: + """ + This is very similar to Version.__str__, but has one subtle difference + with the way it handles the release segment. + """ + if isinstance(version, str): + try: + parsed = Version(version) + except InvalidVersion: + # Legacy versions cannot be normalized + return version + else: + parsed = version + + parts = [] + + # Epoch + if parsed.epoch != 0: + parts.append(f"{parsed.epoch}!") + + # Release segment + release_segment = ".".join(str(x) for x in parsed.release) + if strip_trailing_zero: + # NB: This strips trailing '.0's to normalize + release_segment = re.sub(r"(\.0)+$", "", release_segment) + parts.append(release_segment) + + # Pre-release + if parsed.pre is not None: + parts.append("".join(str(x) for x in parsed.pre)) + + # Post-release + if parsed.post is not None: + parts.append(f".post{parsed.post}") + + # Development release + if parsed.dev is not None: + parts.append(f".dev{parsed.dev}") + + # Local version segment + if parsed.local is not None: + parts.append(f"+{parsed.local}") + + return "".join(parts) + + +def parse_wheel_filename( + filename: str, +) -> Tuple[NormalizedName, Version, BuildTag, FrozenSet[Tag]]: + if not filename.endswith(".whl"): + raise InvalidWheelFilename( + f"Invalid wheel filename (extension must be '.whl'): {filename}" + ) + + filename = filename[:-4] + dashes = filename.count("-") + if dashes not in (4, 5): + raise InvalidWheelFilename( + f"Invalid wheel filename (wrong number of parts): {filename}" + ) + + parts = filename.split("-", dashes - 2) + name_part = parts[0] + # See PEP 427 for the rules on escaping the project name. + if "__" in name_part or re.match(r"^[\w\d._]*$", name_part, re.UNICODE) is None: + raise InvalidWheelFilename(f"Invalid project name: {filename}") + name = canonicalize_name(name_part) + + try: + version = Version(parts[1]) + except InvalidVersion as e: + raise InvalidWheelFilename( + f"Invalid wheel filename (invalid version): {filename}" + ) from e + + if dashes == 5: + build_part = parts[2] + build_match = _build_tag_regex.match(build_part) + if build_match is None: + raise InvalidWheelFilename( + f"Invalid build number: {build_part} in '{filename}'" + ) + build = cast(BuildTag, (int(build_match.group(1)), build_match.group(2))) + else: + build = () + tags = parse_tag(parts[-1]) + return (name, version, build, tags) + + +def parse_sdist_filename(filename: str) -> Tuple[NormalizedName, Version]: + if filename.endswith(".tar.gz"): + file_stem = filename[: -len(".tar.gz")] + elif filename.endswith(".zip"): + file_stem = filename[: -len(".zip")] + else: + raise InvalidSdistFilename( + f"Invalid sdist filename (extension must be '.tar.gz' or '.zip'):" + f" {filename}" + ) + + # We are requiring a PEP 440 version, which cannot contain dashes, + # so we split on the last dash. + name_part, sep, version_part = file_stem.rpartition("-") + if not sep: + raise InvalidSdistFilename(f"Invalid sdist filename: {filename}") + + name = canonicalize_name(name_part) + + try: + version = Version(version_part) + except InvalidVersion as e: + raise InvalidSdistFilename( + f"Invalid sdist filename (invalid version): {filename}" + ) from e + + return (name, version) diff --git a/pylib/packaging/version.py b/pylib/packaging/version.py new file mode 100644 index 00000000..5faab9bd --- /dev/null +++ b/pylib/packaging/version.py @@ -0,0 +1,563 @@ +# This file is dual licensed under the terms of the Apache License, Version +# 2.0, and the BSD License. See the LICENSE file in the root of this repository +# for complete details. +""" +.. testsetup:: + + from packaging.version import parse, Version +""" + +import itertools +import re +from typing import Any, Callable, NamedTuple, Optional, SupportsInt, Tuple, Union + +from ._structures import Infinity, InfinityType, NegativeInfinity, NegativeInfinityType + +__all__ = ["VERSION_PATTERN", "parse", "Version", "InvalidVersion"] + +LocalType = Tuple[Union[int, str], ...] + +CmpPrePostDevType = Union[InfinityType, NegativeInfinityType, Tuple[str, int]] +CmpLocalType = Union[ + NegativeInfinityType, + Tuple[Union[Tuple[int, str], Tuple[NegativeInfinityType, Union[int, str]]], ...], +] +CmpKey = Tuple[ + int, + Tuple[int, ...], + CmpPrePostDevType, + CmpPrePostDevType, + CmpPrePostDevType, + CmpLocalType, +] +VersionComparisonMethod = Callable[[CmpKey, CmpKey], bool] + + +class _Version(NamedTuple): + epoch: int + release: Tuple[int, ...] + dev: Optional[Tuple[str, int]] + pre: Optional[Tuple[str, int]] + post: Optional[Tuple[str, int]] + local: Optional[LocalType] + + +def parse(version: str) -> "Version": + """Parse the given version string. + + >>> parse('1.0.dev1') + + + :param version: The version string to parse. + :raises InvalidVersion: When the version string is not a valid version. + """ + return Version(version) + + +class InvalidVersion(ValueError): + """Raised when a version string is not a valid version. + + >>> Version("invalid") + Traceback (most recent call last): + ... + packaging.version.InvalidVersion: Invalid version: 'invalid' + """ + + +class _BaseVersion: + _key: Tuple[Any, ...] + + def __hash__(self) -> int: + return hash(self._key) + + # Please keep the duplicated `isinstance` check + # in the six comparisons hereunder + # unless you find a way to avoid adding overhead function calls. + def __lt__(self, other: "_BaseVersion") -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key < other._key + + def __le__(self, other: "_BaseVersion") -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key <= other._key + + def __eq__(self, other: object) -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key == other._key + + def __ge__(self, other: "_BaseVersion") -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key >= other._key + + def __gt__(self, other: "_BaseVersion") -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key > other._key + + def __ne__(self, other: object) -> bool: + if not isinstance(other, _BaseVersion): + return NotImplemented + + return self._key != other._key + + +# Deliberately not anchored to the start and end of the string, to make it +# easier for 3rd party code to reuse +_VERSION_PATTERN = r""" + v? + (?: + (?:(?P[0-9]+)!)? # epoch + (?P[0-9]+(?:\.[0-9]+)*) # release segment + (?P
                                          # pre-release
+            [-_\.]?
+            (?Palpha|a|beta|b|preview|pre|c|rc)
+            [-_\.]?
+            (?P[0-9]+)?
+        )?
+        (?P                                         # post release
+            (?:-(?P[0-9]+))
+            |
+            (?:
+                [-_\.]?
+                (?Ppost|rev|r)
+                [-_\.]?
+                (?P[0-9]+)?
+            )
+        )?
+        (?P                                          # dev release
+            [-_\.]?
+            (?Pdev)
+            [-_\.]?
+            (?P[0-9]+)?
+        )?
+    )
+    (?:\+(?P[a-z0-9]+(?:[-_\.][a-z0-9]+)*))?       # local version
+"""
+
+VERSION_PATTERN = _VERSION_PATTERN
+"""
+A string containing the regular expression used to match a valid version.
+
+The pattern is not anchored at either end, and is intended for embedding in larger
+expressions (for example, matching a version number as part of a file name). The
+regular expression should be compiled with the ``re.VERBOSE`` and ``re.IGNORECASE``
+flags set.
+
+:meta hide-value:
+"""
+
+
+class Version(_BaseVersion):
+    """This class abstracts handling of a project's versions.
+
+    A :class:`Version` instance is comparison aware and can be compared and
+    sorted using the standard Python interfaces.
+
+    >>> v1 = Version("1.0a5")
+    >>> v2 = Version("1.0")
+    >>> v1
+    
+    >>> v2
+    
+    >>> v1 < v2
+    True
+    >>> v1 == v2
+    False
+    >>> v1 > v2
+    False
+    >>> v1 >= v2
+    False
+    >>> v1 <= v2
+    True
+    """
+
+    _regex = re.compile(r"^\s*" + VERSION_PATTERN + r"\s*$", re.VERBOSE | re.IGNORECASE)
+    _key: CmpKey
+
+    def __init__(self, version: str) -> None:
+        """Initialize a Version object.
+
+        :param version:
+            The string representation of a version which will be parsed and normalized
+            before use.
+        :raises InvalidVersion:
+            If the ``version`` does not conform to PEP 440 in any way then this
+            exception will be raised.
+        """
+
+        # Validate the version and parse it into pieces
+        match = self._regex.search(version)
+        if not match:
+            raise InvalidVersion(f"Invalid version: '{version}'")
+
+        # Store the parsed out pieces of the version
+        self._version = _Version(
+            epoch=int(match.group("epoch")) if match.group("epoch") else 0,
+            release=tuple(int(i) for i in match.group("release").split(".")),
+            pre=_parse_letter_version(match.group("pre_l"), match.group("pre_n")),
+            post=_parse_letter_version(
+                match.group("post_l"), match.group("post_n1") or match.group("post_n2")
+            ),
+            dev=_parse_letter_version(match.group("dev_l"), match.group("dev_n")),
+            local=_parse_local_version(match.group("local")),
+        )
+
+        # Generate a key which will be used for sorting
+        self._key = _cmpkey(
+            self._version.epoch,
+            self._version.release,
+            self._version.pre,
+            self._version.post,
+            self._version.dev,
+            self._version.local,
+        )
+
+    def __repr__(self) -> str:
+        """A representation of the Version that shows all internal state.
+
+        >>> Version('1.0.0')
+        
+        """
+        return f""
+
+    def __str__(self) -> str:
+        """A string representation of the version that can be rounded-tripped.
+
+        >>> str(Version("1.0a5"))
+        '1.0a5'
+        """
+        parts = []
+
+        # Epoch
+        if self.epoch != 0:
+            parts.append(f"{self.epoch}!")
+
+        # Release segment
+        parts.append(".".join(str(x) for x in self.release))
+
+        # Pre-release
+        if self.pre is not None:
+            parts.append("".join(str(x) for x in self.pre))
+
+        # Post-release
+        if self.post is not None:
+            parts.append(f".post{self.post}")
+
+        # Development release
+        if self.dev is not None:
+            parts.append(f".dev{self.dev}")
+
+        # Local version segment
+        if self.local is not None:
+            parts.append(f"+{self.local}")
+
+        return "".join(parts)
+
+    @property
+    def epoch(self) -> int:
+        """The epoch of the version.
+
+        >>> Version("2.0.0").epoch
+        0
+        >>> Version("1!2.0.0").epoch
+        1
+        """
+        return self._version.epoch
+
+    @property
+    def release(self) -> Tuple[int, ...]:
+        """The components of the "release" segment of the version.
+
+        >>> Version("1.2.3").release
+        (1, 2, 3)
+        >>> Version("2.0.0").release
+        (2, 0, 0)
+        >>> Version("1!2.0.0.post0").release
+        (2, 0, 0)
+
+        Includes trailing zeroes but not the epoch or any pre-release / development /
+        post-release suffixes.
+        """
+        return self._version.release
+
+    @property
+    def pre(self) -> Optional[Tuple[str, int]]:
+        """The pre-release segment of the version.
+
+        >>> print(Version("1.2.3").pre)
+        None
+        >>> Version("1.2.3a1").pre
+        ('a', 1)
+        >>> Version("1.2.3b1").pre
+        ('b', 1)
+        >>> Version("1.2.3rc1").pre
+        ('rc', 1)
+        """
+        return self._version.pre
+
+    @property
+    def post(self) -> Optional[int]:
+        """The post-release number of the version.
+
+        >>> print(Version("1.2.3").post)
+        None
+        >>> Version("1.2.3.post1").post
+        1
+        """
+        return self._version.post[1] if self._version.post else None
+
+    @property
+    def dev(self) -> Optional[int]:
+        """The development number of the version.
+
+        >>> print(Version("1.2.3").dev)
+        None
+        >>> Version("1.2.3.dev1").dev
+        1
+        """
+        return self._version.dev[1] if self._version.dev else None
+
+    @property
+    def local(self) -> Optional[str]:
+        """The local version segment of the version.
+
+        >>> print(Version("1.2.3").local)
+        None
+        >>> Version("1.2.3+abc").local
+        'abc'
+        """
+        if self._version.local:
+            return ".".join(str(x) for x in self._version.local)
+        else:
+            return None
+
+    @property
+    def public(self) -> str:
+        """The public portion of the version.
+
+        >>> Version("1.2.3").public
+        '1.2.3'
+        >>> Version("1.2.3+abc").public
+        '1.2.3'
+        >>> Version("1.2.3+abc.dev1").public
+        '1.2.3'
+        """
+        return str(self).split("+", 1)[0]
+
+    @property
+    def base_version(self) -> str:
+        """The "base version" of the version.
+
+        >>> Version("1.2.3").base_version
+        '1.2.3'
+        >>> Version("1.2.3+abc").base_version
+        '1.2.3'
+        >>> Version("1!1.2.3+abc.dev1").base_version
+        '1!1.2.3'
+
+        The "base version" is the public version of the project without any pre or post
+        release markers.
+        """
+        parts = []
+
+        # Epoch
+        if self.epoch != 0:
+            parts.append(f"{self.epoch}!")
+
+        # Release segment
+        parts.append(".".join(str(x) for x in self.release))
+
+        return "".join(parts)
+
+    @property
+    def is_prerelease(self) -> bool:
+        """Whether this version is a pre-release.
+
+        >>> Version("1.2.3").is_prerelease
+        False
+        >>> Version("1.2.3a1").is_prerelease
+        True
+        >>> Version("1.2.3b1").is_prerelease
+        True
+        >>> Version("1.2.3rc1").is_prerelease
+        True
+        >>> Version("1.2.3dev1").is_prerelease
+        True
+        """
+        return self.dev is not None or self.pre is not None
+
+    @property
+    def is_postrelease(self) -> bool:
+        """Whether this version is a post-release.
+
+        >>> Version("1.2.3").is_postrelease
+        False
+        >>> Version("1.2.3.post1").is_postrelease
+        True
+        """
+        return self.post is not None
+
+    @property
+    def is_devrelease(self) -> bool:
+        """Whether this version is a development release.
+
+        >>> Version("1.2.3").is_devrelease
+        False
+        >>> Version("1.2.3.dev1").is_devrelease
+        True
+        """
+        return self.dev is not None
+
+    @property
+    def major(self) -> int:
+        """The first item of :attr:`release` or ``0`` if unavailable.
+
+        >>> Version("1.2.3").major
+        1
+        """
+        return self.release[0] if len(self.release) >= 1 else 0
+
+    @property
+    def minor(self) -> int:
+        """The second item of :attr:`release` or ``0`` if unavailable.
+
+        >>> Version("1.2.3").minor
+        2
+        >>> Version("1").minor
+        0
+        """
+        return self.release[1] if len(self.release) >= 2 else 0
+
+    @property
+    def micro(self) -> int:
+        """The third item of :attr:`release` or ``0`` if unavailable.
+
+        >>> Version("1.2.3").micro
+        3
+        >>> Version("1").micro
+        0
+        """
+        return self.release[2] if len(self.release) >= 3 else 0
+
+
+def _parse_letter_version(
+    letter: Optional[str], number: Union[str, bytes, SupportsInt, None]
+) -> Optional[Tuple[str, int]]:
+
+    if letter:
+        # We consider there to be an implicit 0 in a pre-release if there is
+        # not a numeral associated with it.
+        if number is None:
+            number = 0
+
+        # We normalize any letters to their lower case form
+        letter = letter.lower()
+
+        # We consider some words to be alternate spellings of other words and
+        # in those cases we want to normalize the spellings to our preferred
+        # spelling.
+        if letter == "alpha":
+            letter = "a"
+        elif letter == "beta":
+            letter = "b"
+        elif letter in ["c", "pre", "preview"]:
+            letter = "rc"
+        elif letter in ["rev", "r"]:
+            letter = "post"
+
+        return letter, int(number)
+    if not letter and number:
+        # We assume if we are given a number, but we are not given a letter
+        # then this is using the implicit post release syntax (e.g. 1.0-1)
+        letter = "post"
+
+        return letter, int(number)
+
+    return None
+
+
+_local_version_separators = re.compile(r"[\._-]")
+
+
+def _parse_local_version(local: Optional[str]) -> Optional[LocalType]:
+    """
+    Takes a string like abc.1.twelve and turns it into ("abc", 1, "twelve").
+    """
+    if local is not None:
+        return tuple(
+            part.lower() if not part.isdigit() else int(part)
+            for part in _local_version_separators.split(local)
+        )
+    return None
+
+
+def _cmpkey(
+    epoch: int,
+    release: Tuple[int, ...],
+    pre: Optional[Tuple[str, int]],
+    post: Optional[Tuple[str, int]],
+    dev: Optional[Tuple[str, int]],
+    local: Optional[LocalType],
+) -> CmpKey:
+
+    # When we compare a release version, we want to compare it with all of the
+    # trailing zeros removed. So we'll use a reverse the list, drop all the now
+    # leading zeros until we come to something non zero, then take the rest
+    # re-reverse it back into the correct order and make it a tuple and use
+    # that for our sorting key.
+    _release = tuple(
+        reversed(list(itertools.dropwhile(lambda x: x == 0, reversed(release))))
+    )
+
+    # We need to "trick" the sorting algorithm to put 1.0.dev0 before 1.0a0.
+    # We'll do this by abusing the pre segment, but we _only_ want to do this
+    # if there is not a pre or a post segment. If we have one of those then
+    # the normal sorting rules will handle this case correctly.
+    if pre is None and post is None and dev is not None:
+        _pre: CmpPrePostDevType = NegativeInfinity
+    # Versions without a pre-release (except as noted above) should sort after
+    # those with one.
+    elif pre is None:
+        _pre = Infinity
+    else:
+        _pre = pre
+
+    # Versions without a post segment should sort before those with one.
+    if post is None:
+        _post: CmpPrePostDevType = NegativeInfinity
+
+    else:
+        _post = post
+
+    # Versions without a development segment should sort after those with one.
+    if dev is None:
+        _dev: CmpPrePostDevType = Infinity
+
+    else:
+        _dev = dev
+
+    if local is None:
+        # Versions without a local segment should sort before those with one.
+        _local: CmpLocalType = NegativeInfinity
+    else:
+        # Versions with a local segment need that segment parsed to implement
+        # the sorting rules in PEP440.
+        # - Alpha numeric segments sort before numeric segments
+        # - Alpha numeric segments sort lexicographically
+        # - Numeric segments sort numerically
+        # - Shorter versions sort before longer versions when the prefixes
+        #   match exactly
+        _local = tuple(
+            (i, "") if isinstance(i, int) else (NegativeInfinity, i) for i in local
+        )
+
+    return epoch, _release, _pre, _post, _dev, _local