23745351279

Committed 30 Mar 2026 12:44PM UTC coverage: 92.915% (-0.008%) from 92.923%

Build # 23745351279

Build Type

Pull #23197

github

Committed by

web-flow

Commit Message

Merge 910623766 into da60c6486

Pull Request Pull Request #23197: Add experimental uv-based PEX builder (`[python].pex_builder`)

Coverage Stats

123 of 139 new or added lines in 5 files covered. (88.49%)

1 existing line in 1 file now uncovered.

91643 of 98631 relevant lines covered (92.92%)

4.05 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

94.94

/src/python/pants/backend/python/util_rules/pex.py

# Copyright 2019 Pants project contributors (see CONTRIBUTORS.md).
# Licensed under the Apache License, Version 2.0 (see LICENSE).

from __future__ import annotations

import dataclasses
import json
import logging
import os
import shlex
import zlib
from collections.abc import Iterable, Iterator, Mapping, Sequence
from dataclasses import dataclass
from pathlib import PurePath
from textwrap import dedent  # noqa: PNT20
from typing import TypeVar

import packaging.specifiers
import packaging.version
from packaging.requirements import Requirement

from pants.backend.python.subsystems import uv as uv_subsystem
from pants.backend.python.subsystems.setup import PexBuilder, PythonSetup
from pants.backend.python.subsystems.uv import download_uv_binary
from pants.backend.python.target_types import (
    Executable,
    MainSpecification,
    PexCompletePlatformsField,
    PexLayout,
    PythonRequirementFindLinksField,
    PythonRequirementsField,
)
from pants.backend.python.util_rules import pex_cli, pex_requirements
from pants.backend.python.util_rules.interpreter_constraints import InterpreterConstraints
from pants.backend.python.util_rules.pex_cli import PexCliProcess, PexPEX, maybe_log_pex_stderr
from pants.backend.python.util_rules.pex_environment import (
    CompletePexEnvironment,
    PexEnvironment,
    PexSubsystem,
    PythonExecutable,
)
from pants.backend.python.util_rules.pex_requirements import (
    EntireLockfile,
    LoadedLockfile,
    LoadedLockfileRequest,
    Lockfile,
    Resolve,
    ResolvePexConfigRequest,
    determine_resolve_pex_config,
    get_lockfile_for_resolve,
    load_lockfile,
    validate_metadata,
)
from pants.backend.python.util_rules.pex_requirements import (
    PexRequirements as PexRequirements,  # Explicit re-export.
)
from pants.build_graph.address import Address
from pants.core.environments.target_types import EnvironmentTarget
from pants.core.target_types import FileSourceField, ResourceSourceField
from pants.core.util_rules.stripped_source_files import StrippedFileNameRequest, strip_file_name
from pants.core.util_rules.stripped_source_files import rules as stripped_source_rules
from pants.core.util_rules.system_binaries import BashBinary
from pants.engine.addresses import UnparsedAddressInputs
from pants.engine.collection import Collection, DeduplicatedCollection
from pants.engine.engine_aware import EngineAwareParameter
from pants.engine.environment import EnvironmentName
from pants.engine.fs import (
    EMPTY_DIGEST,
    AddPrefix,
    CreateDigest,
    Digest,
    Directory,
    FileContent,
    MergeDigests,
    RemovePrefix,
)
from pants.engine.internals.graph import (
    hydrate_sources,
    resolve_targets,
    resolve_unparsed_address_inputs,
)
from pants.engine.internals.graph import transitive_targets as transitive_targets_get
from pants.engine.internals.native_engine import Snapshot
from pants.engine.intrinsics import (
    add_prefix,
    create_digest,
    digest_to_snapshot,
    get_digest_contents,
    merge_digests,
    remove_prefix,
)
from pants.engine.process import (
    Process,
    ProcessCacheScope,
    ProcessResult,
    execute_process_or_raise,
    fallible_to_exec_result_or_raise,
)
from pants.engine.rules import collect_rules, concurrently, implicitly, rule
from pants.engine.target import HydrateSourcesRequest, SourcesField, TransitiveTargetsRequest
from pants.engine.unions import UnionMembership, union
from pants.util.frozendict import FrozenDict
from pants.util.logging import LogLevel
from pants.util.strutil import bullet_list, pluralize, softwrap

logger = logging.getLogger(__name__)


@union(in_scope_types=[EnvironmentName])
@dataclass(frozen=True)
class PythonProvider:
    """Union which should have 0 or 1 implementations registered which provide Python.

    Subclasses should provide a rule from their subclass type to `PythonExecutable`.
    """

    interpreter_constraints: InterpreterConstraints


@rule(polymorphic=True)
async def get_python_executable(
    provider: PythonProvider, env_name: EnvironmentName
) -> PythonExecutable:
    raise NotImplementedError()


class PexPlatforms(DeduplicatedCollection[str]):
    sort_input = True

    def generate_pex_arg_list(self) -> list[str]:
        args = []
        for platform in self:
            args.extend(["--platform", platform])
        return args


class CompletePlatforms(DeduplicatedCollection[str]):
    sort_input = True

    def __init__(self, iterable: Iterable[str] = (), *, digest: Digest = EMPTY_DIGEST):
        super().__init__(iterable)
        self._digest = digest

    @classmethod
    def from_snapshot(cls, snapshot: Snapshot) -> CompletePlatforms:
        return cls(snapshot.files, digest=snapshot.digest)

    @property
    def digest(self) -> Digest:
        return self._digest

    def generate_pex_arg_list(self) -> Iterator[str]:
        for path in self:
            yield "--complete-platform"
            yield path


@rule
async def digest_complete_platform_addresses(
    addresses: UnparsedAddressInputs,
) -> CompletePlatforms:
    original_file_targets = await resolve_targets(**implicitly(addresses))
    original_files_sources = await concurrently(
        hydrate_sources(
            HydrateSourcesRequest(
                tgt.get(SourcesField),
                for_sources_types=(
                    FileSourceField,
                    ResourceSourceField,
                ),
                enable_codegen=True,
            ),
            **implicitly(),
        )
        for tgt in original_file_targets
    )
    snapshot = await digest_to_snapshot(
        **implicitly(MergeDigests(sources.snapshot.digest for sources in original_files_sources))
    )
    return CompletePlatforms.from_snapshot(snapshot)


@rule
async def digest_complete_platforms(
    complete_platforms: PexCompletePlatformsField,
) -> CompletePlatforms:
    return await digest_complete_platform_addresses(complete_platforms.to_unparsed_address_inputs())


@dataclass(frozen=True)
class PexRequest(EngineAwareParameter):
    output_filename: str
    internal_only: bool
    layout: PexLayout
    python: PythonExecutable | None
    requirements: PexRequirements | EntireLockfile
    interpreter_constraints: InterpreterConstraints
    platforms: PexPlatforms
    complete_platforms: CompletePlatforms
    sources: Digest | None
    additional_inputs: Digest
    main: MainSpecification | None
    inject_args: tuple[str, ...]
    inject_env: FrozenDict[str, str]
    additional_args: tuple[str, ...]
    pex_path: tuple[Pex, ...]
    description: str | None = dataclasses.field(compare=False)
    cache_scope: ProcessCacheScope

    def __init__(
        self,
        *,
        output_filename: str,
        internal_only: bool,
        layout: PexLayout | None = None,
        python: PythonExecutable | None = None,
        requirements: PexRequirements | EntireLockfile = PexRequirements(),
        interpreter_constraints=InterpreterConstraints(),
        platforms=PexPlatforms(),
        complete_platforms=CompletePlatforms(),
        sources: Digest | None = None,
        additional_inputs: Digest | None = None,
        main: MainSpecification | None = None,
        inject_args: Iterable[str] = (),
        inject_env: Mapping[str, str] = FrozenDict(),
        additional_args: Iterable[str] = (),
        pex_path: Iterable[Pex] = (),
        description: str | None = None,
        cache_scope: ProcessCacheScope = ProcessCacheScope.SUCCESSFUL,
    ) -> None:
        """A request to create a PEX from its inputs.

        :param output_filename: The name of the built Pex file, which typically should end in
            `.pex`.
        :param internal_only: Whether we ever materialize the Pex and distribute it directly
            to end users, such as with the `binary` goal. Typically, instead, the user never
            directly uses the Pex, e.g. with `lint` and `test`. If True, we will use a Pex setting
            that results in faster build time but compatibility with fewer interpreters at runtime.
        :param layout: The filesystem layout to create the PEX with.
        :param python: A particular PythonExecutable to use, which must match any relevant
            interpreter_constraints.
        :param requirements: The requirements that the PEX should contain.
        :param interpreter_constraints: Any constraints on which Python versions may be used.
        :param platforms: Which abbreviated platforms should be supported. Setting this value will
            cause interpreter constraints to not be used at PEX build time because platforms already
            constrain the valid Python versions, e.g. by including `cp36m` in the platform string.
            Unfortunately this also causes interpreter constraints to not be embedded in the built
            PEX for use at runtime which can lead to problems.
            See: https://github.com/pantsbuild/pants/issues/13904.
        :param complete_platforms: Which complete platforms should be supported. Setting this value
            will cause interpreter constraints to not be used at PEX build time because complete
            platforms completely constrain the valid Python versions. Unfortunately this also causes
            interpreter constraints to not be embedded in the built PEX for use at runtime which can
            lead to problems. See: https://github.com/pantsbuild/pants/issues/13904.
        :param sources: Any source files that should be included in the Pex.
        :param additional_inputs: Any inputs that are not source files and should not be included
            directly in the Pex, but should be present in the environment when building the Pex.
        :param main: The main for the built Pex, equivalent to Pex's `-e` or '-c' flag. If
            left off, the Pex will open up as a REPL.
        :param inject_args: Command line arguments to freeze in to the PEX.
        :param inject_env: Environment variables to freeze in to the PEX.
        :param additional_args: Any additional Pex flags.
        :param pex_path: Pex files to add to the PEX_PATH.
        :param description: A human-readable description to render in the dynamic UI when building
            the Pex.
        :param cache_scope: The cache scope for the underlying pex cli invocation process.
        """
        object.__setattr__(self, "output_filename", output_filename)
        object.__setattr__(self, "internal_only", internal_only)
        # Use any explicitly requested layout, or Packed for internal PEXes (which is a much
        # friendlier layout for the CAS than Zipapp.)
        object.__setattr__(
            self, "layout", layout or (PexLayout.PACKED if internal_only else PexLayout.ZIPAPP)
        )
        object.__setattr__(self, "python", python)
        object.__setattr__(self, "requirements", requirements)
        object.__setattr__(self, "interpreter_constraints", interpreter_constraints)
        object.__setattr__(self, "platforms", platforms)
        object.__setattr__(self, "complete_platforms", complete_platforms)
        object.__setattr__(self, "sources", sources)
        object.__setattr__(self, "additional_inputs", additional_inputs or EMPTY_DIGEST)
        object.__setattr__(self, "main", main)
        object.__setattr__(self, "inject_args", tuple(inject_args))
        object.__setattr__(self, "inject_env", FrozenDict(inject_env))
        object.__setattr__(self, "additional_args", tuple(additional_args))
        object.__setattr__(self, "pex_path", tuple(pex_path))
        object.__setattr__(self, "description", description)
        object.__setattr__(self, "cache_scope", cache_scope)

        self.__post_init__()

    def __post_init__(self):
        if self.internal_only and self.platforms:
            raise ValueError(
                softwrap(
                    f"""
                    Internal only PEXes can only constrain interpreters with interpreter_constraints.
                    Given platform constraints {self.platforms} for internal only pex request:
                    {self}.
                    """
                )
            )
        if self.internal_only and self.complete_platforms:
            raise ValueError(
                softwrap(
                    f"""
                    Internal only PEXes can only constrain interpreters with interpreter_constraints.
                    Given complete_platform constraints {self.complete_platforms} for internal only
                    pex request: {self}.
                    """
                )
            )
        if self.python and self.platforms:
            raise ValueError(
                softwrap(
                    f"""
                    Only one of platforms or a specific interpreter may be set. Got
                    both {self.platforms} and {self.python}.
                    """
                )
            )
        if self.python and self.complete_platforms:
            raise ValueError(
                softwrap(
                    f"""
                    Only one of complete_platforms or a specific interpreter may be set. Got
                    both {self.complete_platforms} and {self.python}.
                    """
                )
            )
        if self.python and self.interpreter_constraints:
            raise ValueError(
                softwrap(
                    f"""
                    Only one of interpreter_constraints or a specific interpreter may be set. Got
                    both {self.interpreter_constraints} and {self.python}.
                    """
                )
            )

    def debug_hint(self) -> str:
        return self.output_filename


@dataclass(frozen=True)
class OptionalPexRequest:
    maybe_pex_request: PexRequest | None


@dataclass(frozen=True)
class Pex:
    """Wrapper for a digest containing a pex file created with some filename."""

    digest: Digest
    name: str
    python: PythonExecutable | None


@dataclass(frozen=True)
class OptionalPex:
    maybe_pex: Pex | None


@rule(desc="Find Python interpreter for constraints", level=LogLevel.DEBUG)
async def find_interpreter(
    interpreter_constraints: InterpreterConstraints,
    pex_subsystem: PexSubsystem,
    env_target: EnvironmentTarget,
    union_membership: UnionMembership,
) -> PythonExecutable:
    python_providers = union_membership.get(PythonProvider)
    if len(python_providers) > 1:
        raise ValueError(
            softwrap(
                f"""
                Too many Python provider plugins were registered. We expected 0 or 1, but found
                {len(python_providers)}. Providers were:

                {bullet_list(repr(provider.__class__) for provider in python_providers)}
                """
            )
        )
    if python_providers:
        python_provider = next(iter(python_providers))
        python = await get_python_executable(
            **implicitly({python_provider(interpreter_constraints): PythonProvider})
        )
        return python

    formatted_constraints = " OR ".join(str(constraint) for constraint in interpreter_constraints)
    result = await fallible_to_exec_result_or_raise(
        **implicitly(
            PexCliProcess(
                description=f"Find interpreter for constraints: {formatted_constraints}",
                subcommand=(),
                # Here, we run the Pex CLI with no requirements, which just selects an interpreter.
                # Normally, this would start an isolated repl. By passing `--`, we force the repl to
                # instead act as an interpreter (the selected one) and tell us about itself. The upshot
                # is we run the Pex interpreter selection logic unperturbed but without resolving any
                # distributions.
                extra_args=(
                    *interpreter_constraints.generate_pex_arg_list(),
                    "--",
                    "-c",
                    # N.B.: The following code snippet must be compatible with Python 2.7 and
                    # Python 3.5+.
                    #
                    # When hashing, we pick 8192 for efficiency of reads and fingerprint updates
                    # (writes) since it's a common OS buffer size and an even multiple of the
                    # hash block size.
                    dedent(
                        """\
                    import hashlib, os, sys

                    python = os.path.realpath(sys.executable)
                    print(python)

                    hasher = hashlib.sha256()
                    with open(python, "rb") as fp:
                      for chunk in iter(lambda: fp.read(8192), b""):
                          hasher.update(chunk)
                    print(hasher.hexdigest())
                    """
                    ),
                ),
                level=LogLevel.DEBUG,
                cache_scope=env_target.executable_search_path_cache_scope(),
            )
        )
    )
    path, fingerprint = result.stdout.decode().strip().splitlines()

    maybe_log_pex_stderr(result.stderr, pex_subsystem.verbosity)

    return PythonExecutable(path=path, fingerprint=fingerprint)


@dataclass(frozen=True)
class BuildPexResult:
    result: ProcessResult
    pex_filename: str
    digest: Digest
    python: PythonExecutable | None

    def create_pex(self) -> Pex:
        return Pex(digest=self.digest, name=self.pex_filename, python=self.python)


@dataclass
class _BuildPexPythonSetup:
    python: PythonExecutable | None
    argv: list[str]


@rule
async def _determine_pex_python_and_platforms(request: PexRequest) -> _BuildPexPythonSetup:
    # NB: If `--platform` is specified, this signals that the PEX should not be built locally.
    # `--interpreter-constraint` only makes sense in the context of building locally. These two
    # flags are mutually exclusive. See https://github.com/pex-tool/pex/issues/957.
    if request.platforms or request.complete_platforms:
        # Note that this means that this is not an internal-only pex.
        # TODO(#9560): consider validating that these platforms are valid with the interpreter
        #  constraints.
        return _BuildPexPythonSetup(
            None,
            [
                *request.platforms.generate_pex_arg_list(),
                *request.complete_platforms.generate_pex_arg_list(),
            ],
        )

    if request.python:
        python = request.python
    else:
        python = await find_interpreter(request.interpreter_constraints, **implicitly())

    if request.python or request.internal_only:
        # Sometimes we want to build and run with a specific interpreter (either because request
        # demanded it, or because it's an internal-only PEX). We will have already validated that
        # there were no platforms.
        return _BuildPexPythonSetup(python, ["--python", python.path])

    else:
        # Otherwise, we don't want to force compatibility with a particular interpreter (as in, the
        # resulting PEX should follow the ICs), but we _do_ want to tell PEX about at least one
        # interpreter that is compatible, to ensure that an interpreter installed/managed by
        # provider backends are visible (in the extreme case, a machine may have no Python
        # interpreters installed at all, and just rely on Pants' provider backends to install them,
        # and thus pex searching $PATH will find nothing).
        return _BuildPexPythonSetup(
            python,
            [
                *request.interpreter_constraints.generate_pex_arg_list(),
                "--python-path",
                python.path,
            ],
        )


_UV_VENV_DIR = "__uv_venv"


@dataclass(frozen=True)
class _UvVenvRequest:
    """Request to build a pre-populated venv using uv for PEX --venv-repository."""

    req_strings: tuple[str, ...]
    requirements: PexRequirements | EntireLockfile
    python_path: str
    description: str


@dataclass(frozen=True)
class _UvVenvResult:
    """Result of building a uv venv."""

    venv_digest: Digest | None


def _check_uv_preconditions(
    request: PexRequest,
    req_strings: tuple[str, ...],
    pex_python_setup: _BuildPexPythonSetup,
) -> str | None:
    """Check whether the uv builder can be used for this PEX request.

    Returns None if all preconditions are met, or a warning message describing
    why uv cannot be used.
    """
    label = request.description or request.output_filename
    if not req_strings:
        return (
            f"pex_builder=uv: no individual requirement strings for {label} "
            "(e.g. using a whole-lockfile resolve or no third-party deps). "
            "Falling back to the default PEX/pip builder."
        )
    if request.platforms or request.complete_platforms:
        return (
            f"pex_builder=uv: cross-platform build detected for {label}. "
            "Falling back to the default PEX/pip builder."
        )
    if pex_python_setup.python is None:
        return (
            f"pex_builder=uv: no local Python interpreter available for {label}. "
            "Falling back to the default PEX/pip builder."
        )
    return None


@rule
async def _build_uv_venv(
    uv_request: _UvVenvRequest,
    pex_env: PexEnvironment,
) -> _UvVenvResult:
    """Build a pre-populated venv using uv for use with PEX --venv-repository."""
    downloaded_uv = await download_uv_binary(**implicitly())

    logger.debug(
        "pex_builder=uv: using uv builder for %s",
        uv_request.description,
    )

    # Try to extract the full resolved package list from the lockfile
    # so we can pass pinned versions with --no-deps (reproducible).
    # Fall back to letting uv resolve transitively if no lockfile.
    all_resolved_reqs: tuple[str, ...] = ()
    if isinstance(uv_request.requirements, PexRequirements) and isinstance(
        uv_request.requirements.from_superset, Resolve
    ):
        lockfile = await get_lockfile_for_resolve(
            uv_request.requirements.from_superset, **implicitly()
        )
        loaded_lockfile = await load_lockfile(LoadedLockfileRequest(lockfile), **implicitly())
        if loaded_lockfile.is_pex_native:
            try:
                digest_contents = await get_digest_contents(loaded_lockfile.lockfile_digest)
                lockfile_bytes = next(
                    c.content for c in digest_contents if c.path == loaded_lockfile.lockfile_path
                )
                lockfile_data = json.loads(lockfile_bytes)
                all_resolved_reqs = tuple(
                    f"{req['project_name']}=={req['version']}"
                    for resolve in lockfile_data.get("locked_resolves", ())
                    for req in resolve.get("locked_requirements", ())
                )
            except (json.JSONDecodeError, KeyError, StopIteration) as e:
                logger.warning(
                    "pex_builder=uv: failed to parse lockfile for %s: %s. "
                    "Falling back to transitive uv resolution.",
                    uv_request.description,
                    e,
                )
                all_resolved_reqs = ()

    uv_reqs = all_resolved_reqs or uv_request.req_strings

    if all_resolved_reqs:
        logger.debug(
            "pex_builder=uv: using %d pinned packages from lockfile with --no-deps for %s",
            len(all_resolved_reqs),
            uv_request.description,
        )
    else:
        logger.debug(
            "pex_builder=uv: no lockfile available, using transitive uv resolution for %s",
            uv_request.description,
        )

    reqs_file = "__uv_requirements.txt"
    reqs_content = "\n".join(uv_reqs) + "\n"
    reqs_digest = await create_digest(CreateDigest([FileContent(reqs_file, reqs_content.encode())]))

    complete_pex_env = pex_env.in_sandbox(working_directory=None)
    uv_cache_dir = ".cache/uv_cache"
    uv_env = {
        **complete_pex_env.environment_dict(python_configured=True),
        "UV_CACHE_DIR": uv_cache_dir,
        "UV_NO_CONFIG": "1",
    }
    uv_caches = {
        **complete_pex_env.append_only_caches,
        "uv_cache": uv_cache_dir,
    }
    uv_tmpdir = "__uv_tmp"
    tmpdir_digest = await create_digest(CreateDigest([Directory(uv_tmpdir)]))

    python_path = uv_request.python_path
    venv_dir = _UV_VENV_DIR

    uv_input = await merge_digests(MergeDigests([downloaded_uv.digest, reqs_digest, tmpdir_digest]))

    # Step 1: Create venv with uv.
    venv_result = await execute_process_or_raise(
        **implicitly(
            Process(
                argv=(
                    downloaded_uv.exe,
                    "venv",
                    venv_dir,
                    "--python",
                    python_path,
                ),
                input_digest=uv_input,
                output_directories=(venv_dir,),
                env={**uv_env, "TMPDIR": uv_tmpdir},
                append_only_caches=uv_caches,
                description=f"Create uv venv for {uv_request.description}",
                level=LogLevel.DEBUG,
                cache_scope=ProcessCacheScope.SUCCESSFUL,
            )
        )
    )

    # Step 2: Install dependencies into the venv.
    install_input = await merge_digests(MergeDigests([uv_input, venv_result.output_digest]))

    install_argv: tuple[str, ...] = (
        downloaded_uv.exe,
        "pip",
        "install",
        "--python",
        os.path.join(venv_dir, "bin", "python"),
        "-r",
        reqs_file,
        *(("--no-deps",) if all_resolved_reqs else ()),
        *downloaded_uv.args_for_uv_pip_install,
    )

    uv_install_result = await execute_process_or_raise(
        **implicitly(
            Process(
                argv=install_argv,
                input_digest=install_input,
                output_directories=(venv_dir,),
                env={**uv_env, "TMPDIR": uv_tmpdir},
                append_only_caches=uv_caches,
                description=f"uv pip install for {uv_request.description}",
                level=LogLevel.DEBUG,
                cache_scope=ProcessCacheScope.SUCCESSFUL,
            )
        )
    )

    return _UvVenvResult(
        venv_digest=uv_install_result.output_digest,
    )


@dataclass
class _BuildPexRequirementsSetup:
    digests: list[Digest]
    argv: list[str]
    concurrency_available: int


@dataclass(frozen=True)
class PexRequirementsInfo:
    req_strings: tuple[str, ...]
    find_links: tuple[str, ...]


@rule
async def get_req_strings(pex_reqs: PexRequirements) -> PexRequirementsInfo:
    addrs: list[Address] = []
    specs: list[str] = []
    req_strings: list[str] = []
    find_links: set[str] = set()
    for req_str_or_addr in pex_reqs.req_strings_or_addrs:
        if isinstance(req_str_or_addr, Address):
            addrs.append(req_str_or_addr)
        else:
            assert isinstance(req_str_or_addr, str)
            # Require a `//` prefix, to distinguish address specs from
            # local or VCS requirements.
            if req_str_or_addr.startswith(os.path.sep * 2):
                specs.append(req_str_or_addr)
            else:
                req_strings.append(req_str_or_addr)
    if specs:
        addrs_from_specs = await resolve_unparsed_address_inputs(
            UnparsedAddressInputs(
                specs,
                owning_address=None,
                description_of_origin=pex_reqs.description_of_origin,
            ),
            **implicitly(),
        )
        addrs.extend(addrs_from_specs)
    if addrs:
        transitive_targets = await transitive_targets_get(
            TransitiveTargetsRequest(addrs), **implicitly()
        )
        req_strings.extend(
            PexRequirements.req_strings_from_requirement_fields(
                tgt[PythonRequirementsField]
                for tgt in transitive_targets.closure
                if tgt.has_field(PythonRequirementsField)
            )
        )
        find_links.update(
            find_links
            for tgt in transitive_targets.closure
            if tgt.has_field(PythonRequirementFindLinksField)
            for find_links in tgt[PythonRequirementFindLinksField].value or ()
        )
    return PexRequirementsInfo(tuple(sorted(req_strings)), tuple(sorted(find_links)))


async def _get_entire_lockfile_and_requirements(
    requirements: EntireLockfile | PexRequirements,
) -> tuple[LoadedLockfile | None, tuple[str, ...]]:
    lockfile: Lockfile | None = None
    complete_req_strings: tuple[str, ...] = tuple()
    # TODO: This is clunky, but can be simplified once we get rid of old-style tool
    #  lockfiles, because we can unify EntireLockfile and Resolve.
    if isinstance(requirements, EntireLockfile):
        complete_req_strings = requirements.complete_req_strings or tuple()
        lockfile = requirements.lockfile
    elif (
        isinstance(requirements.from_superset, Resolve)
        and requirements.from_superset.use_entire_lockfile
    ):
        lockfile = await get_lockfile_for_resolve(requirements.from_superset, **implicitly())
    if not lockfile:
        return None, complete_req_strings
    loaded_lockfile = await load_lockfile(LoadedLockfileRequest(lockfile), **implicitly())
    return loaded_lockfile, complete_req_strings


@rule
async def _setup_pex_requirements(
    request: PexRequest, python_setup: PythonSetup
) -> _BuildPexRequirementsSetup:
    resolve_name: str | None
    if isinstance(request.requirements, EntireLockfile):
        resolve_name = request.requirements.lockfile.resolve_name
    elif isinstance(request.requirements.from_superset, Resolve):
        resolve_name = request.requirements.from_superset.name
    else:
        # This implies that, currently, per-resolve options are only configurable for resolves.
        # However, if no resolve is specified, we will still load options that apply to every
        # resolve, like `[python-repos].indexes`.
        resolve_name = None
    resolve_config = await determine_resolve_pex_config(
        ResolvePexConfigRequest(resolve_name), **implicitly()
    )

    pex_lock_resolver_args = list(resolve_config.pex_args())
    pip_resolver_args = [*resolve_config.pex_args(), "--resolver-version", "pip-2020-resolver"]

    loaded_lockfile, complete_req_strings = await _get_entire_lockfile_and_requirements(
        request.requirements
    )
    if loaded_lockfile:
        argv = (
            ["--lock", loaded_lockfile.lockfile_path, *pex_lock_resolver_args]
            if loaded_lockfile.is_pex_native
            # We use pip to resolve a requirements.txt pseudo-lockfile, possibly with hashes.
            else [
                "--requirement",
                loaded_lockfile.lockfile_path,
                "--no-transitive",
                *pip_resolver_args,
            ]
        )
        if loaded_lockfile.metadata and complete_req_strings:
            validate_metadata(
                loaded_lockfile.metadata,
                request.interpreter_constraints,
                loaded_lockfile.original_lockfile,
                complete_req_strings,
                # We're using the entire lockfile, so there is no Pex subsetting operation we
                # can delegate requirement validation to.  So we do our naive string-matching
                # validation.
                validate_consumed_req_strings=True,
                python_setup=python_setup,
                resolve_config=resolve_config,
            )

        return _BuildPexRequirementsSetup(
            [loaded_lockfile.lockfile_digest], argv, loaded_lockfile.requirement_estimate
        )

    assert isinstance(request.requirements, PexRequirements)
    reqs_info = await get_req_strings(request.requirements)

    # TODO: This is not the best heuristic for available concurrency, since the
    # requirements almost certainly have transitive deps which also need building, but it
    # is better than using something hardcoded.
    concurrency_available = len(reqs_info.req_strings)

    if isinstance(request.requirements.from_superset, Pex):
        repository_pex = request.requirements.from_superset
        return _BuildPexRequirementsSetup(
            [repository_pex.digest],
            [*reqs_info.req_strings, "--pex-repository", repository_pex.name],
            concurrency_available,
        )

    elif isinstance(request.requirements.from_superset, Resolve):
        lockfile = await get_lockfile_for_resolve(
            request.requirements.from_superset, **implicitly()
        )
        loaded_lockfile = await load_lockfile(LoadedLockfileRequest(lockfile), **implicitly())

        # NB: This is also validated in the constructor.
        assert loaded_lockfile.is_pex_native
        if not reqs_info.req_strings:
            return _BuildPexRequirementsSetup([], [], concurrency_available)

        if loaded_lockfile.metadata:
            validate_metadata(
                loaded_lockfile.metadata,
                request.interpreter_constraints,
                loaded_lockfile.original_lockfile,
                consumed_req_strings=reqs_info.req_strings,
                # Don't validate user requirements when subsetting a resolve, as Pex's
                # validation during the subsetting is far more precise than our naive string
                # comparison. For example, if a lockfile was generated with `foo==1.2.3`
                # and we want to resolve `foo>=1.0.0` or just `foo` out of it, Pex will do
                # so successfully, while our naive validation would fail.
                validate_consumed_req_strings=False,
                python_setup=python_setup,
                resolve_config=resolve_config,
            )

        return _BuildPexRequirementsSetup(
            [loaded_lockfile.lockfile_digest],
            [
                *reqs_info.req_strings,
                "--lock",
                loaded_lockfile.lockfile_path,
                *pex_lock_resolver_args,
            ],
            concurrency_available,
        )

    # We use pip to perform a normal resolve.
    digests = []
    argv = [
        *reqs_info.req_strings,
        *pip_resolver_args,
        *(f"--find-links={find_links}" for find_links in reqs_info.find_links),
    ]
    if request.requirements.constraints_strings:
        constraints_file = "__constraints.txt"
        constraints_content = "\n".join(request.requirements.constraints_strings)
        digests.append(
            await create_digest(
                CreateDigest([FileContent(constraints_file, constraints_content.encode())])
            )
        )
        argv.extend(["--constraints", constraints_file])
    return _BuildPexRequirementsSetup(digests, argv, concurrency_available=concurrency_available)


@rule(level=LogLevel.DEBUG)
async def build_pex(
    request: PexRequest,
    python_setup: PythonSetup,
    pex_subsystem: PexSubsystem,
    pex_env: PexEnvironment,
) -> BuildPexResult:
    """Returns a PEX with the given settings."""

    if not request.python and not request.interpreter_constraints:
        # Blank ICs in the request means that the caller wants us to use the ICs configured
        # for the resolve (falling back to the global ICs).
        resolve_name = ""
        if isinstance(request.requirements, PexRequirements) and isinstance(
            request.requirements.from_superset, Resolve
        ):
            resolve_name = request.requirements.from_superset.name
        elif isinstance(request.requirements, EntireLockfile):
            resolve_name = request.requirements.lockfile.resolve_name

        if resolve_name:
            request = dataclasses.replace(
                request,
                interpreter_constraints=InterpreterConstraints(
                    python_setup.resolves_to_interpreter_constraints.get(
                        resolve_name,
                        python_setup.interpreter_constraints,
                    )
                ),
            )

    source_dir_name = "source_files"

    pex_python_setup_req = _determine_pex_python_and_platforms(request)
    requirements_setup_req = _setup_pex_requirements(**implicitly({request: PexRequest}))
    sources_digest_as_subdir_req = add_prefix(
        AddPrefix(request.sources or EMPTY_DIGEST, source_dir_name)
    )
    if isinstance(request.requirements, PexRequirements):
        (
            pex_python_setup,
            requirements_setup,
            sources_digest_as_subdir,
            req_info,
        ) = await concurrently(
            pex_python_setup_req,
            requirements_setup_req,
            sources_digest_as_subdir_req,
            get_req_strings(request.requirements),
        )
        req_strings = req_info.req_strings
    else:
        pex_python_setup, requirements_setup, sources_digest_as_subdir = await concurrently(
            pex_python_setup_req,
            requirements_setup_req,
            sources_digest_as_subdir_req,
        )
        req_strings = ()

    # Experimental: build PEX via uv + --venv-repository.
    # When opted in, we use uv to create a pre-populated venv and let PEX
    # package from it instead of resolving with pip.
    # See: https://github.com/pantsbuild/pants/issues/20679
    uv_venv_digest: Digest | None = None

    use_uv_builder = python_setup.pex_builder == PexBuilder.uv
    # uv builder only applies to non-internal PEXes with requirements and a
    # local interpreter (not cross-platform builds).
    if use_uv_builder and not request.internal_only:
        fallback_reason = _check_uv_preconditions(request, req_strings, pex_python_setup)
        if fallback_reason:
            logger.warning(fallback_reason)
        else:
            assert pex_python_setup.python is not None
            uv_result = await _build_uv_venv(
                _UvVenvRequest(
                    req_strings=req_strings,
                    requirements=request.requirements,
                    python_path=pex_python_setup.python.path,
                    description=request.description or request.output_filename,
                ),
                **implicitly(),
            )
            uv_venv_digest = uv_result.venv_digest

            # Replace requirements_setup: pass requirement strings + --venv-repository
            # so PEX subsets from the uv-populated venv instead of resolving with pip.
            requirements_setup = _BuildPexRequirementsSetup(
                digests=[],
                argv=[*req_strings, f"--venv-repository={_UV_VENV_DIR}"],
                concurrency_available=requirements_setup.concurrency_available,
            )
    elif use_uv_builder and request.internal_only:
        logger.debug(
            "pex_builder=uv: skipping for internal-only PEX %s. Using the default PEX/pip builder.",
            request.description or request.output_filename,
        )

    output_chroot = os.path.dirname(request.output_filename)
    if output_chroot:
        output_file = request.output_filename
        strip_output_chroot = False
    else:
        # In principle a cache should always be just a cache, but existing
        # tests in this repo make the assumption that they can look into a
        # still intact cache and see the same thing as was there before, which
        # requires this to be deterministic and not random.  adler32, because
        # it is in the stlib, fast, and doesn't need to be cryptographic.
        output_chroot = f"pex-dist-{zlib.adler32(request.output_filename.encode()):08x}"
        strip_output_chroot = True
        output_file = os.path.join(output_chroot, request.output_filename)

    argv = [
        "--output-file",
        output_file,
        *request.additional_args,
    ]

    if uv_venv_digest is not None:
        # When using --venv-repository, PEX does not allow any custom target
        # flags (--python, --interpreter-constraint, --platform). The target is
        # implicitly the venv interpreter.
        pass
    else:
        argv.extend(pex_python_setup.argv)

    if request.main is not None:
        argv.extend(request.main.iter_pex_args())
        if isinstance(request.main, Executable):
            # Unlike other MainSpecifiecation types (that can pass spec as-is to pex),
            # Executable must be an actual path relative to the sandbox.
            # request.main.spec is a python source file including its spec_path.
            # To make it relative to the sandbox, we strip the source root
            # and add the source_dir_name (sources get prefixed with that below).
            stripped = await strip_file_name(StrippedFileNameRequest(request.main.spec))
            argv.append(os.path.join(source_dir_name, stripped.value))

    argv.extend(
        f"--inject-args={shlex.quote(injected_arg)}" for injected_arg in request.inject_args
    )
    argv.extend(f"--inject-env={k}={v}" for k, v in sorted(request.inject_env.items()))

    # TODO(John Sirois): Right now any request requirements will shadow corresponding pex path
    #  requirements, which could lead to problems. Support shading python binaries.
    #  See: https://github.com/pantsbuild/pants/issues/9206
    if request.pex_path:
        argv.extend(["--pex-path", ":".join(pex.name for pex in request.pex_path)])

    if request.internal_only:
        # An internal-only runs on a single machine, and pre-installing wheels is wasted work in
        # that case (see https://github.com/pex-tool/pex/issues/2292#issuecomment-1854582647 for
        # analysis).
        argv.append("--no-pre-install-wheels")

    argv.append(f"--sources-directory={source_dir_name}")

    # Include any additional arguments and input digests required by the requirements.
    argv.extend(requirements_setup.argv)

    merged_digest = await merge_digests(
        MergeDigests(
            (
                request.complete_platforms.digest,
                sources_digest_as_subdir,
                request.additional_inputs,
                *requirements_setup.digests,
                *(pex.digest for pex in request.pex_path),
                *([uv_venv_digest] if uv_venv_digest else []),
            )
        )
    )

    argv.extend(["--layout", request.layout.value])

    result = await fallible_to_exec_result_or_raise(
        **implicitly(
            PexCliProcess(
                subcommand=(),
                extra_args=argv,
                additional_input_digest=merged_digest,
                description=_build_pex_description(request, req_strings, python_setup.resolves),
                output_files=None,
                output_directories=[output_chroot],
                concurrency_available=requirements_setup.concurrency_available,
                cache_scope=request.cache_scope,
            )
        )
    )

    maybe_log_pex_stderr(result.stderr, pex_subsystem.verbosity)

    if strip_output_chroot:
        output_digest = await remove_prefix(RemovePrefix(result.output_digest, output_chroot))
    else:
        output_digest = result.output_digest

    digest = (
        await merge_digests(
            MergeDigests((output_digest, *(pex.digest for pex in request.pex_path)))
        )
        if request.pex_path
        else output_digest
    )

    return BuildPexResult(
        result=result,
        pex_filename=request.output_filename,
        digest=digest,
        python=pex_python_setup.python,
    )


def _build_pex_description(
    request: PexRequest, req_strings: Sequence[str], resolve_to_lockfile: Mapping[str, str]
) -> str:
    if request.description:
        return request.description

    if isinstance(request.requirements, EntireLockfile):
        lockfile = request.requirements.lockfile
        desc_suffix = f"from {lockfile.url}"
    else:
        if not req_strings:
            return f"Building {request.output_filename}"
        elif isinstance(request.requirements.from_superset, Pex):
            repo_pex = request.requirements.from_superset.name
            return softwrap(
                f"""
                Extracting {pluralize(len(req_strings), "requirement")}
                to build {request.output_filename} from {repo_pex}:
                {", ".join(req_strings)}
                """
            )
        elif isinstance(request.requirements.from_superset, Resolve):
            # At this point we know this is a valid user resolve, so we can assume
            # it's available in the dict. Nonetheless we use get() so that any weird error
            # here gives a bad message rather than an outright crash.
            lockfile_path = resolve_to_lockfile.get(request.requirements.from_superset.name, "")
            return softwrap(
                f"""
                Building {pluralize(len(req_strings), "requirement")}
                for {request.output_filename} from the {lockfile_path} resolve:
                {", ".join(req_strings)}
                """
            )
        else:
            desc_suffix = softwrap(
                f"""
                with {pluralize(len(req_strings), "requirement")}:
                {", ".join(req_strings)}
                """
            )
    return f"Building {request.output_filename} {desc_suffix}"


@rule
async def create_pex(request: PexRequest) -> Pex:
    result = await build_pex(request, **implicitly())
    return result.create_pex()


@rule
async def create_optional_pex(request: OptionalPexRequest) -> OptionalPex:
    if request.maybe_pex_request is None:
        return OptionalPex(None)
    result = await create_pex(request.maybe_pex_request)
    return OptionalPex(result)


@dataclass(frozen=True)
class Script:
    path: PurePath

    @property
    def argv0(self) -> str:
        return f"./{self.path}" if self.path.parent == PurePath() else str(self.path)


@dataclass(frozen=True)
class VenvScript:
    script: Script
    content: FileContent


@dataclass(frozen=True)
class VenvScriptWriter:
    complete_pex_env: CompletePexEnvironment
    pex: Pex
    venv_dir: PurePath

    @classmethod
    def create(
        cls, complete_pex_env: CompletePexEnvironment, pex: Pex, venv_rel_dir: PurePath
    ) -> VenvScriptWriter:
        # N.B.: We don't know the working directory that will be used in any given
        # invocation of the venv scripts; so we deal with working_directory once in an
        # `adjust_relative_paths` function inside the script to save rule authors from having to do
        # CWD offset math in every rule for all the relative paths their process depends on.
        venv_dir = complete_pex_env.pex_root / venv_rel_dir
        return cls(complete_pex_env=complete_pex_env, pex=pex, venv_dir=venv_dir)

    def _create_venv_script(
        self,
        bash: BashBinary,
        *,
        script_path: PurePath,
        venv_executable: PurePath,
    ) -> VenvScript:
        env_vars = (
            f"{name}={shlex.quote(value)}"
            for name, value in self.complete_pex_env.environment_dict(
                python_configured=True
            ).items()
        )

        target_venv_executable = shlex.quote(str(venv_executable))
        venv_dir = shlex.quote(str(self.venv_dir))
        execute_pex_args = " ".join(
            f"$(adjust_relative_paths {shlex.quote(arg)})"
            for arg in self.complete_pex_env.create_argv(self.pex.name, python=self.pex.python)
        )

        script = dedent(
            f"""\
            #!{bash.path}
            set -euo pipefail

            # N.B.: This relies on BASH_SOURCE which has been available since bash-3.0, released in
            # 2004. It will either contain the absolute path of the venv script or it will contain
            # the relative path from the CWD to the venv script. Either way, we know the venv script
            # parent directory is the sandbox root directory.
            SANDBOX_ROOT="${{BASH_SOURCE%/*}}"

            function adjust_relative_paths() {{
                local value0="$1"
                shift
                if [ "${{value0:0:1}}" == "/" ]; then
                    # Don't relativize absolute paths.
                    echo "${{value0}}" "$@"
                else
                    # N.B.: We convert all relative paths to paths relative to the sandbox root so
                    # this script works when run with a PWD set somewhere else than the sandbox
                    # root.
                    #
                    # There are two cases to consider. For the purposes of example, assume PWD is
                    # `/tmp/sandboxes/abc123/foo/bar`; i.e.: the rule API sets working_directory to
                    # `foo/bar`. Also assume `config/tool.yml` is the relative path in question.
                    #
                    # 1. If our BASH_SOURCE is  `/tmp/sandboxes/abc123/pex_shim.sh`; so our
                    #    SANDBOX_ROOT is `/tmp/sandboxes/abc123`, we calculate
                    #    `/tmp/sandboxes/abc123/config/tool.yml`.
                    # 2. If our BASH_SOURCE is instead `../../pex_shim.sh`; so our SANDBOX_ROOT is
                    #    `../..`, we calculate `../../config/tool.yml`.
                    echo "${{SANDBOX_ROOT}}/${{value0}}" "$@"
                fi
            }}

            export {" ".join(env_vars)}
            export PEX_ROOT="$(adjust_relative_paths ${{PEX_ROOT}})"

            execute_pex_args="{execute_pex_args}"
            target_venv_executable="$(adjust_relative_paths {target_venv_executable})"
            venv_dir="$(adjust_relative_paths {venv_dir})"

            # Let PEX_TOOLS invocations pass through to the original PEX file since venvs don't come
            # with tools support.
            if [ -n "${{PEX_TOOLS:-}}" ]; then
              exec ${{execute_pex_args}} "$@"
            fi

            # If the seeded venv has been removed from the PEX_ROOT, we re-seed from the original
            # `--venv` mode PEX file.
            if [ ! -e "${{venv_dir}}" ]; then
                PEX_INTERPRETER=1 ${{execute_pex_args}} -c ''
            fi

            exec "${{target_venv_executable}}" "$@"
            """
        )
        return VenvScript(
            script=Script(script_path),
            content=FileContent(path=str(script_path), content=script.encode(), is_executable=True),
        )

    def exe(self, bash: BashBinary) -> VenvScript:
        """Writes a safe shim for the venv's executable `pex` script."""
        script_path = PurePath(f"{self.pex.name}_pex_shim.sh")
        return self._create_venv_script(
            bash, script_path=script_path, venv_executable=self.venv_dir / "pex"
        )

    def bin(self, bash: BashBinary, name: str) -> VenvScript:
        """Writes a safe shim for an executable or script in the venv's `bin` directory."""
        script_path = PurePath(f"{self.pex.name}_bin_{name}_shim.sh")
        return self._create_venv_script(
            bash,
            script_path=script_path,
            venv_executable=self.venv_dir / "bin" / name,
        )

    def python(self, bash: BashBinary) -> VenvScript:
        """Writes a safe shim for the venv's python binary."""
        return self.bin(bash, "python")


@dataclass(frozen=True)
class VenvPex:
    digest: Digest
    append_only_caches: FrozenDict[str, str] | None
    pex_filename: str
    pex: Script
    python: Script
    bin: FrozenDict[str, Script]
    venv_rel_dir: str


@dataclass(frozen=True)
class VenvPexRequest:
    pex_request: PexRequest
    complete_pex_env: CompletePexEnvironment
    bin_names: tuple[str, ...] = ()
    site_packages_copies: bool = False

    def __init__(
        self,
        pex_request: PexRequest,
        complete_pex_env: CompletePexEnvironment,
        bin_names: Iterable[str] = (),
        site_packages_copies: bool = False,
    ) -> None:
        """A request for a PEX that runs in a venv and optionally exposes select venv `bin` scripts.

        :param pex_request: The details of the desired PEX.
        :param complete_pex_env: The complete PEX environment the pex will be run in.
        :param bin_names: The names of venv `bin` scripts to expose for execution.
        :param site_packages_copies: `True` to use copies (hardlinks when possible) of PEX
            dependencies when installing them in the venv site-packages directory. By default this
            is `False` and symlinks are used instead which is a win in the time and space dimensions
            but results in a non-standard venv structure that does trip up some libraries.
        """
        object.__setattr__(self, "pex_request", pex_request)
        object.__setattr__(self, "complete_pex_env", complete_pex_env)
        object.__setattr__(self, "bin_names", tuple(bin_names))
        object.__setattr__(self, "site_packages_copies", site_packages_copies)


@rule
async def wrap_venv_prex_request(
    pex_request: PexRequest, pex_environment: PexEnvironment
) -> VenvPexRequest:
    # Allow creating a VenvPex from a plain PexRequest when no extra bin scripts need to be exposed.
    return VenvPexRequest(pex_request, pex_environment.in_sandbox(working_directory=None))


@rule
async def create_venv_pex(
    request: VenvPexRequest, bash: BashBinary, pex_environment: PexEnvironment
) -> VenvPex:
    # VenvPex is motivated by improving performance of Python tools by eliminating traditional PEX
    # file startup overhead.
    #
    # To achieve the minimal overhead (on the order of 1ms) we discard:
    # 1. Using Pex default mode:
    #    Although this does reduce initial tool execution overhead, it still leaves a minimum
    #    O(100ms) of overhead per subsequent tool invocation. Fundamentally, Pex still needs to
    #    execute its `sys.path` isolation bootstrap code in this case.
    # 2. Using the Pex `venv` tool:
    #    The idea here would be to create a tool venv as a Process output and then use the tool
    #    venv as an input digest for all tool invocations. This was tried and netted ~500ms of
    #    overhead over raw venv use.
    #
    # Instead we use Pex's `--venv` mode. In this mode you can run the Pex file and it will create a
    # venv on the fly in the PEX_ROOT as needed. Since the PEX_ROOT is a named_cache, we avoid the
    # digest materialization overhead present in 2 above. Since the venv is naturally isolated we
    # avoid the `sys.path` isolation overhead of Pex itself present in 1 above.
    #
    # This does leave O(50ms) of overhead though for the PEX bootstrap code to detect an already
    # created venv in the PEX_ROOT and re-exec into it. To eliminate this overhead we execute the
    # `pex` venv script in the PEX_ROOT directly. This is not robust on its own though, since the
    # named caches store might be pruned at any time. To guard against that case we introduce a shim
    # bash script that checks to see if the `pex` venv script exists in the PEX_ROOT and re-creates
    # the PEX_ROOT venv if not. Using the shim script to run Python tools gets us down to the ~1ms
    # of overhead we currently enjoy.

    pex_request = request.pex_request
    seeded_venv_request = dataclasses.replace(
        pex_request,
        additional_args=pex_request.additional_args
        + (
            "--venv",
            "prepend",
            "--seed",
            "verbose",
            pex_environment.venv_site_packages_copies_option(
                use_copies=request.site_packages_copies
            ),
        ),
    )
    venv_pex_result = await build_pex(seeded_venv_request, **implicitly())
    # Pex verbose --seed mode outputs the absolute path of the PEX executable as well as the
    # absolute path of the PEX_ROOT.  In the --venv case this is the `pex` script in the venv root
    # directory.
    seed_info = json.loads(venv_pex_result.result.stdout.decode())
    abs_pex_root = PurePath(seed_info["pex_root"])
    abs_pex_path = PurePath(seed_info["pex"])
    venv_rel_dir = abs_pex_path.relative_to(abs_pex_root).parent

    venv_script_writer = VenvScriptWriter.create(
        complete_pex_env=request.complete_pex_env,
        pex=venv_pex_result.create_pex(),
        venv_rel_dir=venv_rel_dir,
    )
    pex = venv_script_writer.exe(bash)
    python = venv_script_writer.python(bash)
    scripts = {bin_name: venv_script_writer.bin(bash, bin_name) for bin_name in request.bin_names}
    scripts_digest = await create_digest(
        CreateDigest(
            (
                pex.content,
                python.content,
                *(venv_script.content for venv_script in scripts.values()),
            )
        )
    )
    input_digest = await merge_digests(
        MergeDigests((venv_script_writer.pex.digest, scripts_digest))
    )
    append_only_caches = (
        venv_pex_result.python.append_only_caches if venv_pex_result.python else None
    )

    return VenvPex(
        digest=input_digest,
        append_only_caches=append_only_caches,
        pex_filename=venv_pex_result.pex_filename,
        pex=pex.script,
        python=python.script,
        bin=FrozenDict((bin_name, venv_script.script) for bin_name, venv_script in scripts.items()),
        venv_rel_dir=venv_rel_dir.as_posix(),
    )


@dataclass(frozen=True)
class PexProcess:
    pex: Pex
    argv: tuple[str, ...]
    description: str = dataclasses.field(compare=False)
    level: LogLevel
    input_digest: Digest | None
    working_directory: str | None
    extra_env: FrozenDict[str, str]
    output_files: tuple[str, ...] | None
    output_directories: tuple[str, ...] | None
    timeout_seconds: int | None
    execution_slot_variable: str | None
    concurrency_available: int
    cache_scope: ProcessCacheScope

    def __init__(
        self,
        pex: Pex,
        *,
        description: str,
        argv: Iterable[str] = (),
        level: LogLevel = LogLevel.INFO,
        input_digest: Digest | None = None,
        working_directory: str | None = None,
        extra_env: Mapping[str, str] | None = None,
        output_files: Iterable[str] | None = None,
        output_directories: Iterable[str] | None = None,
        timeout_seconds: int | None = None,
        execution_slot_variable: str | None = None,
        concurrency_available: int = 0,
        cache_scope: ProcessCacheScope = ProcessCacheScope.SUCCESSFUL,
    ) -> None:
        object.__setattr__(self, "pex", pex)
        object.__setattr__(self, "argv", tuple(argv))
        object.__setattr__(self, "description", description)
        object.__setattr__(self, "level", level)
        object.__setattr__(self, "input_digest", input_digest)
        object.__setattr__(self, "working_directory", working_directory)
        object.__setattr__(self, "extra_env", FrozenDict(extra_env or {}))
        object.__setattr__(self, "output_files", tuple(output_files) if output_files else None)
        object.__setattr__(
            self, "output_directories", tuple(output_directories) if output_directories else None
        )
        object.__setattr__(self, "timeout_seconds", timeout_seconds)
        object.__setattr__(self, "execution_slot_variable", execution_slot_variable)
        object.__setattr__(self, "concurrency_available", concurrency_available)
        object.__setattr__(self, "cache_scope", cache_scope)


@rule
async def setup_pex_process(request: PexProcess, pex_environment: PexEnvironment) -> Process:
    pex = request.pex
    complete_pex_env = pex_environment.in_sandbox(working_directory=request.working_directory)
    argv = complete_pex_env.create_argv(pex.name, *request.argv, python=pex.python)
    env = {
        **complete_pex_env.environment_dict(python_configured=pex.python is not None),
        **request.extra_env,
    }
    input_digest = (
        await merge_digests(MergeDigests((pex.digest, request.input_digest)))
        if request.input_digest
        else pex.digest
    )
    append_only_caches = (
        request.pex.python.append_only_caches if request.pex.python else FrozenDict({})
    )
    return Process(
        argv,
        description=request.description,
        level=request.level,
        input_digest=input_digest,
        working_directory=request.working_directory,
        env=env,
        output_files=request.output_files,
        output_directories=request.output_directories,
        append_only_caches={
            **complete_pex_env.append_only_caches,
            **append_only_caches,
        },
        timeout_seconds=request.timeout_seconds,
        execution_slot_variable=request.execution_slot_variable,
        concurrency_available=request.concurrency_available,
        cache_scope=request.cache_scope,
    )


@dataclass(unsafe_hash=True)
class VenvPexProcess:
    venv_pex: VenvPex
    argv: tuple[str, ...]
    description: str = dataclasses.field(compare=False)
    level: LogLevel
    input_digest: Digest | None
    working_directory: str | None
    extra_env: FrozenDict[str, str]
    output_files: tuple[str, ...] | None
    output_directories: tuple[str, ...] | None
    timeout_seconds: int | None
    execution_slot_variable: str | None
    concurrency_available: int
    cache_scope: ProcessCacheScope
    append_only_caches: FrozenDict[str, str]

    def __init__(
        self,
        venv_pex: VenvPex,
        *,
        description: str,
        argv: Iterable[str] = (),
        level: LogLevel = LogLevel.INFO,
        input_digest: Digest | None = None,
        working_directory: str | None = None,
        extra_env: Mapping[str, str] | None = None,
        output_files: Iterable[str] | None = None,
        output_directories: Iterable[str] | None = None,
        timeout_seconds: int | None = None,
        execution_slot_variable: str | None = None,
        concurrency_available: int = 0,
        cache_scope: ProcessCacheScope = ProcessCacheScope.SUCCESSFUL,
        append_only_caches: Mapping[str, str] | None = None,
    ) -> None:
        object.__setattr__(self, "venv_pex", venv_pex)
        object.__setattr__(self, "argv", tuple(argv))
        object.__setattr__(self, "description", description)
        object.__setattr__(self, "level", level)
        object.__setattr__(self, "input_digest", input_digest)
        object.__setattr__(self, "working_directory", working_directory)
        object.__setattr__(self, "extra_env", FrozenDict(extra_env or {}))
        object.__setattr__(self, "output_files", tuple(output_files) if output_files else None)
        object.__setattr__(
            self, "output_directories", tuple(output_directories) if output_directories else None
        )
        object.__setattr__(self, "timeout_seconds", timeout_seconds)
        object.__setattr__(self, "execution_slot_variable", execution_slot_variable)
        object.__setattr__(self, "concurrency_available", concurrency_available)
        object.__setattr__(self, "cache_scope", cache_scope)
        object.__setattr__(self, "append_only_caches", FrozenDict(append_only_caches or {}))


@rule
async def setup_venv_pex_process(
    request: VenvPexProcess, pex_environment: PexEnvironment
) -> Process:
    venv_pex = request.venv_pex
    pex_bin = (
        os.path.relpath(venv_pex.pex.argv0, request.working_directory)
        if request.working_directory
        else venv_pex.pex.argv0
    )
    argv = (pex_bin, *request.argv)
    input_digest = (
        await merge_digests(MergeDigests((venv_pex.digest, request.input_digest)))
        if request.input_digest
        else venv_pex.digest
    )
    append_only_caches: FrozenDict[str, str] = FrozenDict(
        **pex_environment.in_sandbox(
            working_directory=request.working_directory
        ).append_only_caches,
        **request.append_only_caches,
        **(FrozenDict({}) if venv_pex.append_only_caches is None else venv_pex.append_only_caches),
    )
    return Process(
        argv=argv,
        description=request.description,
        level=request.level,
        input_digest=input_digest,
        working_directory=request.working_directory,
        env=request.extra_env,
        output_files=request.output_files,
        output_directories=request.output_directories,
        append_only_caches=append_only_caches,
        timeout_seconds=request.timeout_seconds,
        execution_slot_variable=request.execution_slot_variable,
        concurrency_available=request.concurrency_available,
        cache_scope=request.cache_scope,
    )


@dataclass(frozen=True)
class PexDistributionInfo:
    """Information about an individual distribution in a PEX file, as reported by `PEX_TOOLS=1
    repository info -v`."""

    project_name: str
    version: packaging.version.Version
    requires_python: packaging.specifiers.SpecifierSet | None
    # Note: These are parsed from metadata written by the pex tool, and are always
    #   a valid packaging.requirements.Requirement.
    requires_dists: tuple[Requirement, ...]


DefaultT = TypeVar("DefaultT")


class PexResolveInfo(Collection[PexDistributionInfo]):
    """Information about all distributions resolved in a PEX file, as reported by `PEX_TOOLS=1
    repository info -v`."""

    def find(
        self, name: str, default: DefaultT | None = None
    ) -> PexDistributionInfo | DefaultT | None:
        """Returns the PexDistributionInfo with the given name, first one wins."""
        try:
            return next(info for info in self if info.project_name == name)
        except StopIteration:
            return default


def parse_repository_info(repository_info: str) -> PexResolveInfo:
    def iter_dist_info() -> Iterator[PexDistributionInfo]:
        for line in repository_info.splitlines():
            info = json.loads(line)
            requires_python = info["requires_python"]
            yield PexDistributionInfo(
                project_name=info["project_name"],
                version=packaging.version.Version(info["version"]),
                requires_python=(
                    packaging.specifiers.SpecifierSet(requires_python)
                    if requires_python is not None
                    else None
                ),
                requires_dists=tuple(Requirement(req) for req in sorted(info["requires_dists"])),
            )

    return PexResolveInfo(sorted(iter_dist_info(), key=lambda dist: dist.project_name))


@rule
async def determine_venv_pex_resolve_info(venv_pex: VenvPex) -> PexResolveInfo:
    process_result = await fallible_to_exec_result_or_raise(
        **implicitly(
            VenvPexProcess(
                venv_pex,
                argv=["repository", "info", "-v"],
                extra_env={"PEX_TOOLS": "1"},
                input_digest=venv_pex.digest,
                description=f"Determine distributions found in {venv_pex.pex_filename}",
                level=LogLevel.DEBUG,
            )
        )
    )
    return parse_repository_info(process_result.stdout.decode())


@rule
async def determine_pex_resolve_info(pex_pex: PexPEX, pex: Pex) -> PexResolveInfo:
    process_result = await fallible_to_exec_result_or_raise(
        **implicitly(
            PexProcess(
                pex=Pex(digest=pex_pex.digest, name=pex_pex.exe, python=pex.python),
                argv=[pex.name, "repository", "info", "-v"],
                input_digest=pex.digest,
                extra_env={"PEX_MODULE": "pex.tools"},
                description=f"Determine distributions found in {pex.name}",
                level=LogLevel.DEBUG,
            )
        )
    )
    return parse_repository_info(process_result.stdout.decode())


def rules():
    return [
        *collect_rules(),
        *pex_cli.rules(),
        *pex_requirements.rules(),
        *uv_subsystem.rules(),  # Also in register.py; engine deduplicates.
        *stripped_source_rules(),
    ]

pantsbuild / pants / 23745351279

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous