Source code for pip._internal.operations.prepare

"""Prepares a distribution for installation
"""

# The following comment should be removed at some point in the future.
# mypy: strict-optional=False

import logging
import mimetypes
import os
import shutil
from typing import Dict, Iterable, List, Optional

from pip._vendor.packaging.utils import canonicalize_name

from pip._internal.distributions import make_distribution_for_install_requirement
from pip._internal.distributions.installed import InstalledDistribution
from pip._internal.exceptions import (
    DirectoryUrlHashUnsupported,
    HashMismatch,
    HashUnpinned,
    InstallationError,
    NetworkConnectionError,
    PreviousBuildDirError,
    VcsHashUnsupported,
)
from pip._internal.index.package_finder import PackageFinder
from pip._internal.metadata import BaseDistribution
from pip._internal.models.link import Link
from pip._internal.models.wheel import Wheel
from pip._internal.network.download import BatchDownloader, Downloader
from pip._internal.network.lazy_wheel import (
    HTTPRangeRequestUnsupported,
    dist_from_wheel_url,
)
from pip._internal.network.session import PipSession
from pip._internal.operations.build.build_tracker import BuildTracker
from pip._internal.req.req_install import InstallRequirement
from pip._internal.utils.hashes import Hashes, MissingHashes
from pip._internal.utils.logging import indent_log
from pip._internal.utils.misc import display_path, hide_url, is_installable_dir
from pip._internal.utils.temp_dir import TempDirectory
from pip._internal.utils.unpacking import unpack_file
from pip._internal.vcs import vcs

logger = logging.getLogger(__name__)


def _get_prepared_distribution(
    req: InstallRequirement,
    build_tracker: BuildTracker,
    finder: PackageFinder,
    build_isolation: bool,
    check_build_deps: bool,
) -> BaseDistribution:
    """Prepare a distribution for installation."""
    abstract_dist = make_distribution_for_install_requirement(req)
    with build_tracker.track(req):
        abstract_dist.prepare_distribution_metadata(
            finder, build_isolation, check_build_deps
        )
    return abstract_dist.get_metadata_distribution()


def unpack_vcs_link(link: Link, location: str, verbosity: int) -> None:
    vcs_backend = vcs.get_backend_for_scheme(link.scheme)
    assert vcs_backend is not None
    vcs_backend.unpack(location, url=hide_url(link.url), verbosity=verbosity)


class File:
    def __init__(self, path: str, content_type: Optional[str]) -> None:
        self.path = path
        if content_type is None:
            self.content_type = mimetypes.guess_type(path)[0]
        else:
            self.content_type = content_type


def get_http_url(
    link: Link,
    download: Downloader,
    download_dir: Optional[str] = None,
    hashes: Optional[Hashes] = None,
) -> File:
    temp_dir = TempDirectory(kind="unpack", globally_managed=True)
    # If a download dir is specified, is the file already downloaded there?
    already_downloaded_path = None
    if download_dir:
        already_downloaded_path = _check_download_dir(link, download_dir, hashes)

    if already_downloaded_path:
        from_path = already_downloaded_path
        content_type = None
    else:
        # let's download to a tmp dir
        from_path, content_type = download(link, temp_dir.path)
        if hashes:
            hashes.check_against_path(from_path)

    return File(from_path, content_type)


def get_file_url(
    link: Link, download_dir: Optional[str] = None, hashes: Optional[Hashes] = None
) -> File:
    """Get file and optionally check its hash."""
    # If a download dir is specified, is the file already there and valid?
    already_downloaded_path = None
    if download_dir:
        already_downloaded_path = _check_download_dir(link, download_dir, hashes)

    if already_downloaded_path:
        from_path = already_downloaded_path
    else:
        from_path = link.file_path

    # If --require-hashes is off, `hashes` is either empty, the
    # link's embedded hash, or MissingHashes; it is required to
    # match. If --require-hashes is on, we are satisfied by any
    # hash in `hashes` matching: a URL-based or an option-based
    # one; no internet-sourced hash will be in `hashes`.
    if hashes:
        hashes.check_against_path(from_path)
    return File(from_path, None)


[docs]def unpack_url( link: Link, location: str, download: Downloader, verbosity: int, download_dir: Optional[str] = None, hashes: Optional[Hashes] = None, ) -> Optional[File]: """Unpack link into location, downloading if required. :param hashes: A Hashes object, one of whose embedded hashes must match, or HashMismatch will be raised. If the Hashes is empty, no matches are required, and unhashable types of requirements (like VCS ones, which would ordinarily raise HashUnsupported) are allowed. """ # non-editable vcs urls if link.is_vcs: unpack_vcs_link(link, location, verbosity=verbosity) return None assert not link.is_existing_dir() # file urls if link.is_file: file = get_file_url(link, download_dir, hashes=hashes) # http urls else: file = get_http_url( link, download, download_dir, hashes=hashes, ) # unpack the archive to the build dir location. even when only downloading # archives, they have to be unpacked to parse dependencies, except wheels if not link.is_wheel: unpack_file(file.path, location, file.content_type) return file
def _check_download_dir( link: Link, download_dir: str, hashes: Optional[Hashes] ) -> Optional[str]: """Check download_dir for previously downloaded file with correct hash If a correct file is found return its path else None """ download_path = os.path.join(download_dir, link.filename) if not os.path.exists(download_path): return None # If already downloaded, does its hash match? logger.info("File was already downloaded %s", download_path) if hashes: try: hashes.check_against_path(download_path) except HashMismatch: logger.warning( "Previously-downloaded file %s has bad hash. Re-downloading.", download_path, ) os.unlink(download_path) return None return download_path
[docs]class RequirementPreparer: """Prepares a Requirement""" def __init__( self, build_dir: str, download_dir: Optional[str], src_dir: str, build_isolation: bool, check_build_deps: bool, build_tracker: BuildTracker, session: PipSession, progress_bar: str, finder: PackageFinder, require_hashes: bool, use_user_site: bool, lazy_wheel: bool, verbosity: int, ) -> None: super().__init__() self.src_dir = src_dir self.build_dir = build_dir self.build_tracker = build_tracker self._session = session self._download = Downloader(session, progress_bar) self._batch_download = BatchDownloader(session, progress_bar) self.finder = finder # Where still-packed archives should be written to. If None, they are # not saved, and are deleted immediately after unpacking. self.download_dir = download_dir # Is build isolation allowed? self.build_isolation = build_isolation # Should check build dependencies? self.check_build_deps = check_build_deps # Should hash-checking be required? self.require_hashes = require_hashes # Should install in user site-packages? self.use_user_site = use_user_site # Should wheels be downloaded lazily? self.use_lazy_wheel = lazy_wheel # How verbose should underlying tooling be? self.verbosity = verbosity # Memoized downloaded files, as mapping of url: path. self._downloaded: Dict[str, str] = {} # Previous "header" printed for a link-based InstallRequirement self._previous_requirement_header = ("", "")
[docs] def _get_linked_req_hashes(self, req: InstallRequirement) -> Hashes: # By the time this is called, the requirement's link should have # been checked so we can tell what kind of requirements req is # and raise some more informative errors than otherwise. # (For example, we can raise VcsHashUnsupported for a VCS URL # rather than HashMissing.) if not self.require_hashes: return req.hashes(trust_internet=True) # We could check these first 2 conditions inside unpack_url # and save repetition of conditions, but then we would # report less-useful error messages for unhashable # requirements, complaining that there's no hash provided. if req.link.is_vcs: raise VcsHashUnsupported() if req.link.is_existing_dir(): raise DirectoryUrlHashUnsupported() # Unpinned packages are asking for trouble when a new version # is uploaded. This isn't a security check, but it saves users # a surprising hash mismatch in the future. # file:/// URLs aren't pinnable, so don't complain about them # not being pinned. if req.original_link is None and not req.is_pinned: raise HashUnpinned() # If known-good hashes are missing for this requirement, # shim it with a facade object that will provoke hash # computation and then raise a HashMissing exception # showing the user what the hash should be. return req.hashes(trust_internet=False) or MissingHashes()
[docs] def _fetch_metadata_using_lazy_wheel( self, link: Link, ) -> Optional[BaseDistribution]: """Fetch metadata using lazy wheel, if possible.""" if not self.use_lazy_wheel: return None if self.require_hashes: logger.debug("Lazy wheel is not used as hash checking is required") return None if link.is_file or not link.is_wheel: logger.debug( "Lazy wheel is not used as %r does not points to a remote wheel", link, ) return None wheel = Wheel(link.filename) name = canonicalize_name(wheel.name) logger.info( "Obtaining dependency information from %s %s", name, wheel.version, ) url = link.url.split("#", 1)[0] try: return dist_from_wheel_url(name, url, self._session) except HTTPRangeRequestUnsupported: logger.debug("%s does not support range requests", url) return None
[docs] def _complete_partial_requirements( self, partially_downloaded_reqs: Iterable[InstallRequirement], parallel_builds: bool = False, ) -> None: """Download any requirements which were only fetched by metadata.""" # Download to a temporary directory. These will be copied over as # needed for downstream 'download', 'wheel', and 'install' commands. temp_dir = TempDirectory(kind="unpack", globally_managed=True).path # Map each link to the requirement that owns it. This allows us to set # `req.local_file_path` on the appropriate requirement after passing # all the links at once into BatchDownloader. links_to_fully_download: Dict[Link, InstallRequirement] = {} for req in partially_downloaded_reqs: assert req.link links_to_fully_download[req.link] = req batch_download = self._batch_download( links_to_fully_download.keys(), temp_dir, ) for link, (filepath, _) in batch_download: logger.debug("Downloading link %s to %s", link, filepath) req = links_to_fully_download[link] req.local_file_path = filepath # This step is necessary to ensure all lazy wheels are processed # successfully by the 'download', 'wheel', and 'install' commands. for req in partially_downloaded_reqs: self._prepare_linked_requirement(req, parallel_builds)
[docs] def prepare_linked_requirement( self, req: InstallRequirement, parallel_builds: bool = False ) -> BaseDistribution: """Prepare a requirement to be obtained from req.link.""" assert req.link link = req.link self._log_preparing_link(req) with indent_log(): # Check if the relevant file is already available # in the download directory file_path = None if self.download_dir is not None and link.is_wheel: hashes = self._get_linked_req_hashes(req) file_path = _check_download_dir(req.link, self.download_dir, hashes) if file_path is not None: # The file is already available, so mark it as downloaded self._downloaded[req.link.url] = file_path else: # The file is not available, attempt to fetch only metadata wheel_dist = self._fetch_metadata_using_lazy_wheel(link) if wheel_dist is not None: req.needs_more_preparation = True return wheel_dist # None of the optimizations worked, fully prepare the requirement return self._prepare_linked_requirement(req, parallel_builds)
[docs] def prepare_linked_requirements_more( self, reqs: Iterable[InstallRequirement], parallel_builds: bool = False ) -> None: """Prepare linked requirements more, if needed.""" reqs = [req for req in reqs if req.needs_more_preparation] for req in reqs: # Determine if any of these requirements were already downloaded. if self.download_dir is not None and req.link.is_wheel: hashes = self._get_linked_req_hashes(req) file_path = _check_download_dir(req.link, self.download_dir, hashes) if file_path is not None: self._downloaded[req.link.url] = file_path req.needs_more_preparation = False # Prepare requirements we found were already downloaded for some # reason. The other downloads will be completed separately. partially_downloaded_reqs: List[InstallRequirement] = [] for req in reqs: if req.needs_more_preparation: partially_downloaded_reqs.append(req) else: self._prepare_linked_requirement(req, parallel_builds) # TODO: separate this part out from RequirementPreparer when the v1 # resolver can be removed! self._complete_partial_requirements( partially_downloaded_reqs, parallel_builds=parallel_builds, )
[docs] def _prepare_linked_requirement( self, req: InstallRequirement, parallel_builds: bool ) -> BaseDistribution: assert req.link link = req.link self._ensure_link_req_src_dir(req, parallel_builds) hashes = self._get_linked_req_hashes(req) if link.is_existing_dir(): local_file = None elif link.url not in self._downloaded: try: local_file = unpack_url( link, req.source_dir, self._download, self.verbosity, self.download_dir, hashes, ) except NetworkConnectionError as exc: raise InstallationError( "Could not install requirement {} because of HTTP " "error {} for URL {}".format(req, exc, link) ) else: file_path = self._downloaded[link.url] if hashes: hashes.check_against_path(file_path) local_file = File(file_path, content_type=None) # For use in later processing, # preserve the file path on the requirement. if local_file: req.local_file_path = local_file.path dist = _get_prepared_distribution( req, self.build_tracker, self.finder, self.build_isolation, self.check_build_deps, ) return dist
[docs] def save_linked_requirement(self, req: InstallRequirement) -> None: assert self.download_dir is not None assert req.link is not None link = req.link if link.is_vcs or (link.is_existing_dir() and req.editable): # Make a .zip of the source_dir we already created. req.archive(self.download_dir) return if link.is_existing_dir(): logger.debug( "Not copying link to destination directory " "since it is a directory: %s", link, ) return if req.local_file_path is None: # No distribution was downloaded for this requirement. return download_location = os.path.join(self.download_dir, link.filename) if not os.path.exists(download_location): shutil.copy(req.local_file_path, download_location) download_path = display_path(download_location) logger.info("Saved %s", download_path)
[docs] def prepare_editable_requirement( self, req: InstallRequirement, ) -> BaseDistribution: """Prepare an editable requirement.""" assert req.editable, "cannot prepare a non-editable req as editable" logger.info("Obtaining %s", req) with indent_log(): if self.require_hashes: raise InstallationError( "The editable requirement {} cannot be installed when " "requiring hashes, because there is no single file to " "hash.".format(req) ) req.ensure_has_source_dir(self.src_dir) req.update_editable() dist = _get_prepared_distribution( req, self.build_tracker, self.finder, self.build_isolation, self.check_build_deps, ) req.check_if_exists(self.use_user_site) return dist
[docs] def prepare_installed_requirement( self, req: InstallRequirement, skip_reason: str, ) -> BaseDistribution: """Prepare an already-installed requirement.""" assert req.satisfied_by, "req should have been satisfied but isn't" assert skip_reason is not None, ( "did not get skip reason skipped but req.satisfied_by " "is set to {}".format(req.satisfied_by) ) logger.info( "Requirement %s: %s (%s)", skip_reason, req, req.satisfied_by.version ) with indent_log(): if self.require_hashes: logger.debug( "Since it is already installed, we are trusting this " "package without checking its hash. To ensure a " "completely repeatable environment, install into an " "empty virtualenv." ) return InstalledDistribution(req).get_metadata_distribution()