Since pip is a command-line-tool, it does not have an official, supported, importable API.

However, this does not mean that people haven't tried to import pip, usually to end up with much headache when pip's maintainers do routine refactoring.

The goal of this project is to provide an importable pip API, which is fully compliant with the recommended method of using pip from your program.

How? By providing an importable API that wraps command-line calls to pip, this library can be used as a drop-in replacement for existing uses of pip's internal API.

This goal means that any new API added here must have the following equivalents:

• some internal pip API (or combination of internal APIs)
• some CLI calls (or combination of CLI calls)

Any functionality that is not currently possible from internal pip API or CLI calls is out of scope.

You can install pip-api with either pip or with conda.

With pip:

python -m pip install pip-api

With conda:

conda install -c conda-forge pip-api

Not all commands are supported in all versions of pip and on all platforms. If the command you are trying to use is not compatible, pip_api will raise a pip_api.exceptions.Incompatible exception for your program to catch.

• pip_api.version()

  Returns the pip version as a string, e.g. "9.0.1"

• pip_api.installed_distributions(local=False)

  Returns a list of all installed distributions as a Distribution object with the following attributes:

  • Distribution.name (string): The name of the installed distribution
  • Distribution.version (packaging.version.Version): The version of the installed distribution
  • Distribution.location (string): The location of the installed distribution
  • Distribution.editable (bool): Whether the distribution is editable or not Optionally takes a local parameter to filter out globally-installed packages

• pip_api.parse_requirements(filename, options=None, include_invalid=False, strict_hashes=False)

  Takes a path to a filename of a Requirements file. Returns a mapping from package name to a pip_api.Requirement object (subclass of packaging.requirements.Requirement) with the following attributes:

  • Requirement.name (string): The name of the requirement.
  • Requirement.extras (set): A set of extras that the requirement specifies.
  • Requirement.specifier (packaging.specifiers.SpecifierSet): A SpecifierSet of the version specified by the requirement.
  • Requirement.marker (packaging.markers.Marker): A Marker of the marker for the requirement. Can be None.
  • Requirement.hashes (dict): A mapping of hashes for the requirement, corresponding to --hash=... options.
  • Requirement.editable (bool): Whether the requirement is editable, corresponding to -e ...
  • Requirement.filename (str): The filename that the requirement originates from.
  • Requirement.lineno (int): The source line that the requirement was parsed from.

  Optionally takes an options parameter to override the regex used to skip requirements lines. Optionally takes an include_invalid parameter to return an UnparsedRequirement in the event that a requirement cannot be parsed correctly. Optionally takes a strict_hashes parameter to require that all requirements have associated hashes.

• pip_api.hash(filename, algorithm='sha256')

  Returns the resulting hash digest as a string. Valid algorithm parameters are 'sha256', 'sha384', and 'sha512'

• pip_api.installed_distributions(local=False, paths=[])

  As described above, but with an extra optional paths parameter to provide a list of locations to look for installed distributions. Attempting to use the paths parameter with pip<19.2 will result in a PipError.

This library is in use by a number of other tools, including:

• pip-audit, to analyze dependencies for known vulnerabilities
• pytest-reqs, to compare requirements files with test dependencies
• hashin, to add hash pinning to requirements files
• ...and many more.