PEP 328

PEP 328
Title	PEP 328
Status	Final
Author	Marc-Andre Lemburg
Type	Standards Track
Created	2003-10-15
Python version	2.5

PEP 328 is a Python Enhancement Proposal that formalized changes to the import statement and package import semantics. The proposal reconciled legacy import behaviors with package-oriented designs from projects and institutions such as Zope, Google, Red Hat, Canonical, Mozilla Foundation while interacting with contemporary discussions among developers at Python Software Foundation, Guido van Rossum, ActiveState, NumPy, and Django. It influenced toolchains and ecosystems including setuptools, pip, Gentoo, Debian, and Fedora during transitions across versions like Python 2.3, Python 2.4, and Python 2.5.

Background and Rationale

PEP 328 arose amid debates involving figures and organizations such as Guido van Rossum, Barry Warsaw, Tim Peters, Fred L. Drake Jr., Marc-Andre Lemburg, and groups at PSF, Zope Corporation, Enthought, Intel, IBM about namespace clarity and package layout. Legacy import behaviors conflicted with large codebases maintained by projects like Twisted, Plone, Mercurial, OpenStack, and Trac, and with packaging standards advanced by PEP 302, PEP 328 sought to align import mechanics with expectations from ecosystems represented by SourceForge, GitHub, Bitbucket, and distributions such as Ubuntu. Historical precedents and related proposals from PEP 241 and PEP 328 discussions intersected with module loader work by contributors linked to PEP 302, PEP 314 and influenced governance deliberations at Python Enhancement Proposals steering groups.

Specification and Syntax Changes

The proposal specified syntax alterations to the import statement and the semantics of relative imports, engaging maintainers from CPython, Jython, IronPython, PyPy, and Stackless Python. It introduced explicit relative import syntax using leading dots to indicate current and parent package levels, an approach debated alongside designs from PEP 328-related discussions, and compared to behaviors in languages and platforms such as Java, C#, Perl, and Ruby. The specification defined how package __path__ resolution and absolute imports should interact with module finders and loaders as implemented by teams at ActiveState and contributors from Google Summer of Code and Open Source Initiative projects, coordinating with standards used by setuptools, distutils, and pkg_resources.

Impact on Import Semantics and Module Resolution

Adoption changed resolution order and name-binding semantics, affecting runtime interactions in projects like Django, Twisted, NumPy, SciPy, Pandas, Matplotlib, and frameworks maintained by organizations such as Canonical, Red Hat, and Oracle. Module search behaviors influenced packaging practices endorsed by PyPI, Wheel, Egg, and distribution maintainers at Debian, Fedora, and Gentoo, while loader APIs implemented by contributors to PEP 302 and extension modules written for CPython shifted expectations. The revised semantics impacted interoperability with foreign runtimes and embedding contexts used by Android, iOS, Windows, macOS, and server platforms maintained by companies like Amazon Web Services and Heroku.

Compatibility and Migration Considerations

Migration guidance cited effects on large applications maintained by teams at Google, Facebook, Microsoft Research, NASA, and academic groups at institutions such as MIT, Stanford University, University of Cambridge, and University of California, Berkeley. Backward compatibility strategies involved transitional flags, tooling changes in setuptools and pip, and vendor packagers at Red Hat and Canonical coordinating upgrades. The proposal prompted linting and static analysis tool updates used by projects like PyLint, PyChecker, Bandit, and integrated development environments such as PyCharm, Eclipse, Visual Studio, and Sublime Text.

Reception and Implementation in CPython

The PEP was implemented in CPython with review and commits by maintainers including contributors from Python Software Foundation, Kenneth Reitz, Raymond Hettinger, and Armin Ronacher in repositories hosted on platforms like SourceForge and later GitHub. Community response included endorsements from maintainers of Django, Zope, Twisted, and package authors in the Scientific Python ecosystem such as Travis Oliphant and Enthought engineers, while debates continued in mailing lists and issue trackers associated with PEP discussions, python-dev, and conferences like PyCon, EuroPython, SciPy, and OSCON.

Examples and Use Cases

Common use cases addressed package-local imports in applications like Django projects, libraries such as NumPy and Pandas, command-line tools maintained by Canonical and Red Hat, and plugins for systems like Sphinx, Mercurial, and Trac. The explicit relative import syntax clarified code in hierarchical packages used by research groups at Lawrence Berkeley National Laboratory, CERN, and Los Alamos National Laboratory, and in commercial products from IBM, Oracle, and Microsoft. Developers adopted patterns demonstrated in tutorials at conferences including PyCon, EuroPython, SciPy, and in documentation maintained by Python Software Foundation and organizations like Real Python.

Category:Python Enhancement Proposals