Cython not found, cython extensions will not be generated…

Error message

Cython not found, cython extensions will not be generated. To use cythonized extensions, pip install cython and run python setup.py build_ext --inplace

What it means

When the Cython package is not importable, Beam's setup.py catches ImportError and warns that cythonized extensions will not be built. The build falls back to cythonize returning an empty list, so the install succeeds but without optimized native extensions (e.g. Cython-based coders/transforms run in slower pure-Python mode).

Solutions

Install cython: pip install cython, then run python setup.py build_ext --inplace.
Install numpy as well (the cythonize wrapper needs it for include dirs): pip install cython numpy.
Prefer installing an official apache-beam wheel (pip install apache-beam) which ships pre-built extensions and skips this path entirely.

Example fix

// before
$ python setup.py build_ext --inplace   # cython missing
// after
$ pip install cython numpy
$ python setup.py build_ext --inplace

Defensive patterns

Strategy: fallback

Validate before calling

import importlib.util
if importlib.util.find_spec('cython') is None or importlib.util.find_spec('numpy') is None:
    print('Warning: extensions will be skipped; pip install cython numpy')

Try / catch

try:
    run_build()
except SystemExit:
    print('Falling back to pure-Python install (slower): pip install apache-beam')

Prevention

Install build extras: pip install cython numpy before source builds
Prefer official apache-beam wheels with prebuilt extensions
Add a build-time assertion for cython in CI

When it happens

Trigger: Running 'python setup.py build_ext --inplace' or installing from source without the cython package installed in the build environment.

Common situations: Fresh virtualenv without dev dependencies; building Beam wheels in CI images that omit build extras; users installing from a git checkout rather than a wheel.

Understand the failure class

Background: "X is not installed. Please install it with pip install Y": missing optional dependency errors — ImportError/ValueError raised when a library's optional extra was never installed — this error's family across 22 libraries.

Related errors

AI-assisted analysis of apache/beam@12126d8942 (2026-09-13). Data as JSON: /api/errors/6c2dca949f5b2c64. Report an issue: GitHub.

Appendix: source

Thrown at sdks/python/setup.py:141

            _CYTHON_VERSION, REQUIRED_CYTHON_VERSION
        )
    )
except PackageNotFoundError:
  # do nothing if Cython is not installed
  pass

try:
  # pylint: disable=wrong-import-position
  from Cython.Build import cythonize as cythonize0

  def cythonize(*args, **kwargs):
    import numpy
    extensions = cythonize0(*args, **kwargs)
    for e in extensions:
      e.include_dirs.append(numpy.get_include())
    return extensions
except ImportError:
  warnings.warn(
      "Cython not found, cython extensions will not be generated. " \
      "To use cythonized extensions, pip install cython and run " \
      "python setup.py build_ext --inplace"
  )
  cythonize = lambda *args, **kwargs: []

# [BEAM-8181] pyarrow cannot be installed on 32-bit Windows platforms.
if sys.platform == 'win32' and sys.maxsize <= 2**32:
  pyarrow_dependency = ['']
else:
  pyarrow_dependency = [
    # Generally try to cover versions released in the last two years.
    # Update python/sdks/tox.ini to cover the same pyarrow versions
    # when updating the bounds here.
    'pyarrow>=14.0.1,<26.0.0',
  ]

# Exclude pandas<=1.4.2 since it doesn't work with numpy 1.24.x.

View on GitHub (pinned to 12126d8942)