Advanced Usage#

This page describes some advanced usage scenarios of Colour.


Various environment variables can be used to modify Colour behaviour at runtime:

  • COLOUR_SCIENCE__DEFAULT_INT_DTYPE: Set the default integer dtype for most of Colour computations. Possible values are int32 and int64 (default). Changing the integer dtype will almost certainly break Colour! With great power comes great responsibility.

  • COLOUR_SCIENCE__DEFAULT_FLOAT_DTYPE: Set the float dtype for most of Colour computations. Possible values are float16, float32 and float64 (default). Changing the float dtype might result in various Colour functionality breaking entirely. With great power comes great responsibility.

  • COLOUR_SCIENCE__DISABLE_CACHING: Disable the caches that can be disabled, useful for debugging purposes.

  • COLOUR_SCIENCE__COLOUR__IMPORT_VAAB_COLOUR: Import vaab/colour injection into Colour namespace. This solves the clash with vaab/colour by loading a known subset of the objects given by vaab/colour-0.1.5 into our namespace.

  • COLOUR_SCIENCE__COLOUR__SHOW_WARNINGS_WITH_TRACEBACK: Result in the warnings.showwarning() definition to be replaced with the colour.utilities.show_warning() definition and thus providing complete traceback from the point where the warning occurred.

  • COLOUR_SCIENCE__FILTER_RUNTIME_WARNINGS: Filter Colour runtime warnings.

  • COLOUR_SCIENCE__FILTER_USAGE_WARNINGS: Filter Colour usage warnings.

  • COLOUR_SCIENCE__FILTER_COLOUR_WARNINGS: Filter Colour warnings, this also filters Colour usage and runtime warnings.


JEnv File#

Colour will also read the ~/.colour-science/colour-science.jenv JSON file if it exists. The syntax is that of a mapping of environment variable and values as follows:



Colour uses various internal caches to improve speed and prevent redundant processes, notably for spectral related computations.

The internal caches are managed with the colour.utilities.CACHE_REGISTRY cache registry object:

import colour

{'colour.colorimetry.spectrum._CACHE_RESHAPED_SDS_AND_MSDS': '0 item(s)',
 'colour.colorimetry.tristimulus_values._CACHE_LAGRANGE_INTERPOLATING_COEFFICIENTS': '0 '
 'colour.colorimetry.tristimulus_values._CACHE_SD_TO_XYZ': '0 item(s)',
 'colour.colorimetry.tristimulus_values._CACHE_TRISTIMULUS_WEIGHTING_FACTORS': '0 '
 'colour.quality.cfi2017._CACHE_TCS_CIE2017': '0 item(s)',
 'colour.volume.macadam_limits._CACHE_OPTIMAL_COLOUR_STIMULI_XYZ': '0 item(s)',
 'colour.volume.macadam_limits._CACHE_OPTIMAL_COLOUR_STIMULI_XYZ_TRIANGULATIONS': '0 '
 'colour.volume.spectrum._CACHE_OUTER_SURFACE_XYZ': '0 item(s)',
 'colour.volume.spectrum._CACHE_OUTER_SURFACE_XYZ_POINTS': '0 item(s)'}

See colour.utilities.CacheRegistry class documentation for more information on how to manage the cache registry.

Using Colour without Scipy#

With the release of Colour 0.3.8, SciPy became a requirement.

Scipy is notoriously hard to compile, especially on Windows. Some Digital Content Creation (DCC) applications are shipping Python interpreters compiled with versions of Visual Studio such as 2011 or 2015. Those are incompatible with the Python Wheels commonly built with Visual Studio 2008 (Python 2.7) or Visual Studio 2017 (Python 3.6).

It is however possible to use Colour in a partially broken state and mock Scipy by using the module.

Assuming it is available for import, a typical usage would be as follows:

import sys
from mock_for_colour import MockModule

for module in (
    sys.modules[str(module)] = MockModule(str(module))

import colour

xyY = (0.4316, 0.3777, 0.1008)
array([ 0.11518475,  0.1008    ,  0.05089373])

Or directly using the mock_scipy_for_colour definition:

from mock_for_colour import mock_scipy_for_colour


import colour

xyY = (0.4316, 0.3777, 0.1008)
array([ 0.11518475,  0.1008    ,  0.05089373])

Anything relying on the spectral code will be unusable, but a great amount of useful functionality will still be available.