2.15

2.15

• BREAK: ∆E HCT used an extremely small JND by default to yield tonal palettes that were comparable to Google's
  Material. This was not inline with other gamut mapping function defaults. The default is now a more appropriate
  value of 2. Users that relied on ∆E HCT to help generate tonal pallets with HCT should now use the jnd parameter
  to set the JND to 0.02 in order to generate tonal pallets more like Google.
• NEW: fit() now accepts a jnd option to control the JND limit when gamut mapping with lch-chroma,
  oklch-chroma, and hct-chroma.
• NEW: to_string() can now accept a dictionary of arguments to control gamut mapping via the fit argument.
• FIX: Update lch-chroma epsilon to be consistent with other gamut mapping plugin conventions.

2.14.1

2.14.1

• FIX: More precision for HCT conversion for better round trip conversions.

2.14

2.14

• NEW: normalize() will now also normalize a cylindrical color model with negative chroma/saturation to its
  positive chroma/saturation form, assuming one exists.
• NEW: Gamut clipping is performed on a cylindrical color's normalized form ensuring that a color which is in
  gamut but has a negative chroma/saturation will be mapped more correctly.
• NEW: Do not clamp user input of lightness and chroma in various spaces. Clamping will only occur during
  conversion if the algorithm requires it.
• NEW: Channels can be accessed by get and set using their numerical value (as a string input).
• NEW: Color space plugins that specify the gamut space via GAMUT_CHECK must use that color space as a
  reference when when gamut mapping or clipping by default.
• NEW: New color space plugin attribute CLIP_SPACE added which will override the space specified by
  GAMUT_CHECK to force clipping in the origin space even if a gamut mapping space is defined. This is only used when
  it is advantageous to clip in the origin space, e.g. when faster and still practical.
• NEW: Deprecate non-standard CAM16 (Jab) space. People should use the standard CAM16 JMh or the CAM16 UCS, SCD,
  or LCD Jab spaces. The non-standard Jab is still available via coloraide.spaces.cam16.CAM16, but it is no longer
  available in coloraide.everything and will be removed at a future time.
• NEW: Add new channel aliases: j for jz in Jzazbz and JzCzhz. Also add c for cz and h for hz in
  JzCzhz.
• NEW: HSL will now always return positive saturation for wide gamut colors via its conversion.
• FIX: Fix a an issue with the CAM16 model's transformation that prevented good round trip with negative
  lightness.
• FIX: Ensure that when harmony auto creates a cylindrical space from a rectangular space that it checks
  achromatic status in the original color space.
• FIX: ∆E HCT should use colorfulness, not chroma, in the calculation.
• FIX: Don't return scientific notation when serializing colors.
• FIX: Small fix for Rec. 2100 PQ conversion algorithm.
• FIX: The oRGB color space should be gamut mapped in srgb as it is a transform of the sRGB space.
• FIX: Because Okhsl and Okhsv have a rough sRGB approximation and not precise, they are instead gamut mapped to
  their own gamut by default.
• FIX: Much more accurate ICtCp matrices.
• FIX: Fix typing of deeply nested arrays in algebra.
• FIX: Fix issue with HCT undefined channel resolver.
• FIX: Proper handling of negative lightness for DIN99o.
• FIX: Circular mean should return positive values.

2.13.1

2.13.1

• FIX: Minor typing regressions and fixes.

2.13

2.13

• NEW: Performance related enhancements in high traffic calculations.
• NEW: Use matmul instead of dot in calculations to not confuse math savvy people.
• FIX: Some typing fixes/improvements.
• FIX: Minor fixes to algebra library.

2.12

2.12

• NEW: When serializing, percent can now take a sequence of booleans to indicate which channels are desired to
  be represented as a percentage, alpha included.
• NEW: color() serializing now supports string output with percent.
• FIX: When serializing, the alpha channel is no longer handled special with a minimum value of 5. Precision is
  equally applied to all channels.

2.11

2.11

• NEW: Add new css-linear interpolator that provides compatibility with the CSS specification. This deviates
  from the default linear interpolator in how undefined hues are resolved for interpolation, particularly noticeable
  with longer hue resolution.
• NEW: Add new INTERPOLATOR class option to change the default interpolator that is used.

2.10

2.10

• NEW: Declare official support for Python 3.12.
• NEW: Color.steps and Color.discrete now accept delta_e_args to allow configuring the underlying distance
  algorithm when using the delta_e option.
• NEW: CIE Lab, both D50 and D65, are now derived from a CIELab class. CIE LCh, both D50 and D65, are also
  now derived from a CIELCh class. This makes it easy to determine a CIE Lab or CIE LCh space from other Lab-like
  spaces.
• NEW: ∆E^*^76, ∆E^*^94, ∆E^*^00, and ∆E^*^cmc all accept a new parameter called space which allows
  the user to specify a registered Lab color space name (one that is derived from the CIELab class) to use as the
  distancing color space. This allows a user to use D50 Lab (or any other variant) for distancing if required.
• FIX: For consistency, ∆E^*^94 and ∆E^*^cmc now use Lab D65 by default just like ∆E^*^76 and
  ∆E^*^00. This fixes an issue where the docs indicated that they use D65, but in actuality they were using D50.

2.9.1.post1

2.9.1.post1

• FIX: Fix incorrect changelog mention of recent fix being for HSL instead of HWB.

2.9.1

2.9.1

• FIX: Average should allow controlling powerless be disabled by default for backwards compatibility.
• FIX: HWB should use the algorithm defined in CSS that allows for round tripping even in the negative lightness
  direction. Previously we were converting directly from HSV.