[css-color-4] Discrepancy between § 19.1. ΔE2000 implementation and ISO 11664-6:2014/Sharma et al., 2005 #13322

The sample implementation in section of 19.1. ΔE2000 of CIEDE2000 has a discrepancy in the computation of the mean hue angle $\overline{h'}$ versus the equation in ISO/CIE 11664-6:2014 (I do not have the 2022 version available) which reference the paper by Sharma et al., 2005.

The offending line exists on line 103 of the code snippet:

if (Cdash1 == 0 && Cdash2 == 0) {

where $\overline{h'}$ (hdash in code) is set to hsum if Cdash1 and Cdash2 is zero, which differs from the specification condition that $C'_1 \cdot{} C'_2 = 0$, which would be implemented in code as:

if (Cdash1 * Cdash2 === 0) {

The above condition is correctly implemented for the computation of $Δh'$ (Δh in code) as seen on line 75 of the implementation.

Mathematically, the discrepancy does not matter in the final computation of the ΔE2000 value since having either $C'_1$ or $C'_2$ be 0 means that the $ΔH$ value is 0, and the mean hue angle $\overline{h'}$ is always eventually multiplied with $ΔH$, making the discrepancy disappear in the final result.

However, the intermediate values $Δθ$, $T$, and $R_T$ do show the discrepancy and may not match testing data that expect all the intermediate values to match as well.

A specific test pair of:
Reference L = 50.00, a = 0.000, b = 0.000
and
Sample L = 50.00, a = 10.00, b = -114.3

can be used to show large deviations between the two specifications for the $Δθ$, $T$, and $R_T$ values (as well as $\overline{h'}$), with $Δθ = 30.000$ for ISO/CIE 11664-6:2014 and $Δθ = 1.6669$ for the sample implementation in section 19.1. ΔE2000.

I was wondering what the rationale behind the change in implementation is, as the code comments don't seem to note anything about this discrepancy.