Re: Doc string and operation of color-distance

[Mark Oteiza]

On 14/09/17 at 09:45pm, Eli Zaretskii wrote:
> Mark, the new color-distance says something strange in its doc string:
> 
>   (color-distance COLOR1 COLOR2 &optional FRAME METRIC)
> 
>   Return an integer distance between COLOR1 and COLOR2 on FRAME.
>   COLOR1 and COLOR2 may be either strings containing the color name,
>   or lists of the form (RED GREEN BLUE), each in the range 0 to 65535 
> inclusive.
>   If FRAME is unspecified or nil, the current frame is used.
>   If METRIC is unspecified or nil, a modified L*u*v* metric is used.
> 
> The last sentence should say "non-nil", I think

No, the 'nil' is correct--if no metric is given as an argument,
color_distance is used which, as the comments say, is a modified L*u*v*
metric.

> and it should
> document that METRIC is supposed to be a function of 2 colors.

True, that would be helpful.

> Also, this is unexpected:
> 
>   (color-distance "red" "blue" nil 'lcms-cam02-ucs)
>   => (error "Invalid color" "red")
> 
> I think the new lcms2 functions should support colors specified as
> strings, because all the other color-related functions do.

I thought (but did not test) the body of color-distance did the lookups.
Now I see what's going on.