Re: [css-images] gradient midpoints

On Thu, Apr 17, 2014 at 11:07 AM, Tab Atkins Jr. <jackalmage@gmail.com>wrote:

> On Thu, Apr 17, 2014 at 10:58 AM, Rik Cabanier <cabanier@gmail.com> wrote:
> > At the CSS F2F in Seattle, we decided to use exponential interpolation
> > function to calculate the color ramps when there is a gradient midpoint.
> [1]
>
> Don't worry, I've got this tracked to edit in:
> https://github.com/tabatkins/specs/issues/29  ^_^
>
> > Right now the spec [2] states:
> >
> > Similarly, the color of a color stop can be omitted. The causes the
> color to
> > be automatically computed to halfway between the two surrounding stops,
> so
> > that the "midpoint" of a transition can be easily adjusted.
> >
> > which specifies linear interpolation that results in banding.
> > Could we add the right formula here?
> > To calculate a color C between:
> >
> > C0 = color stop on the left
> > C1 = color stop on the right
> > N = value between 0 and 1 describing the distance of the midpoint
> between C0
> > and C1
> > X = value between 0 and 1 describing the distance of the C between C0
> and C1
> >
> > The formula is: C = C0 + pow(x, log(.5)/log(X)) * (C1 - C0)
>
> This formula isn't correct, but I know what you mean. ^_^  (Your
> CodePen had useful code in it.)
>

oops, good catch!

C = C0 + pow(X, log(.5)/log(N)) * (C1 - C0)


> > The spec also has this wording:
> >
> > If multiple stops in a row lack a color, they space themselves out
> equally
> > in "color space", giving more control over the smoothness of the
> transition.
> >
> > Is this to add multiple midpoints?
>
> It has nothing to do with the conception of "midpoints" that you
> proposed at the f2f.  The current state of the spec (since I haven't
> yet edited in the f2f resolution) has the behavior of a color stop
> being symmetrical between missing a position and missing a color -
> either way, you just fill in the missing values by spacing them out
> between the surrounding specified values.


ah, I see
so, if a gradient goes from 0 to 1 and there are 2 stops at 10% and 90%,
stop 1 would get 33% and stop 2 66%.
Would there be linear interpolation at that point?