[vox-tech] [OT] Math question: help me make a rainbow

harke rharke at earthlink.net
Thu Apr 2 01:22:30 PDT 2009


On Tuesday 31 March 2009, Bill Kendrick wrote:
> Tux Paint has a magic tool called "Rainbow."  It's not really, it's just
> a color-cycling circle-shaped paintbrush.  A few years ago, a couple of
> the other TP developers joked that we need a tool to draw real rainbows.
> (Draw an arc, and the rainbow gradient appears over it.)
>
> So I've decided it might be fun to take a stab at it.  My main problem is
> I'm rusty with math (trig., geometry, etc. ... you don't use those much in
> PHP or MySQL. ;) Heck, you don't use them much in most of the kinds of
> games I've worked on over the years.)
>
> Here's how I imagine the UI working.  Click mouse to select one endpoint
> of the rainbow.  Drag to where you want the other endpoint, and release
> mouse.  Rainbow gets drawn.
>
> The rainbow would be an upside-down-"U" shape (with a 1:1 aspect, e.g.
> it'd be a portion of a perfect circle).  I think for artistic purposes,
> it shouldn't always be a complete 180-degree arc, though.  So here's
> how I complicate the UI :)
>
> P1 (aka x1,y1) is where you click.  P2 (aka x2,y2) is where you release.
>
> 0,0 is the top left of the screen.  Angles go from 0 (bottom/right of arc)
> thru 90 (top/center of arc) to 180 (bottom/left of arc).
>
> If y1 >= y2, then P1 represents the 180th degree of the arc
> (going counter-clockwise from the far right of the circle), and
> P2 represents the Nth degree of the arc (where 0 <= N < 180)
>
> If y1 < y2, then P1 represents the Nth degree of the arc,
> and P2 represents the 0th degree.
>
> This will let you draw, e.g. (bad ASCII ahead):
>
>      /==P2
>     /
>
>   P1
>
> as well as the full semi-circular arc:
>
>     /==\
>    /    \
>
>   P1    P2
>
> and of course, if P2 is lower, then:
>
>
>     /==\
>    P1   \
>
>         P2
>
> In both cases, there's an invisible P3, which is at 0 degrees or 180 deg.,
> depending on which point ended up being lower.  The center of the circle
> is not necessarily exactly at (x1+((x2-x1)/2) , max(y1,y2)) which is where
> I'm getting stuck.
>
> My initial though was to find the vector V between P1 and P2.
> Going off at the normal of V, from its center, I believe I get a line
> that strikes the center of my would-be circle, when it crosses the
> max(y1,y2) height down the screen.
>
> But I was up late thinking about this, and worrying about atan() and
> intersect tests, and starting to think there's a simpler way.
>
>
> So, given two points on an arc of a circle, P1 and P2, how do I find
> the center and radius of the circle?  (Then all I need to do is determine
> the angle (N) of one of the points (P1 or P2), while I know that the
> angle of the other (P2 or P1) is 0 or 180, depending on my UI rules above.
>
> Whew.  Did I make ANY sense? :)

If I understand you correctly, you want the line through the
lowest of the two points to be a diameter of the circle. In that case,
of course, the center of the circle is on that line. You immediately
have the y-coordinate. You can compute the x-coordinate by
writing out an expression for the distance between the first point
and the center and the distance between the second point and the center.
Note for the lower point, this will just be the difference of
the two x-coordinates. Set these two distances equal as they are
the radius of the desired circle. This equation can be solved for the
x-coordinate of the center. (Left as an exercise)

Richard harke


More information about the vox-tech mailing list