[racket] 2htdp/image: colors and alphas

From: Robby Findler (robby at eecs.northwestern.edu)
Date: Mon Dec 27 08:47:25 EST 2010

On Mon, Dec 27, 2010 at 7:13 AM, Stephen Bloch <sbloch at adelphi.edu> wrote:
> I'm trying to port the picturing-programs teachpack to still work with 5.1.  (BTW, approximately how long do I have to do that before 5.1 is released?)

We have not started the release process yet, but the release is
sometime in Jan. Generally releases are every 3 months. Ryan should
know the precise dates.

> Right now it crashes because the API for make-bitmap has changed -- for the better, I think, but I still need to fix things.

Which make-bitmap would that be? Are you talking about the internal,
private constructor?

> I've got a "map-image" function that creates a [bitmap] image the same width and height as an existing image by mapping a color->color function on its pixels, and a "build-image" function that creates a [bitmap] image of a specified width and height by mapping a specified function on the coordinates of its pixels.  I also wrote (but didn't really document) a "map-masked-image" and "build-masked-image" which do the same, but the mapped function is also allowed to return #f, in which case the specified pixel is made transparent (i.e. alpha=0).  I can tweak those to work with the new make-bitmap, but the process got me wondering...
> It would be nice if I could allow students to generate alphas other than 1 and 0.  It seems to me the cleanest way to do this is with an alpha-color struct (which existed in htdp/image, but seems to be gone from 2htdp/image).  I can do that myself, putting in glue code between it and 2htdp/image to convert back and forth among various color-like-things (color, alpha-color, string, symbol, color%, and #f for transparency), but 2htdp/image is already doing a lot of that, and it feels really inefficient to re-do it.
> What would you think of ALL "color" structs actually having an alpha component, with the constructor allowing alpha to default to 1?

I think you should be operating at the racket/draw bitmap level, not
the 2htdp/image bitmap level. That is, use the render-image function
to get a bitmap, and then make a loop that calls their function and
fills in a bytes and then turn that bytes back into a bitmap. This
will save creating a bunch of intermediate data and is a simpler
function overall.


Posted on the users mailing list.