Rotated size algorithm got lost from image API spec

[zimeon]

Current draft has in implementation notes:

"As described in Rotation, in order to retain the size of the requested image contents, rotation will change the width and height dimensions of the returned image file. A formula for calculating the dimensions of the returned image file for a given rotation can be found here."

with no link for "here". Need to fish that out and decide whether to add link or put inline.

[zimeon]

Also missing in the live 1.1: http://iiif.io/api/image/1.1/#implementation

[jpstroop]

That's just a matter of linking to here:

https://github.com/IIIF/iiif.io/blob/master/source/api/annex/notes/rotation.md

no? The big image seems like it would be worth keeping it in the annex.

[zimeon]

In 1.1 on old server http://www-sul.stanford.edu/iiif/image-api/1.1/#implementation the link goes to http://www-sul.stanford.edu/iiif/image-api/1.1/rotation_calc.html

Seems that with this picture it would have to stay in a separate document.... but I don't like this calculation anyway. What does it say about 270deg, is that the >90deg case...NO. I see no reason why one needs to split the calculation on 90deg boundaries and if one does it needs to be every 90deg boundary. One can instead do:

     w_new = abs(w*cos(angle)) + abs(h*sin(angle))
     h_new = abs(h*cos(angle)) + abs(w*sin(angle))

or to put that in a runnable form:

#!/usr/bin/env python
from math import radians,cos,sin
w = 20
h = 10
for degrees in [-60.0,-45.0,-30.0,0.0,30.0,45.0,60.0,90.0,110.0,180.0,270.0,300.0]:
    angle = radians(degrees)
    w_new = abs(w*cos(angle)) + abs(h*sin(angle))
    h_new = abs(h*cos(angle)) + abs(w*sin(angle))
    print "@ %4d -> (%.1f,%.1f)" % (degrees,w_new,h_new)

[jpstroop]

I'm OK with moving this inline if @zimeon wants to do the rewrite and drop the annex. Others?

[zimeon]

My pull request #253 does not include deletion of the the old rotation calc note. I see two options:

1. link it back into 1.1 and think about fixing sometime
2. remove leaving the broken ref in 1.1

If we want to remove I think it is just a matter of deleting

source/api/annex/notes/rotation.md
source/api/annex/notes/iiif-rotated-img-size.png

I don't think they are linked from anywhere:

simeon@Cider iiif.io>grep -r --exclude="./.git/" --exclude="./site/" iiif-rotated-img-size .
./source/api/annex/notes/rotation.md:[rotation-ill]: ./iiif-rotated-img-size.png "Formula for calculating image size of rotated image"
simeon@Cider iiif.io>grep -r --exclude="./.git/" --exclude="./site/" notes/rotation .

[jpstroop]

As I've said (somewhere, I think...starting to lose track). I'm completely OK with only providing a formula for rotation and getting rid of the annex.

Does anyone else have an opinion?

[azaroth42]

I think we should leave 1.1 as is, which means keeping the note somewhere.
And I'm happy to include the algorithm in Appendix A... but perhaps it's worth expanding to A.1 A.2 etc with the number of notes we're getting, so we can reference them separately?

That algorithm is much more efficient than mine:

                maxx = image.size[0]
                maxy = image.size[1]
                cx = int(image.size[0]/2)  # x0 above
                cy = int(image.size[1]/2)  # y0 above
                theta = math.radians(rotation)

                toplx = cx + (0-cx)*math.cos(theta) + (0-cy)*math.sin(theta)
                toply = cy - (0-cx)*math.sin(theta) + (0-cy)*math.cos(theta)                
                toprx = cx + (0-cx)*math.cos(theta) + (maxy-cy)*math.sin(theta)
                topry = cy - (0-cx)*math.sin(theta) + (maxy-cy)*math.cos(theta)
                botlx = cx + (maxx-cx)*math.cos(theta) + (0-cy)*math.sin(theta)
                botly = cy - (maxx-cx)*math.sin(theta) + (0-cy)*math.cos(theta)    
                botrx = cx + (maxx-cx)*math.cos(theta) + (maxy-cy)*math.sin(theta)
                botry = cy - (maxx-cx)*math.sin(theta) + (maxy-cy)*math.cos(theta)                

                nminx = min(toplx, toprx, botlx, botrx)
                nmaxx = max(toplx, toprx, botlx, botrx)
                nminy = min(toply, topry, botly, botry)
                nmaxy = max(toply, topry, botly, botry)

                rx = nmaxx - nminx
                ry = nmaxy - nminy

[zimeon]

@azaroth42 - I'm not sure about going to A.1 A.2 as that will make the appendix very long with all the headings. How about just changing to a set of numbered bullets as a compromise?

[azaroth42]

Sure :)

On Tue, Jul 1, 2014 at 8:13 AM, Simeon Warner notifications@github.com
wrote:

@azaroth42 https://github.com/azaroth42 - I'm not sure about going to
A.1 A.2 as that will make the appendix very long with all the headings. How
about just changing to a set of numbered bullets as a compromise?

—
Reply to this email directly or view it on GitHub
#251 (comment).

Rob Sanderson
Technology Collaboration Facilitator
Digital Library Systems and Services
Stanford, CA 94305