Re: CORRECTION: change to box-packing algorithm?

(a new personal record: four replies to myself without any intervening posts)

David, if you want the "new" behavior in the old system, just make
sure rowspan=1 for all the boxes; then the lines of text won't
overlap.

  - a

Adam Megacz <[email protected]> writes:
> Adam Megacz <[email protected]> writes:
>>    "a box will be placed as far up and to the left as it can fit and
>>    still be no higher above and no farther to the left than its
>>    preceeding sibling"
>
> Scratch that.
>
> A box will be placed:
>
>   - no higher than its preceeding sibling
>   - either
>       - no farther to the left of its preceeding sibling
>       - below the bottoms of all of its preceeding siblings
>
>   - as far up and to the left as possible subject to the previous two
>     constraints.
>
> Basically this means that the packing algorithm requires O(1) space.
> There's a cool algorithm [*] that can do the original form of the
> box-packing algorithm in O(numboxes log numboxes) space, but it's a
> huge huge pain in the butt.
>
> This is also the (correct) statement of the intuitive text-reflow
> system (when colspan==width).
>
>   - a
>
> [*] it was actually invented by the people who did the original GUI
>     Toolbox for the mac... it represents 1-bit bitmaps as an
>     RLE-encoded array of RLE-encoded arrays.  Turns out to be hugely
>     efficient for bitmaps that are made of big, contiguous rectangles
>     (which is exactly what the "which-cells-are-taken" map is)
>
> _______________________________________________
> core mailing list
> [email protected]
> http://lists.xwt.org/listinfo/core
>

-- 
"It's lucky," he added, after a pause, "that there are such a lot of
islands in the world.  I almost envy you, Mr. Watson."

                                                   -- Mustapha Mond