Re: [css21] Anonymous table objects - box tree?

On 11/24/10 11:42 AM, Øyvind Stenhaug wrote:
> The "Anonymous table objects" section[1] seems to talk about the box
> tree (at some intermediate stage in construction), even though that's
> not defined (and the spec just says there's a "formatting structure"
> which "need not be 'tree-shaped'" and "depends on the implementation").

Yes.  To put it politely, what the spec says is ... not internally 
consistent.  It assumes a tree structure for at least in-flow boxes in a 
number of places.  And it tries hard to assume it in other places, and 
ends up with behavior not being defined as a result.

> "For each 'table-cell' box C in a sequence of consecutive internal table
> and 'table-caption' siblings, if C's parent is not a 'table-row' then
> generate an anonymous 'table-row' box around C and all consecutive
> siblings of C that are 'table-cell' boxes."
>
> So...
> - is "C's parent" actually referring to a link in the formatting
> structure?

Yes.

 > If so, where is the table-caption's box located at this point
> in time

It's a child of the table box.  The assumption is that this algorithm is 
run once all the rest of the formatting tree construction work has 
happened.  That should perhaps be made clearer.

> is it a "consecutive sibling" or not?

Yes.

> - is "C's parent" actually about "the parent of the element that
> generated C"?

No.

> If so, what about anonymous table-cell boxes?

Exactly.

-Boris

Received on Wednesday, 24 November 2010 17:23:00 UTC