On Mon, Aug 20, 2007 at 04:44:16PM +0200, Leuven, E. wrote:
> hi guys,
>
> i made a start with ripping out the way merged cells are handled in
> the tabular code.
>
> atm cells can have attributes
>
> CELL_NORMAL CELL_BEGIN_OF_MULTICOLUMN CELL_PART_OF_MULTICOLUMN
>
> if a range of cells is merged the first cell gets the
> CELL_BEGIN_OF_MULTICOLUMN attribute and the rest
> CELL_PART_OF_MULTICOLUMN.
>
> so we keep a range of cells for a single merged cell. there is a lot
> of bookkeeping necessary to keep things consistent.
>
> i basically got rid of this so that we have one cell per cell. in the
> cell we have the attribute column_span which tells us how many columns
> the cell spans.
>
> since this stuff is all over the place the attached patch is not
> exactly tiny or clean.
>
> it is not finished yet, but most things should be ok.
>
> since this tabular stuff is such a mess and tricky i'd like to ask for
> some input.
>
> basically: is this the way to go?
If you have the feeling that things are simpler/less-of-a-mess
afterwards, then it doesn't sound wrong.
> (i also touched the metrics and drawing stuff, so perhaps a
> knowledgeable person could have a peek?)
>
> thanks, edwin.
Content-Description: tab1.diff
> Index: src/frontends/controllers/ControlTabular.cpp
> ===================================================================
> --- src/frontends/controllers/ControlTabular.cpp (revision 19656)
> +++ src/frontends/controllers/ControlTabular.cpp (working copy)
> @@ -92,7 +92,7 @@
>
> void ControlTabular::toggleTopLine()
> {
> - if (tabular().isMultiColumn(getActiveCell()))
> + if (tabular().cell_info[getActiveCell()].multicolumn)
Why no keep the accessor method?
> Tabular::cellstruct::cellstruct(BufferParams const & bp)
> - : cellno(0),
> - width_of_cell(0),
> - multicolumn(Tabular::CELL_NORMAL),
> + : width_(0),
> + multicolumn(false),
> + column_span(1),
> alignment(LYX_ALIGN_CENTER),
> valignment(LYX_VALIGN_TOP),
> - top_line(true),
> + top_line(false),
> bottom_line(false),
> - left_line(true),
> + left_line(false),
> right_line(false),
> usebox(BOX_NONE),
> rotate(false),
This seems to change behaviour.
Would your approach also allow for cells spanning multiple rows in the
long run?
Wrt the metrics/drawing: To big to make an educated guess whether it
still works as before. So if it works, it works. Make sure you test with
a variety of multicol cells.
Andre'
