Proposal for XSL Tables: Split Cells

Requirement

These kind of tables are very common in China, Korea, and Japan. Academia Sinica considers that it is an internationalization requirement to support them, and invites the XSL FO WG to consider the following proposal.  Alphabetic scripts have longer words than ideographic scripts; consequently splitting header cells diagonally is not in important or common typographical practise for use with alphabetic scripts. However, it is a common practise in East Asia: one regularly sees publications in which every table is set with diagonally-split header-cells.

Occurrence

The definition of table-row should be changed to
    ( table-cell | table-split-cell-container )+

fo:table-split-cell-container

• Common Use

The fo:table-split-cell-container formatting object is used to group table-split-cells to be placed in a table-row.
 
• Type of Area

This formatting object does not direct create any area.
 
• Content distribution and size

Same as fo:table-cell.

A table-split-cell-container cannot be spanned-into by another cell, either vertically or horizontally. This requirement is provided to simplify implementation and design consideration; it is reasonable because split-cells are used for headers rather than data cells.
 

• Content

    ( table-split-cell )+
 
• Properties

Same as fo:table-cell.

An advanced system would have a property to specify the point from which the diagonals radiate.  In the description of table-split-cell below, the top left-hand corner is assumed, as that is the usual case in most modern CJK tables.  However,  tables read right-to-left with row headers on the right-hand-side are still found; these require a top-right radation point.

fo:table-split-cell

• Common Use

The fo:table-split-cell formatting object is used to implement diagonally-split headers.  Visually, the borders of each split cell are lines drawn from the top-left corner (relative to the table alignment) or the parent table-split-cell-container down-and-across to the intersection of the parent split-cell-container and a grid-unit boundary.
 
• Type of Area

A table-split-cell is occupies an area with non-parallel sides: a cell with at least one edge not vertical or horizontal.   Each table-split-cell results in a block area.

Within the polygon, various options are available for determining the size of the block area. The default method is that the largest available rectangle that can be fitted into the table-split-cell is used to determine the block size: this block size would be (closer to) the content-rectangle size rather than then padding rectangle as the purpose of these formatting objects is to utilize space more effectively for scripts with words that occupy a short measure.

(A more advanced formatting system could attempt to fit several block areas into the available space. For ideographic text, these would give the effect of text being written on a sloped line-progression angle, though really it would  be that each block has only one ideograph and is placed going down a slope.)
 

• Content distribution and size

Each table-split-cell is place in clockwise fashion.  The abuttment property references a row or colspec.  The boundary of the table-split-cell is created by the polygon create by starting from the top left-hand point and following the edges of the currently unallocated region of the parent split-table-cell-container until the lower boundary of the row (if the abuttment property specifies a row) or the rightmost boundary of a column (if the abuttment property specifies a column).

At this end point, the polygon is completed by either creating a side which goes to the start-point, or by casting a line at 45 degrees going left and upward until it bisects the enclosing table-split-cell-container.

The last table-split-cell occupies the remaining area.

(More advanced formatting systems may adjust the boundary points of each table-split-cell to give the more equal area to each.  Thus there can be some flexibility in both the start and end-point.  Sophisticated implementations may introduce mechanisms to avoid particularly steep or low-rising diagonals by adusting column or row size or by altering the angle of the diagonal somewhat.)
 

• Content

    ( %block; )+
 
• Properties
• abuttment  ( If it is not present then the next grid boundary is used, clockwise. The last table-split-cell cannot have this property: the last table-split-cell occupies all the remaining space.)

This is an IDREF to either a colspec or a row.

<table-and-caption><table-caption>What are they wearing?</table-caption>

<table>

    <table-header>

        <table-split-cell-container>

            <table-split-cell>Age</table-split-cell>

            <table-split-cell>Sex</table-split-cell>

        </table-split-cell-container>

        <table-cell>0 - 16</table-cell>

        <table-cell>17-100</table-cell>

   </table-header>

    <table-body>

    <table-row>

        <table-cell class="header>Male</table-cell>

        <table-cell>shorts</table-cell>

        <table-cell>tattoos</table-cell>

    </table-row>

    <table-row>

        <table-cell class="header>Female</table-cell>

        <table-cell>skirt</table-cell>

        <table-cell>perfume</table-cell>

    </table-row>

   </table-body>

</table>

</table-and-caption>