ROWSPAN > 1 to row distribution

The algorithm has not been standardized. This is my understanding of how it works.

1. rowspan>1 TDs are sorted:
1. If TD span the same rows, taller TD is distributed first.
2. If one TD is fully enclosed by another, inner TD is distributed first.
3. Otherwise, higher TD is distributed first.
2. Each rowspan>1 TD's height is distributed as following
1. rowspan > 1 TDs have height TDh, span N rows. N rows have total height of Rh. TDh - Rh height, Dh, must be distributed as follows.
2. If percentage resolution size is available (this happens when redistributiong table/section height), percentage rows grow to their percentage size, proportional to (percentage size - current size). Dh shrinks by distributed height. Justification: explicit percentage rows should grow to their percentage.
3. Rows that originate rowspan>1 cells get all the Dh height, distributed evenly. Justification: rowspan>1 rows are likely to need to grow later. If there are multiple rowspan>1 cells, there can be multiple originating rows.
4. Unconstrained non-empty rows grow, proportional to their existing height.
5. If all rows are empty, last row gets all the height. Justification: ???
6. Contstrained rows grow in proportion to current height.

It is unclear what the existing ChromeLegacy/FF algorithms do for distribution over rowspan>1 and empty cells. FF special cases "there is no cell originating in the row with owspan=1 and there are at least 2 cells spanning the row. Chrome Legacy also tries to do something similar, but they disagree on what. TablesNG will try to ship without special cases.

Safari fails most of these tests

Color scheme

Unconstrained rows

Rows whose height is not fixed or percent are unconstrained.

Unconstrained rows Unconstrained rows are redistributed proportionally. Rows are constrained if their height is fixed, or percent.

Unconstrained rows with zero height do not grow.

rowspan>1 is zero height, spanned rows have height.

Unconstrained rows are redistributed proportionally to heights

Fixed rows

Constrained fixed rows do not grow if there are unconstrained ones

Edge grows constrained rows too

Constrained fixed rows grow proportionally to their size if there are no unconstrained rows

Percent rows

Constrained percent rows grow like unconstrained ones when percent resolution size is undefined.

FF always treats percent rows as constrained. Chrome legacy does resolve percentage against final height of the table. I do not think that can work. Edge follows NG.

Percent rows with zero height do not grow.

Legacy Chrome has a strange gap between rows

Order of rowspan distribution

If cells span the same rows, bigger cell is distributed first Not sure how to test this, I think it is just an optimization, there is no observable effect.

FF and Legacy Chrome unexpectedly distribute height evenly between rows in the first test case. Edge and TablesNG do not.

If one cell is fully enclosed by another, inner cell wins.

Not in Edge

First row wins. rowspan-4 distributes 50 to last empty row, row3. rowspan-3 distributes 100px to only nonempty row, row3.

Edge disagrees here.

Rowspan distribution over empty rows.

Rowspans that span non-existent rows Span is truncated so only existing rows are spanned.

Rowspan spans only empty rows Last spanned row gets all the height.

Edge distributes height to all empty rows, not just last.

TD is not considered empty if it has padding, but no content

row with an empty tall cell is not considered empty.

Empty rows with border-spacing big enough for rowspan cell rows are 0 height, cell spans the entire table.

row with a non-empty rowspan>0 cell is empty. Distributes to all rows except start row?

Distribution over rowspan > 1 rows Distribution over rowspan > 1 rows

Distribution of table height over rowspan > 1 rows If there are any unconstrained non-empty rows, they get it. When all rows are empty, last row takes it

Distribution of rowspan over percentage rows Percentage rows are considered empty if they cannot resolve