On Tue, Dec 07 2010, Lars Magne Ingebrigtsen wrote:

> Well, if there's more code being run in the common paths, it's going to
> be slower.  :-)  The balance we're looking for is between "render as
> prettily and readably as possible" vs "render fast enough for actual
> use".  And that's a real trade-off. 

I agree, but I'm just suggesting to optimize things based on real data
rather than throwing "it will be slower that way".

It's obvious that it will be slower if we add more code (thanks). But if
the table code execution time (you say it's slow) and what I'm proposing
to add are having a 90/10 execution time ratio, arguing that adding more
stuff is bloating the execution time is not really helpful. :-D

Anyhow, rather than speaking forever I've implemented what I've in mind.

Here is a patch.