Re: [patch] Sleeping cells, correction

[Jody Goldberg <jody gnome org>]

On Sun, Jul 14, 2002 at 10:51:43PM +0200, Eric Methorst wrote:
> b.t.w.... am I supposed to reply to the entire mailing list here?
Thats fine.  It is useful to have design choices made on the list so
that others can take part.

> > > +cell_is_sleeping(Cell const *cell)
> > > +{
> > > + return mstyle_get_content_sleeping(cell_get_mstyle(cell));
> > >  }
> >
> > ouch this gets expensive.  The quad tree makes style lookups fast,
> > but still not something you want to do 1M times.
> Can that be done via something like: 
> cell->mstyle->is_sleeping
> directly? Or is there some other means of doing it quickly? 
Cells do not have style pointers.  The mstyle_get_content_sleeping
is fast.  cell_get_mstyle (cell) is the bottleneck.  To be useful it
would probably need to be a dependent flag, however that is
significantly more work to move around as styles are changed.  We
could steal a page from Quattro Pro and have a magic function
wrapper.
    =expensive(expression.....)
That would be make support at the dependent level much simpler.

> > The main question in my mind is why/how this would be useful.
> > The notion of a spreadsheet with _partially_ incorrect values is a
> > little scary.
> It is not a _partially_ incorrect value; but a 'explicitly not calculated' 
> value. Usage for example: to calculate primary keys and other values and use 
> those as part of an insert statement (or update) in a database. You certainly 
> donot want any insert untill you're completely done entering all fields. Not 
> only would this result in database constraint errors, but it might also be 
> killing for performance.

The issue here is whether this is the best way to handle these use
cases.  As long as the extension mechanism is a new sheet function
we don't have much of a choice.  However, a longer term goal would
be to support commands also.  This would allow a menu/toolbar item
to control data refresh for external links and recalc for 'sleeping'
cells would be mostly be a nop (dunno how that would affect things
like cut-n-paste where the value has really changed).