Re: [Gzz] 30th-31th (Benja)

[Tuomas Lukka]

On Thu, Aug 01, 2002 at 11:46:08AM +0200, B. Fallenstein wrote:
> On Thursday 01 August 2002 09:08, Tuomas Lukka wrote:
> > On Thu, Aug 01, 2002 at 12:59:20PM +0200, B. Fallenstein wrote:
> > > I've committed a change to ScrollSpanMaker that creates new
> > > TransientTextScrolls when the old ones have gone immutable.
> >
> > Heh, this is exactly the solution I thought of when mulling over
> > this yesterday ;) Glad to see we're on the same track.
> 
> :-)
> 
> > No, FakeSpan works like a span but is saved as a string and has no
> > identity to it - it's there so that you can put computer-generated
> > text to enfilades if you want.
> >
> > The documentation apparently isn't clear enough yet ;)
> 
> It's not, but I understood it anyway (it's not the first time I thought about 
> the problem :) ). But it makes for unresolved difficulties:
> - What if the user copies the text elsewhere? Does it become a real span?

No, never.

> - What if the user enters own text in the same cell? If they make it a "user 
> cell", I'd expect the span to become "real".

Those spans that the user types are real. Those he doesn't, don't.

Think of Fake Spans as spans from a different scroll, every copy of a fake
span is from a different point of that scroll.

> - IF the span becomes real when copying elsewhere, does the origin of the 
> copy become that span, too? If not, when we copy the same fake span two 
> times, it won't be a transclusion.

Yes, it won't, that's the point.

> The point here is: When the user takes computer-generated text and uses it 
> somehow, it arguably becomes real text.

There should be a very explicit conversion process.

> Now, what do we gain if we save fake spans as a String? Some clarity, ok-- 
> the blocks contain only text typed by humans. But else? Is it more efficient 
> to save the text in the space diff instead of the scroll? Why? I think 
> there's no big difference.

There is: consider slurping etc. If using real spans, then all the coordinates
outputted to cells by PP are *always* appended to a span.

But the clarity is the more important issue: as Ted has said, the scroll
should be all the characters types, no less, no more... And that has
a very good ring to it.

> The *real* efficiency problem I do see is when computer-generated text 
> changes: for example, a cell keeping the current mouse position. It would 
> pollute the scrolls with data.

*Exactly*.

> An elegant solution to this would be, as I said, to save only those spans 
> that are currently inside some cell when the space is saved.

Might be but to me it doesn't feel right at all.

> But as I said, it's not an idea I propose for now.

Ok

        Tuomas