[Kst] Vector skip patch ?

[Barth Netterfield]

AFAIKT, This would work in static mode, but in real time count-from-end mode, 
it will not behave well:  If you are counting from the end, with skip on, I 
assert that these three things should be true:
	-My plots should only update every SKIP frames (not every frame, as would 
happen with the patch) and 
	-old data should not change when new data comes in - it should just shift 
into the past.  I have no idea what will happen to old data in terms of 
shifting with this patch.
	-my plot should look like your plot, even if we started them at different 
times.

This was the reasoning behind the old behavior - that new_F0 always be on a 
skip boundry.  One can argue that for defined frame range or for read to end 
mode, new_F0 could remain as set, but for count-from-end, it must be on a 
skip boundry.  But then we are looking at a surprising difference in behavior 
between modes - hence the idea that skip reads always start on skip 
boundries.

Sorry that I was slow on this response.  I will look at your tests now.

cbn

	
On Monday 16 October 2006 13:08, Adam Treat wrote:
> *** DO NOT RELY ON THIS GOING IN AS IS ***
>
> Ok, I've attached the patch, but realize that this is the minimum diff
> necessary to make rvector pass the tests.  *my tests* which have not been
> verified to be correct.  That means that this is subject to change and the
> behavior is subject to change.
>
> Another caveat, not that this will not work for datasources that use the
> skip acceleration.  Right now, dirfile is the only datasource that this is
> tested to work for.  I'm working on a refactor of rvector at the moment
> while I wait to hear back about the testcases.  Again, things could change
> quickly...
>
> *** DO NOT RELY ON THIS GOING IN AS IS ***