Subject: [ggl] Snapping
From: Stephen Leary (sleary)
Date: 2009-11-30 17:11:03
1. Yes - Old habbit.
2. Yes - i will correct.
3. Yes and No. A geometric tolerance is used when for example two vertices
are less than the geometric tolerance a comparison operator will return
true. This is not always the numerical precision so much as the limit of
what we care about in real life. i.e. less than 1mm on a countrywide map
might be the geometric tolerance. Snapping would be 1cm in this case.
4. Higher positional accuracy would be a decision made at a higher level
based on metadata/attribute data. But the ability to perform each snap type
needs to be selectable. I.e. the tools must be available to snap either way
round and the caller should be able to pick.
5. There are no papers on this that i can freely distribute. Most of this
experience is derived from working with 3 different commercial snapping
engines and having developed one myself. The snapping engine i developed was
a 3d one so 2d/2.5d is much simpler by comparison.
6. I'm happy to once I can include everyones comments.
On Mon, Nov 30, 2009 at 9:51 PM, Mateusz Loskot <mateusz_at_[hidden]> wrote:
> Stephen Leary wrote:
> > Not 100% sure my snapping document got through so here is a link to the
> > http://www.vavi.co.uk/~sleary/Snapping_GGL.pdf
> Nice collection of cases.
> A few questions/notes:
> 1. multi-line should probably read multi-linestring as the GGL multi
> geometry is named, to avoid confusion
> 2. "snapping a the left to the geometry on the right" - a typo?
> 3. "Snapping tolerances must be > 10 times the geometric tolerance"
> What is the geometric tolerance?
> Is it an acceptable deviation of location (of single vertex)
> or anything else? Notice, there is no precision model in GGL like in
> GEOS. As far as I can understand the matter, it looks for GGL precision
> (not accuracy) for snapping would be always based on size and coordinate
> space of actual geometry.
> 4. Related to 3, determining higher position accuracy may be tricky
> without precision model, am I correct?
> 5. I miss a reference to classic papers discussing this matter, also
> would be good to have reference or short study on requirements of
> numerical robustness. At least a newbie as myself would have chance to
> learn some details.
> 6. Would you consider convert this paper to the Wiki?
> Best regards,
> Mateusz Loskot, http://mateusz.loskot.net
> Charter Member of OSGeo, http://osgeo.org
> ggl mailing list
-- Stephen Leary -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.osgeo.org/pipermail/ggl/attachments/20091130/f46ddc54/attachment.html
Geometry list run by mateusz at loskot.net