[postgis-devel] ST_Buffer - single sided option

Mon Feb 26 09:43:47 PST 2018

Hi Paul,

I merged in Stephen's patch for this, which uses the "side=left/right"
syntax. He made some nice examples for the docs:

https://postgis.net/docs/manual-dev/ST_Buffer.html

The "sign" and "side" are two ways to refer to the same thing, i.e., a
negative left buffer is the same as a positive right buffer:

https://trac.osgeo.org/postgis/browser/trunk/postgis/lwgeom_geos.c#L966

Dan

On Mon, Feb 26, 2018 at 12:37 PM, Paul Ramsey <pramsey at cleverelephant.ca>
wrote:

> Yes, left/right seems reasonable to me.
> It's not an east/west thing though, it's a line direction thing, so
> "left in the direction of digitization". I haven't experimented what
> that would mean for a polygon... in particular, if you asked for a
> right-side buffer on a CW ring, with a negative buffer distance...
> what do you get? The right/left side and positive/negative distance
> would appear to be interacting somewhat.
> P.
>
> On Fri, Jan 19, 2018 at 1:23 AM, Darafei "Komяpa" Praliaskouski
> <me at komzpa.net> wrote:
> > Hi,
> >
> > I like the idea of changing 1/-1 to left/right, it's more humane.
> > How about side=left/right/both for less presses of shift key? :)
> >
> > As a person who mixes left and right I would also like have a setting
> that
> > doesn't mention "left" and "right" at all.
> > It's visually tricky sometimes, to convert the screen's "east" to "right"
> > for a line that goes "southwards/down", and then write that into query
> from
> > first attempt.
> >
> > I'll quote again a property I expect to hold true without mentioning
> "left"
> > and "right":
> >>> I expect ST_Buffer(ST_Boundary(ST_ForceRHR(polygon)), 10,
> 'single_side')
> >>> to be buffered "outside".
> >
> > Or at least a highlight in docs that will tell me whether it's left or
> right
> > each time I forget :)
> >
> > пт, 19 янв. 2018 г. в 3:30, Stephen Knox <stephenknox73 at gmail.com>:
> >>
> >> Hi Darafei,
> >>
> >> Thanks for that good feedback.
> >>
> >> Actually I think thinking it through I wonder if it would be better to
> >> have the option as single_side=left or single_side=right as I think
> that is
> >> much more intuitive. All the tests I've done on geometries indicate that
> >> geos honours the distance parameter as expected (positive is left hand
> side
> >> of the line and negative right - see
> >> https://geos.osgeo.org/doxygen/classgeos_1_1operation_1_1buffer_1_
> 1BufferParameters.html#adc2afac5dacc613a39a72542bdb8bf21).
> >> The only issue is by default it gives right hand buffers round joins and
> >> left hand buffers mitre joins, but this can be overridden.
> >>
> >> Really this seems to be of greatest use for linestrings, as the buffer
> >> function already had a negative distance parameter for polygons, and
> whether
> >> single_side is true or not does not make any difference. With points
> again
> >> it ignores the parameter and undertakes a normal buffer, which is the
> only
> >> sensible thing to do.
> >>
> >> I'll work on that basis and add in the code, tests and docs for the
> >> revised function and also submit a pull request from Github.
> >>
> >> Thanks again
> >>
> >> Stephen
> >>
> >> On Thu, Jan 18, 2018 at 10:23 AM, Darafei "Komяpa" Praliaskouski
> >> <me at komzpa.net> wrote:
> >>>
> >>> Hi,
> >>>
> >>> Thoughts that I have:
> >>>
> >>>  - this is a cool thing that I've needed multiple times. I worked
> around
> >>> its absence by fiddling with ST_MakeLine(ST_Collect(line,
> >>> ST_OffsetCurve(line, 1))) and then trying to make result valid, which
> >>> sometimes worked;
> >>>  - how do I remember which way is it being offset? I expect
> >>> ST_Buffer(ST_Boundary(ST_ForceRHR(polygon)), 10, 'single_side') to be
> >>> buffered "outside". Does it hold true?
> >>>  - if there are no better ideas on ST_Buffer signature I'd say your
> >>> option is good enough for me;
> >>>  - there's no error reporting mentioning `single_side` as valid option;
> >>>  - can `single_side=1` also work for `single_sided=1` and just
> >>> `single_side`/`single_sided`?
> >>>  - what does it do for line and for point?
> >>>  - does it work on EMPTY, LINESTRING(0 0, 0 0), MULTILINESTRING, and
> >>> self-intersecting one?
> >>>  - how do I reverse the direction of buffering? I understand there's
> >>> negative buffer parameter, does it work? I see you parse parameter to
> be
> >>> single_side=1/0, maybe -1 will also make sense?
> >>>  - out of curiosity, will ST_Buffer(geom,0,'single_side') fix invalid
> >>> geometries? :)
> >>>
> >>> As someone responsible for github in PostGIS, I encourage you to put it
> >>> to GitHub as a PR: that'll get you test coverage report, travis CI
> >>> greenlight, a way to comment on your patch inline, and usual PR merge
> time
> >>> on it recently improved. :)
> >>>
> >>> чт, 18 янв. 2018 г. в 12:56, Stephen Knox <stephenknox73 at gmail.com>:
> >>>>
> >>>> Hi List,
> >>>>
> >>>> I noticed that PostGIS didn't implement the single-sided buffer
> >>>> functionality of Geos (NB. not the Offset Curve functionality, but a
> >>>> polygon), so have submitted a patch to implement that:
> >>>> https://trac.osgeo.org/postgis/ticket/3989
> >>>>
> >>>> I have refrained from adding tests and documentation until I get some
> >>>> agreement that the function signature is acceptable, as I am a new
> >>>> contributor, and I think the ST_Buffer signature is already quit
> "unusual"
> >>>> in that it implements text parameters.
> >>>>
> >>>> I was hoping that post the recent patch releases, a more experienced
> >>>> contributor might be able to look at my (small) patch and give some
> >>>> feedback.
> >>>>
> >>>> I can put it up on Github as well if that would help.
> >>>>
> >>>> Many thanks
> >>>>
> >>>> Stephen
> >>>> _______________________________________________
> >>>> postgis-devel mailing list
> >>>> postgis-devel at lists.osgeo.org
> >>>> https://lists.osgeo.org/mailman/listinfo/postgis-devel
> >>>
> >>>
> >>> _______________________________________________
> >>> postgis-devel mailing list
> >>> postgis-devel at lists.osgeo.org
> >>> https://lists.osgeo.org/mailman/listinfo/postgis-devel
> >>
> >>
> >> _______________________________________________
> >> postgis-devel mailing list
> >> postgis-devel at lists.osgeo.org
> >> https://lists.osgeo.org/mailman/listinfo/postgis-devel
> >
> >
> > _______________________________________________
> > postgis-devel mailing list
> > postgis-devel at lists.osgeo.org
> > https://lists.osgeo.org/mailman/listinfo/postgis-devel
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/postgis-devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20180226/7896bb69/attachment-0001.html>