[mkgmap-dev] Branch is_in ready for a first test

Sun Jan 12 08:58:35 GMT 2020

Hi all,

the message "SCHWERWIEGEND (IsInFunction): .\Baustelle\Speiche_Kanaren_Splitter\88002007.osm.pbf: rounding error? first point is on line but status of first point is not ON" was produced for this node:
https://www.openstreetmap.org/node/640552250#map=18/28.14036/-16.52529

The node is extremely close to the landuse=residential way 255452457. In OSM precision the node is inside the residential area, but in mkgmap precision it is outside.
These special cases really cause trouble. To improve performance, the code first tests the result for the first point. If this result is IN and method is 'any' or when result is OUT and method is 'all' there should be no need to test any further nodes or edges, thus the corresponding result is returned.
Now, with these special cases the quick result may be wrong.
With r4116 I've disabled the quick test. The result is more likely to be correct but performance can be poor when complex areas are involved.

For normal cases the difference should be small, maybe a second per tile.

Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Arndt Röhrig <arndt at speichenkarte.de>
Gesendet: Samstag, 11. Januar 2020 12:02
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] Branch is_in ready for a first test

Good morning,

so far i used "mkgmap:residential" to draw some highways thinner in citys. I replaced it with the is_in function and added some other polys like "school", "industrial" etc. It works fine, great.

I added also a "bicycle=no" on "cemetarys", "schools", "landfills", "quarry" etc.

Thank you all for this work!

Problem #1:
is_in(landuse,residential,any)=true will not work in "Wuppertal". mkmap:residential also no work there either.
I suspect the reason in the osm-data, but i don´t see why. Have anybody an idea?
https://fotos.rennrad-news.de/f3/4/494/494184-sm3yb9gt4tjr-test-medium.png

Problem, #2:
r-4412 report: SCHWERWIEGEND (IsInFunction): .\Baustelle\Speiche_Kanaren_Splitter\88002007.osm.pbf: rounding error? first point is on line but status of first point is not ON
https://speichenkarte.de/88002007.osm.pbf

greetings

Arndt

Gerd Petermann < gpetermann_muenchen at hotmail.com<mailto:gpetermann_muenchen at hotmail.com>> hat am 10. Januar 2020 um 12:07 geschrieben:

Hi Ticker,

thanks, see r4412 : http://www.mkgmap.org.uk/websvn/revision.php?repname=mkgmap&rev=4412

Gerd

________________________________________
Von: mkgmap-dev < mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto:mkgmap-dev-bounces at lists.mkgmap.org.uk>> im Auftrag von Ticker Berkin < rwb-mkgmap at jagit.co.uk<mailto:rwb-mkgmap at jagit.co.uk>>
Gesendet: Freitag, 10. Januar 2020 11:29
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] Branch is_in ready for a first test

Hi Gerd

Here is patch that sets isLineRule in the package on the augmentsWith
call.

I've also used default method for the augmentsWith... in Rule, Op,
OsmConverter and backed out the changes that have now become
unnecessary.

Ticker

On Fri, 2020-01-10 at 09:15 +0000, Gerd Petermann wrote:
Hi Ticker,
yes, it would probably be better to know if the function is for a
polyline or polygon. A patch would be welcomed.
And yes, the expected result for the case you describe is not yet
clear. See b13 and b14 in my example file.
Gerd
________________________________________
Von: mkgmap-dev < mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto:mkgmap-dev-bounces at lists.mkgmap.org.uk>> im Auftrag
von Ticker Berkin < rwb-mkgmap at jagit.co.uk<mailto:rwb-mkgmap at jagit.co.uk>>
Gesendet: Freitag, 10. Januar 2020 09:59
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] Branch is_in ready for a first test
Hi Gerd
I think the logic still needs to know if it is handling a closed way
as
a polygon rule or as a line rule.
Firstly to be able to correctly handle the case where the rule.way
runs
around a hole in the target.polygon - all of the rule.line is in the
target, but only part of the rule.polygon is in the target.
Also there might be different method keywords for the two. For
polygon
rules, "any" and "all" are adequate, but for line rules others might
be
required.
I've just svn updated the is_in branch. Is there still work to do on
non-closed ways?
Otherwise amazing.
Ticker
On Thu, 2020-01-09 at 11:43 +0000, Gerd Petermann wrote:
Hi all,
the branch version r4408 implements the style function is_in with
three parameters tag-key,tag-value,method
- tag-key can be something like landuse or natural or amenity
- tag-value would be the expected value for that tag-key
- method has to be "any" or "all".
Example usage in lines:
# no cycling within a cemetery
highway=* & bicycle!=* & is_in(landuse,cemetery,all) {add
bicycle=dismount}
Before any style rule is evaluated the function evaluates the list
of
all ways found in the input file. When a way matches the given tag
(key+value) and is closed and complete the shape geometry is stored
in a spatial index.
With the current implementation this index is created for
appearance
of the is_in function in a style rule (subject to optimization)
When the style function is called it retrieves the stored shape(s)
which intersect the boundary of the element.
If none is found "false" is returned.
If there are multiple such shapes the are merged where possible, so
that overlapping or touching shapes are combined to a single shape
which may include holes.
The list of (outer) shapes is then tested one after the other until
one shape is found which either contains the element completely or
-
when "any" is used as method - partially. If none is found the
value
"false" is returned.
If a match is found and the combined shapes have holes, a final
test
is done to find out if the element is within the hole.
If method "any" is used and the shape is completey within one hole
"false" is returned.
If method "all" is used and the shape is partly within one hole
"false" is returned.
If we get here true is "returned"
I've also slightly changed the behaviour of the LocationHook and
the
ResidentialHook. The results should be a bit more predictable now
as
searches are done with higher precsion. Still, both try to find a
nearby boundary(shape) if none is found at the exact position.
The ResidentialHook is now automatically disabled when the style
doesn't use mkgmap:residential anywhere.
TODO:
- various possible special cases need more testing
- maybe add more methods for cases where parts of the tested
element
touch the boundary of the shape.
- tuning
- unit tests
As always with branch version the binary can be found at the bottom
of the download page.
http://www.mkgmap.org.uk/download/mkgmap.html
I've tested this with the example file posted before, accept for
b18
all cases work as expected.
Gerd
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk<mailto:mkgmap-dev at lists.mkgmap.org.uk>
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk<mailto:mkgmap-dev at lists.mkgmap.org.uk>
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk<mailto:mkgmap-dev at lists.mkgmap.org.uk>
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev