[mkgmap-dev] Commit r4398: revert changes from r4397 (--is-in-landuse option)

Sat Dec 21 08:16:54 GMT 2019

Hi Jan,

the currently implemented hook never splits a way, it just calculates the tag mkgmap:residential with a very lazy and unprecise test. It was meant to detect buildings inside residential areas. It works quite well for that, but produces more or less unprecise results for the highways in the example file.
There are some other routines which split ways, for example roundabouts or p-shaped roads are split. Maybe you see that effect in your map. The RoadMerger sometimes produces p-shaped ways.

I think it would be possible to split an (unclosed) way where it enters or exits a given polygon type, maybe it would make things easier to handle in the stye. Question is if that should also be done for places where the ways don't share a node with the polygon (can cause line distortion), an example would be w15 which would be split into 3 parts. It's likely that the way would no longer be straight after that because of the poor Garmin resolution.

In fact we want such a split method to be able to create map tiles with a non-rectangular shape, but that is another problem.

I'll post a new file containing a few more special cases in a new thread.

Gerd

________________________________________
Von: jan meisters <jan_m23 at gmx.net>
Gesendet: Samstag, 21. Dezember 2019 00:31
An: Gerd Petermann
Betreff: Re: [mkgmap-dev] Commit r4398: revert changes from r4397 (--is-in-landuse option)

Hi Gerd,

ok, I didn´t know about this radius. That really explains a lot of what I called erratic results.

So straddle means a line to be split accordingly?
I haven´t imagined such a split possible on non existing points, that´s why I mentioned it happening in r4398.
For lines travelling on borders of polygons I´m not sure if straddle is always good.

Treat overlapping/touching polygons as one - yes, reasonable.
But I never really faced the effect on areas when trying the hook.

Jan

Am 20.12.2019 um 23:12 schrieb Gerd Petermann <gpetermann_muenchen at hotmail.com<mailto:gpetermann_muenchen at hotmail.com>>:

Hi Jan,

my understanding of Tickers definition is that straddle means partly in and partly out.
I didn't look at the results produced with any mkgmap version. I want to define what results we want to get.
I agree that the track is outside of the landuse=residential way. Current mkgmap searches a radius of ~ 3 m for a residential area,
that might explain the wrong result.
I agree that we probably want in for w26. Probably it would be best to combine overlapping or touching polygons so that they are treated like a single one.

Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto:mkgmap-dev-bounces at lists.mkgmap.org.uk>> im Auftrag von jan meisters <jan_m23 at gmx.net<mailto:jan_m23 at gmx.net>>
Gesendet: Freitag, 20. Dezember 2019 22:54
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] Commit r4398: revert changes from r4397 (--is-in-landuse option)

Hi Gerd,

great job!

Concerning my english: I understand straddle as effectively outside, correct?
Until now I only focussed on lines. In the samples every scenario of them looks good to me, if straddle=out. If not, hmm.
For w26 i expect in.

Attached is another example, stripped from OSM for residential also.
By 4397 the whole track is rendered as in.
With patched 4398 the line is split at the top west point of the residential llanduse and from there processed as in to the east.
But I would expect it as completely out.
(The original is here: https://www.openstreetmap.org/way/17937104, and I ask for allotments as in)

Jan

Am 20.12.2019 um 19:56 schrieb Arndt Röhrig <arndt at speichenkarte.de<mailto:arndt at speichenkarte.de><mailto:arndt at speichenkarte.de>>:

Hi Gerd,

with r4397 i used this:

mkgmap:lu:cemetery=* &! ((route_mtb=*|route_icn=*|route_ncn=*|route_lcn=*|route_rcn=*) | bicycle~'yes|official|designated|dismount|ok') {set rad=nein}

and this

highway~'unclassified|minor' &!(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10112 resolution 24 continue]
highway~'unclassified|minor' &!(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10113 resolution 23-22 continue]
highway~'unclassified|minor' &!(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10114 resolution 21-21 continue]
highway~'unclassified|minor' &!(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10115 resolution 20-20 continue]
highway~'unclassified|minor' &(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10112 resolution 24 continue]
highway~'unclassified|minor' &(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10113 resolution 23-23 continue]
highway~'unclassified|minor' &(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10114 resolution 22-21 continue]
highway~'unclassified|minor' &(mkgmap:lu:residential=*) &!(tunnel=*) { name'${LangBez}'} [0x10115 resolution 20-20 continue]

I think it´s nice, to expand the function mkgmap:lu:cemetery, so that some other polys (playground, military, deponie etc) work with this.

i don´t understand all mails, because my english is .....(:

Thank you for all the work you do!

best regards

Arndt
Gerd Petermann < gpetermann_muenchen at hotmail.com<mailto:gpetermann_muenchen at hotmail.com><mailto:gpetermann_muenchen at hotmail.com>> hat am 20. Dezember 2019 um 18:32 geschrieben:

Hi all,

attached is an OSM file with various test cases. Assume we want to get the result for landuse=residential.
Please load it in JOSM and search for expected=in or expected=out or expected=straddle and finally expected="?"
and let me know as soon as possible if you don't agree with anything. (each case has a different name for the object to be tested.
The cases with "?" are special, I think b9 and b10 are not important, but the others are very usual in real OSM data.
The existing ResidentialHook would add mkgmap:residential to b13, b14 and w26.

@Ticker:
The residential areas around b13 and b14 show the special case with multipolygons. The hook probably "sees" something like the b13 example when looking at b14 because the multipolygons are cut into pieces without holes before the hook is executed.

@all: Would be very good to get some feedback on this, even if you think that you don't need the hook for your style.

Gerd

________________________________________
Von: mkgmap-dev < mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto: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><mailto:rwb-mkgmap at jagit.co.uk>>
Gesendet: Freitag, 20. Dezember 2019 13:49
An: mkgmap development
Betreff: Re: [mkgmap-dev] Commit r4398: revert changes from r4397 (--is-in-landuse option)

Hi

Given the complexity and cpu cost of calculating this, the various
questions, hence answers possible and it has to be done for all
elements if the --is-in option is given what about making it a function
instead.

This means that it only needs to be computed for elements where the
answer matters and, as a parameter, the question about what the
required is-in means can be asked.

eg, in points:

if building=church and is-in('landuse', 'cemetery')
{add name='Chapel of Rest'}

in lines:

if highway=* and is-in('landuse', 'cemetary', 'fully') {add bicycle=no}

etc.
Above are just approximations of how the parameters might work.
The various levels of is-in that might be required should be
considered, as per Gerd's comments.

For some of the examples that have been given for use of this facility,
I feel they should really be solved by accurate tagging.

@gerd Although it looked like it, I didn't really intend that elements
were tagged with NO if no part was within the polygon.

Ticker

On Fri, 2019-12-20 at 10:43 +0000, Gerd Petermann wrote:
Hi all,
I think I have now the needed methods to be able to distuingish if a
given node is inside, outside or "on boundary" of a polygon.
In (1) Ticker suggested to use these rules to determine what tag the
hook should add either IN, OUT, or STRADDLE as a tag value.
The current implementation in ResidentialHook adds either no tag
(meaning outside) or a tag with the value "yes" or the name of the
found residential area.
The typical rule to use this tag would be
mkgmap:residential=* {do something}
I think we should NOT add the tag with a value OUT, else the above
rule will fail. On the other hand, I assume that nobody use a rule
like
mkgmap:residential=* & mkgmap:residential!=yes {do something with the
name contained in mkgmap:residential}
I think it was not a good idea to use the name, I used it for
debugging.
Open question: Should we really just count points?
Assume you have a U-shaped cemetery and a highway=footway with just
two points starts inside the left upper part and ends inside the
right upper part. Most of the way would be outside but the points are
inside the cemetery. No idea how often this happens, but the result
would be "IN" with Tickers rules. Even when you add (barrier=gate)
nodes on the boundaries of the cemetery the result is still "IN"
I woud call this result wrong and expect a value like STRADDLE.
Gerd
(1) http://gis.19327.n8.nabble.com/Commit-r4397-implement-and-documen
t-new-option-is-in-landuse-value-value-tp5953495p5953703.html
________________________________________
Von: mkgmap-dev < mkgmap-dev-bounces at lists.mkgmap.org.uk<mailto:mkgmap-dev-bounces at lists.mkgmap.org.uk><mailto:mkgmap-dev-bounces at lists.mkgmap.org.uk>> im Auftrag
von svn commit < svn at mkgmap.org.uk<mailto:svn at mkgmap.org.uk><mailto:svn at mkgmap.org.uk>>
Gesendet: Mittwoch, 18. Dezember 2019 09:24
An: mkgmap-svn at lists.mkgmap.org.uk<mailto:mkgmap-svn at lists.mkgmap.org.uk><mailto:mkgmap-svn at lists.mkgmap.org.uk>; mkgmap-dev at lists.mkgmap.org.uk<mailto:mkgmap-dev at lists.mkgmap.org.uk><mailto:mkgmap-dev at lists.mkgmap.org.uk>
Betreff: [mkgmap-dev] Commit r4398: revert changes from r4397 (--is
-in-landuse option)
Version mkgmap-r4398 was committed by gerd on Wed, 18 Dec 2019
revert changes from r4397 (--is-in-landuse option)
The test is too lazy and requires a lot more work.
http://www.mkgmap.org.uk/websvn/revision.php?repname=mkgmap&rev=4398
_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk<mailto: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><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><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><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><mailto:mkgmap-dev at lists.mkgmap.org.uk>
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev