logo separator

[mkgmap-dev] nearby POIs

From Mike Baggaley mike at tvage.co.uk on Sun May 3 08:45:46 BST 2020

HI Gerd, that sounds good to me. Could you create a class containing the
node and the POI and collect these in a list rather than just the POIs or
would that be too much overhead just for logging?

Regards,
Mike

-----Original Message-----
From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com] 
Sent: 03 May 2020 07:19
To: Development list for mkgmap <mkgmap-dev at lists.mkgmap.org.uk>
Subject: Re: [mkgmap-dev] nearby POIs

Hi Jan,

yes, the data structure used for this new option is too simple when multiple
unnamed or equally named POI at very different locations are handled.
The first processed POI is stored and for all further POI the position is
compared with that first. The algo is not able to handle multiple clouds
like in your example.

@Mike: My approach would be to collect all POI, even those which appear at
the same location, in a list. This list would be reduced with a single
method call  in StyledConverter.end().
Small disadvantage: You cannot log the node details.
I'll work on this idea, if you prefer another just go ahead and we can
compare the solutions.

Gerd

________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von jan
meisters <jan_m23 at gmx.net>
Gesendet: Samstag, 2. Mai 2020 22:28
An: Development list for mkgmap
Betreff: Re: [mkgmap-dev] nearby POIs

Hi Mike, hi Gerd,

thanks for the nearby option, I like the new way to unclutter the map.

To my args-file i´ve added:
nearby-poi-rules=*/named:10,*/unnamed:30,0x2f00-0x2f1f:30,0x3200-0x331f:50,0
x13700-0x1370d:30

A lot of the pois seem to be reduced (deleted) as expected, but I stumbled
on tourism=picnic_site and leisure=picnic_table here:
https://www.openstreetmap.org/way/498671746
https://www.openstreetmap.org/way/108892873

These 2 polygon tourism=picnic_site contain picnic pois, tourism in the one
and leisure in the other.
My rule combines the pois in 0x2f0b:
tourism=picnic_site | leisure=picnic_table [0x2f0b resolution 24]

As long as tourism and leisure are collected in one poi, both poi will not
be reduced.
If separated to different pois, tourism is reduced, leisure not.
I can´t get the option to work on both.

Regards
Jan


> Am 29.04.2020 um 18:06 schrieb Gerd Petermann
<gpetermann_muenchen at hotmail.com>:
>
> Hi Mike,
>
> OK, did not see this line
>  fileRules.addAll(Arrays.asList(rules));
>
> So, I think I'll commit the patch tomorrow if nobody else suggests
changes.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike Baggaley <mike at tvage.co.uk>
> Gesendet: Mittwoch, 29. April 2020 17:52
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd,
>
> I've replaced the trim statements with replaceAll in the attached updated
> patch. Having thought a bit more about caching the rules, I've realised in
> the multithreaded environment, there would need to be object locks and
> synchronisation, which would be unnecessarily complicated.
>
> Regarding the two options - I had 30 odd rules in my command line, so the
> config file is better for me, but I expect that most users will probably
> just use one or two rules. Your range suggestion has allowed me to reduce
> the number of rules to about 25. The config file option doesn't override
the
> inline option; if both are provided then both sets of rules are applied -
> hence I don't think the documentation needs changing.
>
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 29 April 2020 07:28
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> regarding error handling I think we can wait for someone to complain. Each
> solution has its pros and cons.
> regarding option handling:
> - I have no idea what users do. It is possible to create very different
maps
> with a single execution of mkgmap. I can't think of any good reason to do
> this, but I also see no way to find out if anybody does. I presume that
many
> users don't read this list.
> - If you see a simple solution to cache the results which works fine with
> multiple threads using different options go ahead.
> - When I suggested --nearby-poi-rules-config=filename I thought that is
> should replace --nearby-poi-rules. Having both options simply complicates
> everything. If you think that it is unlikely to have more than two rules
we
> should better remove --nearby-poi-rules-config=filename, else it should be
> documented that --nearby-poi-rules-config=filename overwrites
> --nearby-poi-rules.  Sorry for suggesting it in the first place.
> - all those trim() statements in combination with substring() look error
> prone. If I got that right you could use rule.replaceAll("\\s+", "") once
to
> remove all whitespace characters?
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Dienstag, 28. April 2020 22:06
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd, please find attached an updated patch that allows for # comments
> either as whole lines or as ends of a line. It also allows whitespace
> between the components of the rules.
>
> At the moment, if an error is found in a rule, this is reported as an
error
> and that rule is ignored. This does result in duplicated error messages,
one
> from each tile. I'm happy to change this to stop processing if you think
> that is better. I'm not keen on ignoring all rules if an error is found in
a
> rule. If the rules file is not found then this will also just report an
> error and continue. Perhaps this one at least would be better if it caused
> termination?
>
> I did think about loading the rules at the beginning, prior to processing
> the tiles, but realised this would break the 'option before tile'
convention
> which allows different options to be passed to each tile (does anyone do
> this for anything other than the description?). It would be pretty
> straightforward to cache the options and resulting rules and if the
options
> are the same for the next tile, use the cached rules instead of
> reprocessing. What do you think?
>
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 28 April 2020 09:21
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> I did not try it, but it seems the configuration file doesn't allow the
> typical "#" as comment. This makes it rather useless. We should at least
> allow comment lines starting with a # to be ignored.
> I've added code to handle this, also untested.
>
> - What should happen when an error was found in the rules?
> + Ignore only the bad rule(s)?
> + Ignore all rules?
> + stop processing?
> We have this problem with other options like --style-option as well, so
it's
> not directly related to the patch. I think It would be good to check this
> before any tile is processed, but the current implementation of the option
> handling makes this really difficult.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Montag, 27. April 2020 16:06
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd, please find attached an updated patch that addresses the points
> below.
>
> Cheers,
> Mike
>
> -----Original Message-----
> From: Mike Baggaley [mailto:mike at tvage.co.uk]
> Sent: 22 April 2020 17:55
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd,
>
> To answer your points:
>
> 1) 0x2f1f was just a POI in my style which I have as a bus stop, and as
bus
> stations often have all individual stops mapped it can be useful to remove
> duplicates (I don't use the default style). Looking at the default style I
> see that it has both bus stops and taxi as 0x2f17. I'm happy with either
the
> example changing to 0x2f17 or the default style changing to 0x2f1f for bus
> stops (I have 0x2f17 just for taxi). The sample doesn't have to match the
> default style, but it would make sense for it to do so.
>
> 2) The code sorts the rules into POI type order, but doesn't sort by
> distance. It would better handle the example you have given if it sorted
> rules with the same POI type by distance. I'll look into this.
>
> 3) Allowing a range of POI types seems like it could be useful. I see that
> --poi-excl-index uses a hyphen to specify a range and propose we use the
> same format. I'll look into this as well.
>
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 22 April 2020 07:07
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> okay, I started to find out if the code works as documented. Here is a
first
> set of questions:
> 1) I started with the given sample
> --nearby-poi-rules=*/named:10,*/unnamed:25,0x2f1f:30
> and found out the the default style doesn't use type 0x2f1f. Is that
> intended?
> 2) Is it allowed to use multiple rules for the same POI type?
> Two rules like
> --nearby-poi-rules=0x2f1f:10:merge-at-mid-point,0x2f1f:1
> are accepted, but I am not sure what to expect when this is done. The
rules
> are applied in the given order, so the second will never match, right?
> 3) The option --poi-excl-index allows to specify ranges of POI types.
> Wouldn't it be good to use the same logic for --nearby-poi-rules?
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Mittwoch, 22. April 2020 01:03
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> HI Gerd, the mods look fine to me. I don't think documenting a config file
> in a sample file is a good idea. I think it is much better if all the
> documentation is in one place. I also think it make sense to keep the
> --nearby-poi-rules option as I expect some users would just want something
> like -- nearby-poi-rules=*:25, and it would be overkill to have to create
a
> config file to do this. I have fixed the formatting issues in the options
> file in the attached version of the patch.
>
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 21 April 2020 08:46
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> attached is a modified version of the patch. I've removed some of the
> changes in StyledConverter and did some cleanup on the NearbyPoiHandler.
>
> I think you got me wrong regarding the config file. With similar to
> --road-name-config=filename I really meant that the config file should be
> provided as an example containing the detailed documentation.
> The option --nearby-poi-rules should be removed, it just adds complexity.
> This will also fix some formatting problems in file options.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Montag, 20. April 2020 16:00
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd, the logging changes were actually part of another patch that is
> aimed at tidying up logging, which causes me significant problems due to
the
> mixed use of System.err.println and log.xxxx. If you want to remove the
> System.error.println change from the patch then that is fine.
>
> Regards,
> Mike
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 20 April 2020 07:25
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> I agree about the changes to debug level because the remaining log.info()
> messages are still meaningful.
>
> You changed some System.err.println(...) to log.info(...).
> This has two effects:
> 1) severity is changed from error to info (which is wrong in these cases),
> the user should always see these messages.
> 2) the normal logging procedure prints the path to the current tile but
the
> tile is not relevant and thus may be confusing.
>
> I am not 100% sure about the messages log.warn("road has < 2
> points",way.getId(),"(discarding)");
> You probably only see them when you disable WrongAngleFixer by modfying
the
> code. What was the reason for the change?
>
> Gerd
>
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Sonntag, 19. April 2020 15:24
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd, the logging changes seem to me to be correct (though unrelated to
> the nearby POI change), reducing the severity of some messages to match
> those elsewhere. I accept that these things are open to interpretation,
but
> would consider messages about entering and leaving functions and similar
> static code location messages to be debug level messages (e.g.
"Maintaining
> highway counters"), those that give information that the code is doing
> something that does not need any action to be informational messages (e.g.
> "Pruning point xxxx"), those that might warrant investigation would be
> warning (e.g. Oneway road xxx goes nowhere" ) and those that indicate some
> sort of real problem would be error (e.g. "shape is not closed with
> identical points ").
>
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 19 April 2020 08:41
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> thanks, will look at it during the next days. Some of the changes to the
> logging in StyledConverter are probably not OK.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Sonntag, 19. April 2020 08:45
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] nearby POIs
>
> Hi Gerd, please find attached an updated patch that provides a config file
> option and puts the nearby POI code into a new class.
>
> Cheers,
> Mike
>
> -----Original Message-----
> From: Mike Baggaley [mailto:mike at tvage.co.uk]
> Sent: 08 April 2020 14:02
> To: 'Development list for mkgmap' <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: RE: [mkgmap-dev] nearby POIs
>
> HI Gerd, I'll look into having a config file option and separating the
code
> into a new class. Probably will be a couple of days before I get back to
you
> with an updated version of the patch.
>
> Cheers,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 07 April 2020 09:07
> To: Development list for mkgmap <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] nearby POIs
>
> Hi Mike,
>
> the syntax for the option parameters is extremely complex. I think it
would
> be better to use a config file for that, similar to the option
> --road-name-config=roadNameConfig.txt we could have a
> --nearby-poi-rules=nearbyPoiConfig.txt
> What do you think?
>
> Gerd
>
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von
Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Montag, 6. April 2020 20:24
> An: 'Development list for mkgmap'
> Betreff: [mkgmap-dev] nearby POIs
>
> Hi Gerd, please find attached a patch for handling the removal of
duplicate
> POIs that are near to each other but not coincident. The existing code
> handles the removal of coincident duplicates, but leaves other duplicates
in
> place. It is very common in OSM for buildings to be tagged as amenities
with
> a single node also being tagged as the same amenity. If using
> --add-pois-to-areas, this can lead to a lot of duplication. The patch
> provides a new command line option with the following syntax:
>
>
--nearby-poi-rules=type[/all|/named|/unnamed]:max-distance[:delete-poi|delet
> e-name|merge-at-mid-point][,...]
>
> This defines a set of rules to follow when a POI is near to another of the
> same type. Each rule consists of three parts separated by colons. The
> first two parts must be provided; the last part can be defaulted.
>
> The first part of the rule is the Garmin POI type code with an optional
> suffix; it determines when the rule is triggered. The type code may be
> specified in decimal or hexadecimal (e.g. 0x2c0b). A rule is triggered
> when processing a POI if the type code of the POI matches the rule type,
> providing there is also a match in the POI name and the first part
> suffix. If the suffix is '/all' (the default) then the match is only made
> on the type. If the suffix is '/named' then the rule is only triggered if
> the POI has a name. If the suffix is '/unnamed' then the rule is only
> triggered if the POI has no name. A wildcard of an asterisk character may
> be used to match any type code. The wildcard may also be combined with a
> suffix to allow separate processing of named and unnamed POIs.
>
> The second part of the rule is the distance in metres which an already
> processed POI must be within for it to be considered to be nearby and
> hence trigger the action part of the rule.
>
> The third part of the rule is the action part and provides three options:
> ::delete-poi - the POIS are considered to be duplicates and the
> duplicate is deleted. This is the default.
> ::delete-name - the POIS are not duplicates, but only a single
> name needs to be displayed.
> ::merge-at-mid-point - the POIS are considered to be duplicates
> but the location of the existing point is moved to a point
> midway between the two points.
>
> Wildcard rules are only applied if no other rule is applicable.
>
> For example:
> :
>
--nearby-poi-rules=*/named:50,*/unnamed:25,0x2f1f/named:50:delete-name,0x2f1
> f:3,0x641d:400:merge-at-mid-point
>
> This has the following effect:
> :       If no other rule applies, a POI with the same name and type and
>        within 50m of one already processed will be deleted
>        If no other rule applies, a POI having no name and of the same type
>        and within 25m of one already processed will be deleted
>        A POI of type 0x2f1f that is within 50m of another POI with the
same
> name and type will have its name deleted
>        A POI of type 0x2f1f that is within 3m of another POI with the same
> type will be deleted
>        A POI of type 0x641d that is within 400m of another POI with the
> same
> type will be deleted and the existing point moved
>
> Note: a POI that matches another in type, name and exact location is
always
> considered a duplicate and deleted.
>
> Regards,
> Mike
>
>
>
> _______________________________________________
> mkgmap-dev mailing list
> 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
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>
>
>
>
> _______________________________________________
> mkgmap-dev mailing list
> 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
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev




More information about the mkgmap-dev mailing list