[LaurentG]

I’m wondering how POI are selected to be present or not in the .poi files.

I have the example of two churches in Briançon (Hautes-Alpes, France), that seem very similar in their OSM definition, but one is present in the .poi files, while the other is not

The one that is present owns the following attributes :
amenity=place_of_worship
building=church
denomination=catholic
name=Église Sainte-Catherine
ref:FR:CEF=05023_01
religion=christian
source=extraction vectorielle v1 cadastre-dgi-fr source : Direction Générale des Impôts – Cadas. Mise à jour : 2010
wikidata=Q29545197

And the one that is missing owns these attributes
amenity=place_of_worship
building=church
check_date=2025-09-21
denomination=catholic
email=paroisse.briancon@diocesedegap.com
heritage=2
heritage:operator=mhs
internet_access=no
layer=1
mhs:inscription_date=1931-10-14
name=Église Notre-Dame, Saint-Nicolas (Ancienne Collégiale)
phone=+33 4 92 20 04 10
ref:mhs=PA00080528
religion=christian
source=cadastre-dgi-fr source : Direction Générale des Impôts – Cadastre ; mise à jour : 2009
source:heritage=data.gouv.fr, Ministère de la Culture – 2016
website=https://www.egliseenbrianconnais.net
wikidata=Q3580438
wikipedia=fr:Église Notre-Dame-et-Saint-Nicolas de Briançon

What makes that the first one is present, but not the second one.
Id‘ like to know to be able to add missing data (if the problem comes from missing data….) in OS to have the second one present in .poi files

Thanks in advance for explanations !

[mbe57]

Bonjour Laurent,
I just tried with france-south-oam poi file and both churches were present in that poi file, dated Aug 22nd, 2025.
Let’s compare file creation dates …
Cheers
Michael

[LaurentG]

Yes, it depends on file generation.
In file dated April 2024, 24th, only the one I mention is there, and not the other.
And in latest rev of the file January 2026, 3rd, NONE of them are here

It’s the reason why I’m really interested in knowing, directly from file’s author, whar are his/her selection criteria !

[mbe57]

„his“ – Christian, should be able to tell/find out in his tool configuration.

[ChristianK]

At a first look:
Both are in the raw poi file.
Found both by the wikidatat IDs in France South.

If these are in the final files i have to check.

Question: how do you found that these objekts are missing ?
Dont you see them in the APP (which one) on the map (with poi activated)
or
have you checked the POI-Files (SQlite3) if the Wikidata IDs are present ?

Best regards
Christian

[ChristianK]

Hi Laurent,

first is Way ID 73576946
second is Way ID 73577250
Both exist in the raw prefiltered POI OSM file
Both are missing in the mapsforge PoiV3 database
Both exist in the old mapsforge PoiV2 database and in the LocusPOI Database

PoiV3 and PoiV2 share the same source, filter, preprozessing, poi_tagmapping

So these Pois are lost somewhere in the mapsforge PoiV3 Poiwriter.
This is a serious issue and I will have to take a breif look at this.
Maybe the poimapping have to be different for the new PoiWriter ????

Thanks for bringing this to my attention.
Best regards
Christian

[LaurentG]

Hi Christian,

to answer your question, I first detected the missing in OruxMaps, with the „Local Search“ feature, and was surprised to do not find it.
So I double checked directly in the .poi file with SQlite Browser, searching for all POI present in a (small…) square of lat/lon coordinates.

I stumbled upon the problem completely by chance. If it led to the detection of a real issue in the generation of .poi files, then it’s a real stroke of luck!

Good luck in finding the real issue !
Best regards

Laurent

• Diese Antwort wurde vor vor 3 Wochen, 3 Tage von LaurentG bearbeitet.

1 Teilnehmer(n) gefällt dieser Beitrag

ChristianK

[ChristianK]

Found it.

There is a new switch in the Mapsforge PoiWriter that skips ways with tagged nodes per default.
This switch kicks in not only if there is a Node tagged the same way as the found way_poi (as I thought) – it looks for any Node-taggings (at least with the Snapshot_Poi_Writer I use)
In the case of these two churches there is an entrance and two Artwork Nodes in the ways that define the church. – so these two churches were skipped.

This switch can be set to false – and the problem is solved.
Both IDs now appear in the PoiV3 DB.

I will update and upload the Pois in the next days.
The France_South Pois will be online tomorrow for testing.

Thanks again
and
Best regards
Christian

2 Teilnehmer(n) gefällt dieser Beitrag

LaurentG, mbe57

[LaurentG]

Not surprised that problem comes from data, and not directly from your tool, since POI files provided directly on https://download.mapsforge.org/ have the same issue… and AFAIK, they are not generated by you….

I’m not sure to understand the reason that motivate such switch. A church with an artwork on it are two „distinct“ POI, the church itself, and the artwork.
And even more, a switch that lead to eliminate a church because its entrance gate is declared on one of the nodes describing the building, doesn’t, IMHO, make any sense…
This is obviouly true not only for churches, but also for all „public“ buildings that can be considered as a POI.

As soon as you have re-generated France_South Pois, I’ll have a look on it.
Afterwards, I don’t know which POI files you plan to re-generate, but I would be personnally very interrested in Alp_West file, since it’s the map I most frequently use !

Thank you so much for your investigation.
Best regards
Laurent

[ChristianK]

Hi Laurent,

I have no idea for what purpose this switch is for – usually EMUX (the man behind Mapsforge) nows what he is doing.

https://github.com/mapsforge/mapsforge/blob/master/docs/POI.md
Nevermind….

All POI files are now on the transfer-box and will be available for download tomorrow, 15.01.2026 from 9:00am
The 5G mobile-connection is a little bit on the expensive side but really awesome…..
🙂

Best regards
Christian

[LaurentG]

I do the bet that in your new POI file, there will be a lot of POI that will re-appear.
In particular, around my home, a lot of churches and/or chapels are not currently in the file, only because one of their node has been identified as the entry gate…. And I guess that it is not only in french Hautes-Alpes that such parameterization is done.

Moreover, a lot of them have sundial („cadran solaire“, in french, not sure of my english wording…) painted on them, and these sundials are generally defined as „ArtWork“… and are one of the chapel’s node.

In such a case, it won’t be one but two POI, skipped in current file, that will appear in the new POI file.

So, it won’t be surprising if the new POI file is a lot bigger than current one….

1 Teilnehmer(n) gefällt dieser Beitrag

ChristianK

[Emux]

This filter has always been there for memory and performance reasons.
You can read more details in the relevant issue.

At some point it became optional, after users requested it:
https://github.com/mapsforge/mapsforge/issues/1456

The library is open source. 🙂
If someone can contribute improvements, I would be happy to review them.

---

Cruiser and Mapsforge developer

[Emux]

Please try creating the entire Germany and France poi files without the filter
and see if it is possible or if memory / performance issues occur.

---

Cruiser and Mapsforge developer

[LaurentG]

Hi Christian,

I just dowloaded the new France-South.Poi file.

It is now 422 Mb vs 410 Mb : +3%
Globally, it contains now 1822767 POI (instead of 1777990) : +2.5%
If we focus on „place_of_worship“, it contains now 32199 of them, instead of 29898 : +7.7%

And in particular, it contains all the ones that were missing around my home !

Thank you so much for that.

Hi EMUX,

since Christian has been able to generate the file, it is an answer to your question, isn’t it ?
I guess (even if I’m not sure…) you are the author of POI files available on https://download.mapsforge.org/

If YES, it would be very great to re-generate them without the filter.
Thanks in advance for that !

• Diese Antwort wurde vor vor 3 Wochen, 1 Tag von LaurentG bearbeitet.

[ChristianK]

Hi Emux,

I have no issues with RAM or performance.
8 Threads in parallel consume about 32GB Ram – no big deal on a 128GB machine.
However, I use a large allowance for Java_Heap – maybe there are problems with the default osmosis java heap configuration.
The poi-writer (0.16) is use for creating the V2 Pois doesnt have this switch and provides unfiltered POIs incl ways with tagged nodes.

Best regards
Christian

[Autor]

Beiträge