markalex2209 (May 27 2024 at 07:29): markalex2209 (May 27 2024 at 07:29):I created a PR to add deposit locations and taromats to Osmalyzer: https://github.com/OSMLatvija/Osmalyzer/pull/32
Here's file with data used there (converted to CSV for ease of working): file
HellMap (May 27 2024 at 09:51):Thanks! I assume nothing matches in Riga for kiosks because the conversion to kiosk/amenity is ongoing from that other topic?
markalex2209 (May 27 2024 at 09:52):Yes. Precisely that.
Riga has a lot of them =((
HellMap (May 27 2024 at 09:56):https://osmlatvija.github.io/Osmalyzer/Depoz%C4%ABta%20punkti%20report.html
HellMap (May 27 2024 at 11:04):What an absolute mess that data is for the fields that are supposedly machine-readable...
markalex2209 (May 27 2024 at 11:12):HellMap said:
What an absolute mess that data is for the fields that are supposedly machine-readable...
I could not agree more.
markalex2209 (May 27 2024 at 11:16):Also, there is a problem I just noticed, that cannot be solved programmatically: their map considers "punkts" as an organizational unit. And thus two kiosks near this Maxima (yet to be converted back, but you'll get what I mean) are not separate items(
And who knows how many more of this is present.
HellMap (May 27 2024 at 11:18):Hmmm... urgh. :melting_face:
HellMap (May 27 2024 at 11:22):The current Correlator© does not deal with multi-matches, but it theoretically could. I am not sure how I even would extend it to do that - this is probably soemthing extremely convoluted as now I need "range of elements that may be grouped as one logical element". I guess one could group semi-identical OSM elements close to each other into a "superelement" and let the correlator treat it as a single point.
HellMap (May 27 2024 at 11:22):I guess we can see how many of these will be on the map once some tagging is fixed, so I'm not adding complex functionality for like 2 stores
HellMap (May 27 2024 at 11:23):Although I suppose if we micromap vending machines that are in stores - they will totally repeat
markalex2209 (May 27 2024 at 11:23):HellMap said:
The current Correlator© does not deal with multi-matches, but it theoretically could. I am not sure how I even would extend it to do that - this is probably soemthing extremely convoluted as now I need "range of elements that may be grouped as one logical element". I guess one could group semi-identical OSM elements close to each other into a "superelement" and let the correlator treat it as a single point.
I feel like it would be easier to try and guess where 1 json object ~ 2 kiosks. At least for this problem.
markalex2209 (May 27 2024 at 11:25):And I've seen your issue about manual ticking issues like "it should be this way": it would be beneficial in this case =)
But I'm not ready to commit to implementing something like this for now.
HellMap (May 27 2024 at 11:34):Is it actually possible to guess from the data?
markalex2209 (May 27 2024 at 11:38):Guess - for sure.
Get a somewhat reliable result - don't know. Will need to reference content of the csv I've sent earlier with reality.
markalex2209 (May 27 2024 at 11:39):And I'm pretty sure that at least for now two kiosks as a part of same location is something out of hand. Doubt it will be more than a handful of them.
Rihards Olups (May 27 2024 at 18:16):HellMap said:
The current Correlator© does not deal with multi-matches, but it theoretically could. I am not sure how I even would extend it to do that - this is probably soemthing extremely convoluted as now I need "range of elements that may be grouped as one logical element". I guess one could group semi-identical OSM elements close to each other into a "superelement" and let the correlator treat it as a single point.
...which could hide accidental items that are duplicated in OSM data?
HellMap (May 27 2024 at 18:20):Yeah, I suppose. This is why I describe this as a convoluted issue and it would need to be smart enough to not cause false negatives. I suppose it would be impossible to tell if someone, for example, mapped a new DP building instead of moving an existing one if it suddenly gets relocated.
Rihards Olups (May 27 2024 at 18:32):Manual exceptions (or just ignoring the known one or two cases) seem to make the most sense :)
HellMap (May 29 2024 at 11:24):Welp, kiosks can have "vidējais" machine in data, so RIP being able to distinguish kiosks even from that. May be the data is old or wrong though. For example, https://www.openstreetmap.org/way/1283887212 is a kiosk but data has 23019 - A - vidējais taromāts - Automatizēta pieņemšana. The kiosk has been there for about 2 years, so I'm assuming it was always a kiosk.
markalex2209 (May 29 2024 at 11:42):Hm. Maybe we should hide Kiosks part for now. I feel like with it's consistency, it might bring more problems than solutions.
Last updated: Jul 04 2024 at 06:43 UTC