as of 2021, Mapillary changed their API-version from 3 to 4 and during that, moved from an alpha-numeric 22-digit ID to a completely numeric one.
Back then, the Tagging-Mailinglist discussed the topic, prompted by a user who was using the Mapillary API to display images of drinking fountains in their app and couldn’t because the display api would no longer accept the legacy ID - with the consensus that something needed to be done.
I have since then worked on creating a “todo” database and finished work on scripts and everything in late August. I will start the edit process today for the first 1000 distinct old values and see how things go.
Important note: The automated edit will only change values in the mapillary-key that either contain the old ID ONLY or an https-link that JUST goes to the same ID. If there are multiple values or specific links to a zoom within the photo, these will not be changed as of right now (There are still 182k unique values where this is the case, so plenty of work to do!)
If you have any questions about the process, there’s a Wiki page for the Edit, including a Discussion Page and a link to the Tagging-ML discussion.
Good afternoon everyone,
after running for 4.5 days, the script has gone through 45k IDs, checking the difference between Oct 2nd and 6th against taginfo (a source that I use to get a relatively precise idea of what to do in the first place) 37.6k old IDs vanished and the same amount of new IDs appeared, so it seems everything is going to plan as of right now.
Will keep y’all updated.
It’s me again …
coming upon day 8 of the transformation and 88k changesets for the script account, we now have more “new” IDs (130871) than “old” IDs (114770) in the world file of 2022-10-10T23:59Z (according to taginfo).
The next step that I’ll be working on will be changing the keys that like to the old thumbnail JPEGs.
Thank you for volunteering to tackle this boring housekeeping task.
Thanks for your encouragement
Regarding Changeset: 127352249 | OpenStreetMap : Overpass has this „qt“ parameter for „out“. You could query for 50 objects which are geographically close to each other: out 50 qt;
Not that it really matters now, it’s fyi only.
The more you know… Thank you!
did you miss me?
As of this night we’re now at 185k valid new values and 60k valid old values for the
mapillary-key in the world-file.
After getting all the changesets the script has created for comparison I found that there some changesets that do not have a change in them. There seems to be an error in detection of whether the CS is empty and shouldn’t be created in the first place.
I will seek to have these redacted from the database after everything is set and done.
At the next stage, I will need to involve the community in a discussion about what happens to the
image key that contain non-working links to mapillary thumbnails.
That’ll be another huge effort to check and I have already been criticised for not doing that in the first place.
So, for today, I’m off to some birthday celebrations.
Good news, and happy birthday!
the largest batch auf edits has been done and I have for now stopped the script because I need to rewrite the selection criteria, that’ll take a few days.
209k objects have been updated, with roughly 7.2k remaining.
I will update on the amount of old keys in the world file once TagInfo gets updated later today.
we’re slowly creeping up the ladder to a complete overhaul of the
There are now about 15k “old” keys left in the world file that. I have now restarted the process of gathering all the objects that contain these old values where a mapping to a new value exists (about 9k keys) and will need to recheck what keys have been missed the first time around (it seems that, for whatever reason, mapillary didn’t deliver the new key when I first checked them but now they do ¯\_(ツ)_/¯
Thank you all for keeping up with the process and providing valuable input!
as of today 221.8k objects have been updated, and only 13.8k values remain, that are not primarily numeric. That might also include full URLs with correct new IDs, which I will not be touching due to specific zoom/direction links.
Nevertheless, there are still more than 9,000 values that have gone “unnoticed” in the first round of object checks, which will now be looked at.
Have a wonderful week everyone!