These scripts were made for this competition.
We count number of modified crossings and road segments per user per place=city.
And with that we rank them to determine who did the most work.
You will need a PostgreSQL database. PostGIS is not needed. And Python 3.7+.
For filtering, you would need two files. One with tags: see konkurs_tags.lst
for an example. Each line has two or three space-delimited parts: object type
(n and w are supported), section name (can be repeated in multiple lines),
and a key or a tag.
If you need cities other than russian 100k+ cities, import your area to PostgreSQL
with osm2pgsql and
style for places,
and run prepare_cities.sh.
So you've got an OSM extract (user names inside are optional). First, find the relevant
hourly replication sequence number. Go here
and find an osc.gz file that's been created right after the last timestamp in the extract.
Now run init_osc.sh with five arguments:
- database name
- extract file name
- tags file name
- regions file name
- sequence number (e.g. 78123)
It will upload filtered objects to the database, and also create a table for tracking the replication sequence.
To update data to the current sequence number, run run_osc.sh. It needs three
arguments: db name, tags and regions file names. When done, check out osc_tracker
table in the database.
Run init.sh with a database name: it will create a timestamp tracking table.
The optional second parameter should be an UTC (!) timestamp in the past.
Then run run.sh once. Fix all errors, wait until it catches up and run it again
(because catching up can take 1 day for 3 days passed). And then add it to crontab.
I recommend running it once or twice an hour, for the last hour of changes is cached
at the Overpass API server.
Basically run stats.sh with a database name and an optional resulting html file name.
Alternatively, if you want a CSV, either add --csv key into the script, or manually
dump the osc_tracker / adiff_tracker to a CSV and then process it with the
generate_user_stats.py script.
Writter by Ilya Zverev, published under WTFPL (and MIT, choose what you like).