Change area decisions at import time #4032
Conversation
Tag v4.21.1
Adding these two features to the local polygon values list in openstreetmap-carto.lua will allow these features to be imported as polygons so that they will render properly
Import railway=station as a polygon
Add tourism=yes as a linestring exception
Import "allotments" as a polygon key
Tag v4.22.0
…ings (#3973) * Add power=cable, man_made=cutline and man_made=pipeline as a linestring only features in the lua file - These features will now be imported only as linestrings, not as polygons, when tagged on a closed way.
Import closed ways taggedwith aerialway=station as polygons for proper rendering
Import aerialway=station as a polygon feature
Add the key `craft=` to the list of keys which are imported as polygons when mapped as closed ways
Tag v4.25.0
Add waterway=tidal_channel to the list of linestring exceptions in openstreetmap-carto.lua so that closed ways with this tag will only be imported as a linestring
Import the key `craft=` as a polygon
A tag like building=no shouldn't cause the object to be considered an area.
Import natural=earth_bank as a linestring
Import closed ways with the key `club=` as polygons
Add golf key to list of polygon keys in openstreetmap-carto.lua so that closed ways with this key will be imported as polygons, except for golf=hole, golf=path and golf=cartpath which are always mapped as lines
Don't form areas from tags like building=no
Add boundary= aboriginal_lands, national_park, and protected_area to polygon keys
Pull request #3727 introduced the new condition osm_id > 0 but did not update the condition for the indexes inteded to be used by the SQL queries of these layers.
|
Originally I was wondering if this needed a benchmarking run, but looking at the diff I don't see anything with performance implications, either at import time or at run time. |
There was a problem hiding this comment.
I've tested these changes. While my system does not have sufficient RAM to import the whole planet, I tried importing the 153 MB Oregon extract (153 MB pbf file "oregon.osm.pbf", downloaded December 16th), 5 times with the current master branch and 5 times with schema_changes.
With schema_changes, the import took 685s, 647s, 669s, 684s, and 671s:
Range 647 to 685
Mean 671 (and median 671)
On master, the import took 665s, 576s, 554s, 607s, and 644s:
Range 554 to 665
Mean 609, median 607
So the range of times overlaps, but on average the new schema is about 10% slower, which is reasonable and usually will not be noticed.
|
I wasn't expecting that much of a performance difference, so I'm running a benchmark myself |
|
I don't find a noticeable performance difference on Geofabrik's Canada extract with |
|
That's great. Is there documentation about how to use these options ( |
I don't use docker, so don't know. |
Fixes #3611