GEOS geometries are necessary for geojson_intersects and geojson_intersects_feature, and the performance improvement from not generating them are negligible.
In #227 we both thought that GEOS geometries for linestrings would be a performance hit but when I benchmarked this change, I found finding a performance difference difficult. An upper bound is a 3% increase to import time in my tests, which were done with
The alternative solution would be passing information to buildAndInsert about if the way needs an area column, geojson_intersects, or geojson_intersects_feature. This would add complexity and possible bugs for minimal gains. With the possible bugs, I feel removing the logic completely is a better solution.
There is another bug present with some geojson files which I am working at reducing, but this fixes #227.
Fixes #227
GEOS geometries are necessary for geojson_intersects and geojson_intersects_feature, and the performance improvement from not generating them are negligible.
In #227 we both thought that GEOS geometries for linestrings would be a performance hit but when I benchmarked this change, I found finding a performance difference difficult. An upper bound is a 3% increase to import time in my tests, which were done with
The alternative solution would be passing information to
buildAndInsert
about if the way needs an area column,geojson_intersects
, orgeojson_intersects_feature
. This would add complexity and possible bugs for minimal gains. With the possible bugs, I feel removing the logic completely is a better solution.There is another bug present with some geojson files which I am working at reducing, but this fixes #227.