How to estimate disk space requirements for flex output in osm2pgsql 2.0.0?

[wcedmisten]

Hello!

I'm running osm2pgsql with a flex output for the planet file (77GB PBF). I'm running this on a VPS with a mounted volume, where I pay by the GB. I'm wondering how much space PostgreSQL needs to run:

osm2pgsql --create --verbose -U postgres -H localhost --slim --output=flex -S /osm2pgsql.lua /extract.osm.pbf

The flex output I'm using is meant to extract all restaurants (nodes, ways, or relations) from the planet.

I've done something similar by running osmium tags-filter with nwr/amenity=restaurant and the output is only ~300MB. But now that I'm running osm2pgsql directly on the planet file, I'm trying to figure out how much disk space is needed to run osm2pgsql directly.

Appendix

/osm2pgsql.lua

local restaurants = osm2pgsql.define_table({
    name = 'restaurants',
    ids = { type = 'any', type_column = 'osm_type', id_column = 'osm_id' },
    columns = {
        { column = 'name', type = 'text' },
        { column = 'cuisine', type = 'text' },
        { column = 'amenity', type = 'text' },
        { column = 'tags', type = 'jsonb' },
        { column = 'geom', type = 'point', projection = '4326' }
}})

function process_poi(object, geom)
    local r = {
        name = object.tags.name,
	    cuisine = object.tags.cuisine,
        amenity = object.tags.amenity,

	    tags = object.tags,
        geom = geom
    }

    if object.tags.amenity == "restaurant" then
    	restaurants:insert(r)
    end
end

function osm2pgsql.process_node(object)
    process_poi(object, object:as_point())
end

function osm2pgsql.process_way(object)
    if object.is_closed and object.tags.building then
        process_poi(object, object:as_polygon():centroid())
    end
end

[joto]

There are some numbers here in the manual. https://osm2pgsql.org/doc/manual.html#sizing

[pnorman]

For the flex output, it depends on the tables. The upper end of the values in the manual would be for something like OpenStreetMap Carto, while something like what you have would be very small.

If you've run it on the planet pre-filtered with osmium then the flex tables will take the same amount of space when run directly on the planet. If you're running in slim mode you will need space for the slim tables and flat nodes. Compared to those the restaurants will be a rounding error.