Partitions always include 0, but no partition of value '0' found in country_name?

[AyushDharDubey]

Hi all,

I was going through the codebase and came across this function in sql_preprocessor:

def _get_partitions(conn: Connection) -> Set[int]:
    """ Get the set of partitions currently in use.
    """
    with conn.cursor() as cur:
        cur.execute('SELECT DISTINCT partition FROM country_name')
        partitions = set([0])
        for row in cur:
            partitions.add(row[0])

    return partitions

One thing that stood out to me is that the value 0 is always inserted into the partitions set upfront, before considering any values returned from country_name.partition.

I assume this is intentional, but I’m not yet familiar enough with the partitioning logic to understand why partition 0 is treated as special here. Moreover, what sticks out is, in my local setup, I don't have the value 0 in my country_name.partition.

Pardon me if this ought to be something obvious. I’m still getting oriented to this project and wanted to affirm my understanding before raising an issue blindly.

Thanks!

[mtmail]

placex.partition 0 is where places not assigned to any country end up. In a planet installation that's about 19.000 records out of 350 million. If you only import one country you might not see any.

Examples are offshore oil platforms, buoys, ship wrecks https://www.openstreetmap.org/node/1973965208

I assume the same partition logic was reused on the country_name table. On a planet installation I also see no records for country_name.partition 0.

[AyushDharDubey]

Thanks, that explains.

On a separate note, shouldn't _get_partitions return an empty set if the partitions do not exist (as when frozen)??

[lonvia]

Partitions always exist. There is a column in placex where you can tell for every entry in which partition they are. The information is not of much use, admittedly, once the database is frozen but it's still there.