OpenStreetMap - Determining number of nodes, ways, and relations in a PBF file?

You can use PyOsmium Python package. -Documentation-

It can be time consuming, because it scans all the data to count all elements.

import osmium

class Counter(osmium.SimpleHandler):
    def __init__(self):
        osmium.SimpleHandler.__init__(self)
        self.num_nodes = 0
        self.num_ways = 0
        self.num_relations = 0

    # this method runs for each node
    def node(self, n):
        self.num_nodes += 1

    # this method runs for each way
    def way(self, w):
        self.num_ways += 1

    # this method runs for each relation
    def relation(self, r):
        self.num_relations += 1    

counter = Counter()
counter.apply_file("path/to/pbf/file.pbf") 

print("Number of nodes : %d"    % counter.num_nodes)
print("Number of ways : %d"     % counter.num_ways)
print("Number of relations: %d" % counter.num_relations)

You can run osm2pgsql with null output, no RAM cache, and no slim cache. For a 2.5GB PBF extract of Canada, it took me about 37 seconds to count the nodes, on an NVMe SSD.

$ osm2pgsql --output null --cache 0 canada-latest.osm.pbf
osm2pgsql version 0.96.0 (64 bit id space)

WARNING: ram cache is disabled. This will likely slow down processing a lot.

Using projection SRS 3857 (Spherical Mercator)
Allocating memory for dense node cache
Allocating dense node cache in one big chunk
Allocating memory for sparse node cache
Sharing dense sparse
Node-cache: cache=0MB, maxblocks=0*65536, allocation method=3

Reading in file: canada-latest.osm.pbf
Using PBF parser.
Processing: Node(327502k 32750.3k/s) Way(18801k 895.29k/s) Relation(0 0.00/s)  parse time: 31s
Node stats: total(327502800), max(6338602157) in 10s
Way stats: total(19016827), max(676910855) in 21s
Relation stats: total(321360), max(9397389) in 0s

Going over pending ways...
        0 ways are pending

Using 12 helper-processes
Finished processing 0 ways in 0 s


Going over pending relations...
        0 relations are pending

Using 12 helper-processes
Finished processing 0 relations in 0 s

node cache: stored: 0(0.00%), storage efficiency: nan% (dense blocks: 0, sparse nodes: 0), hit rate: nan%

Osm2pgsql took 37s overall

Note that there are two numbers, the total for this file, and the max for the OSM database.

The node count can then be multiplied by 8 bytes to get a good estimate for the cache size when you import into PostgreSQL. In this case, 327502800 * 8 bytes = 2.44 GB. The osm2pgsql manual recommends adding 10 to 30% overhead on top of that cache value.