go-osmpbf-filter is a program to filter OpenStreetMap PBF format data files. The filter performs six passes on the input PBF filter, and then writes a new output PBF file. The six passes are:

1. Find the OSMHeaders data block and ensure that we support reading this file.
2. Find ways that match a given tag criteria (eg. leisure=golf_course), and extract their node references.
3. Calculate bounding boxes for each way given the bounding boxes found in (2).
4. Find all the nodes within the bounding boxes from (3).
5. Find all the ways that reference nodes found in (4).
6. Find all the nodes that are referenced by the ways found in (5).

The nodes and ways collected in passes (3), (4) and (5) are then output into a new PBF format data file.

go-osmpbf-filter is written in Go. It is highly concurrent, so you can expect it to use up all your CPU power. go-osmpbf-filter can filter out 1MB of golf course areas from a 1GB PBF file in about 9 minutes, using 789% of a Core i7 processor.

Building go-osmpbf-filter is easy:

cd go-osmpbf-filter
GOPATH=`pwd` go build

Filter a pbf file with the default settings (leisure=golf_course):

go-osmpbf-filter -i alberta.osm.pbf -o alberta-filtered.osm.pbf

Filter searching for a specified tag, like sport=baseball:

go-osmpbf-filter -i japan.osm.pbf -o japan-baseball.osm.pbf -t sport -v baseball

-i

Input file

-o

Output file

-t

Filter tag key

-v

Filter tag value

--high-memory

Cache every decompressed block in memory. This can cause a 25% performance improvement in filtering, but is only recommended for small files. For a ~100MB PBF file, this will increase memory usage by about 100MB; for a 1GB PBF file, memory usage will jump to 8GB or more.