ff6207d200
multipolygons that are not closed (topological) can be handled by using GEOS LineMerge function. It's expected to be slower than the node ID based merger. Maybe it's feasible to use LineMerge as a fallback. |
||
---|---|---|
cache | ||
cmd | ||
config | ||
database | ||
diff | ||
element | ||
expire | ||
geom | ||
import_ | ||
logging | ||
mapping | ||
parser | ||
proj | ||
reader | ||
stats | ||
test | ||
util | ||
writer | ||
.gitignore | ||
LICENSE | ||
Makefile | ||
README.md | ||
example-mapping.json | ||
imposm3.go |
README.md
Imposm 3
Imposm is an importer for OpenStreetMap data. It reads PBF files and imports the data into PostgreSQL/PostGIS. It can also update the DB from diff files.
It is designed to create databases that are optimized for rendering (i.e. generating tiles or for WMS services).
Imposm 3 is written in Go and it is a complete rewrite of the previous Python implementation. Configurations/mappings and cache files are not compatible with Imposm 2, but they share a similar architecture.
It is released as open source under the Apache License 2.0.
The development of Imposm 3 was sponsored by Omniscale and development will continue as resources permit. Please get in touch if you need commercial support or if you need specific features.
Features
- High-performance
- Diff support
- Custom database schemas
- Generalized geometries
In detail
-
High performance: Parallel from the ground up. It distributes parsing and processing to all available CPU cores.
-
Custom database schemas: Creates tables for different data types. This allows easier styling and better performance for rendering in WMS or tile services.
-
Unify values: For example, the boolean values
1
,on
,true
andyes
all becomeTRUE
. -
Filter by tags and values: Only import data you are going to render/use.
-
Efficient nodes cache: It is necessary to store all nodes to build ways and relations. Imposm uses a file-based key-value database to cache this data.
-
Generalized tables: Automatically creates tables with lower spatial resolutions, perfect for rendering large road networks in low resolutions.
-
Limit to polygons: Limit imported geometries to polygons from Shapefiles or GeoJSON, for city/state/country imports.
-
Easy deployment: Single binary with only runtime dependencies to common libs (GEOS, SQLite and LevelDB)
-
Support for table namespace (PostgreSQL schema)
Performance
Imposm 3 is much faster than Imposm 2 and osm2pgsql:
- Makes full use of all available CPU cores
- Bulk inserts into PostgreSQL with
COPY FROM
- Efficient intermediate cache for reduced IO load during ways and relations building
Some import times from a Hetzner EX 4S server (Intel i7-2600 CPU @ 3.40GHz, 32GB RAM and 2TB software RAID1 (2x2TB 7200rpm SATA disks)) for imports of a 20.5GB planet PBF (2013-06-14) with generalized tables:
- 6:30h in normal-mode
- 13h in diff-mode
osm2pgsql required between 2-8 days in a similar benchmark (slide 7) with a smaller planet PBF file (~15GB).
Benchmarks with SSD are TBD.
Import of Europe 11GB PBF with generalized tables:
- 2:20h in normal-mode
Current status
Imposm 3 is in alpha stadium and there is no official release yet. The import itself is working however and it was already used for production databases.
Missing
Compared to Imposm 2:
- Documentation
- Support for other projections than EPSG:3857
- Import of XML files
Other missing features:
- Updating generalized tables in diff-mode
- Automatic download of diff files
- Tile expire list for re-rendering updated areas
- Background mode for diff-import (update DB in background)
- Diff import into custom PG schemas
- Improve parallelization of diff import
Installation
Binary
There are no official releases, but you find development builds at http://imposm.org/static/rel/.
These builds are for x86 64bit Linux and require no further depedecies. Download, untar and start imposm3
.
(Note: These binaries require glibc >= 2.15 at the moment.
Ubuntu 12.04 is recent enough, Debian 7 not. Future binary releases will work on older versions as well.)
Source
There are some dependencies:
Compiler
You need Go >=1.1.
C/C++ libraries
Other dependencies are libleveldb, libgeos and libsqlite3. Imposm 3 was tested with recent versions of these libraries, but you might succeed with older versions. GEOS >=3.2 is recommended, since it became much more robust when handling invalid geometries. For best performance use HyperLevelDB as an in-place replacement for libleveldb.
Go libraries
Imposm3 uses the following libraries. go get
will fetch these:
- https://github.com/jmhodges/levigo
- https://github.com/mattn/go-sqlite3
- https://code.google.com/p/goprotobuf/proto
- https://code.google.com/p/goprotobuf/protoc-gen-go
- https://github.com/lib/pq
For now you need to upgrade lib/pq to the bulk branch:
cd $GOPATH/src/github.com/lib/pq
git remote add olt https://github.com/olt/libpq.git
git fetch olt
git checkout olt/bulk
Other
Fetching Imposm and the Go libraries requires mercurial and git.
Compile
Create a new Go workspace:
mkdir imposm
cd imposm
export GOPATH=`pwd`
Get Imposm 3 and all dependencies:
git clone https://github.com/omniscale/imposm3 src/imposm3
go get imposm3
go install imposm3
Done. You should now have an imposm3 binary in $GOPATH/bin
.
Go compiles to static binaries and so Imposm 3 has no runtime dependencies to Go.
Just copy the imposm3
binary to your server for deployment. The C/C++ libraries listed above are still required though.
Usage
imposm3
has multiple subcommands. Use imposm3 import
for basic imports.
For a simple import:
imposm3 import -connection postgis://user:password@host/database \
-mapping mapping.json -read /path/to/osm.pbf -write
You need a JSON file with the target database mapping. See example-mapping.json
to get an idea what is possible with the mapping.
Imposm creates all new tables inside the import
table schema. So you'll have import.osm_roads
etc. You can change the tables to the public
schema:
imposm3 import -connection postgis://user:passwd@host/database \
-mapping mapping.json -deployproduction
You can write some options into a JSON configuration file:
{
"cachedir": "/var/local/imposm3",
"mapping": "mapping.json",
"connection": "postgis://user:password@localhost:port/database"
}
To use that config:
imposm3 import -config config.json [args...]
For more options see:
imposm3 import -help
Sorry, that's all documentation for the moment.
Support
There is a mailing list at Google Groups for all questions. You can subscribe by sending an email to: imposm+subscribe@googlegroups.com
For commercial support contact Omniscale.
Development
The source code is available at: https://github.com/omniscale/imposm3/
You can report any issues at: https://github.com/omniscale/imposm3/issues
Test
Unit tests
go test imposm3/...
System tests
There is a system test that imports and updates OSM data and verifies the database content.
This test is written in Python and requires nose
, shapely
and psycopg2
. You also need osmosis
to create test PBF files.
There is a Makefile that (re)builds imposm3
and creates all test files if necessary and then runs the test itself.
make test
WARNING: It uses your local PostgeSQL database (import
schema), if you have one. Change the database with the standard PGXXX environment variables.