osm2pgsql Project History

2006

The osm2pgsql project was started in 2006 by Jon Burgess (@jburgess777) and written in C. This makes the project nearly as old as the OpenStreetMap project itself.

2007

In 2007 OSM moved to API 0.5 adding the relation data type. Shortly thereafter osm2pgsql got support for multipolygon relations, allowing polygons to have holes for the first time. Also in 2007 osm2pgsql added support for multiple projections, beyond the WGS84 lat/lon projection and Web Mercator used before.

2008

Support for reading OSM change files was added to osm2pgsql in 2008. From 2009 on the main OSM map was re-generated based on minutely changes. This was a major achievement at a time when most maps on the Internet were updated only a few times per year. The toolchain importing the data with osm2pgsql into a PostgreSQL/PostGIS database and then rendering tiles with Mapnik proves up to the task.

2009

Brian Quinion (@twain47) created the geocoder Nominatim which to this day powers the search on openstreetmap.org and many other map sites. Nominatim uses osm2pgsql as a basis, the special gazetteer output was added for that.

2011

The number of nodes in OpenStreetMap is growing rapidly passing the one billion mark. Osm2pgsql is changed to use 64 bit IDs to allow for the expected growth beyond two billion nodes.

Kai Krueger (@apmon) takes over as maintainer from Jon Burgess.

2012

The flat file persistent node cache for slim mode is introduced, drastically reducing disk usage and improving import and update performance a lot.

2013

The ability to have complex tag transformation on import through Lua scripting is introduced.

2014

Matt Amos (@zerebubuth) together with Kevin Kreiser (@kevinkreiser) start the move from C to C++, a process which will take many years to complete.

Sarah Hoffmann (@lonvia) and Paul Norman (@pnorman) take over as maintainers.

2015

osm2pgsql starts using the libosmium library for parsing OSM data. First parsing of XML and PBF files was switched, later on also the o5m parsing code.

2017

Geometry generation is also switched to libosmium and GEOS is not used any more. This makes especially the multipolygon generation more robust and improves performance.

2018

Support for LuaJIT is added, making imports with complex tag transformations much faster.

2019

With the 1.0 release, osm2pgsql drops processing of so called old-style multipolygons. This allows for a major overhaul of the processing pipeline and much improved import speed.

2020

The new flex output is introduced by Jochen Topf (@joto). It allows a much more flexible configuration of the output tables. The flex output is improved step by step over the next years.

Also in 2020 osm2pgsql finally gets its own website.

2021

The OpenStreetMap project in general and osm2pgsql more specifically is used more and more. The osm2pgsql repository passes 1000 stars on Github.

2022

Osm2pgsql has always been a tool to import OSM into a database. Many projects have built upon that and done further processing of the data once it is in the database, mostly using SQL scripts. Often this is done to generalize the rather detailed OSM data for use in smaller zoom levels. In 2022 support for generalization is added to osm2pgsql with the (experimental) osm2pgsql-gen command. This work is funded by the Prototype Fund.

2023

A new database format for the middle is introduced replacing the somewhat homegrown format with simpler to use JSONB data. This makes it possible for the first time to store all OSM data including all tags, and all attributes and member information for all objects in the database.

2024

Nominatim switches to the flex output, which makes the gazetteer output obsolete. It is removed along with a lot of other obsolete functionality and version 2.0 is released.