Upgrade geocoder to use the new tiger 2010 files

[robe]

What steps will reproduce the problem? 1. the tiger geocoder relies on the 2006 and below which tiger files which were in the proprietary Tiger format. In 2007 census switched to using ESRI shape file format and corrected a lot of data. 2. Most people I know want to use the new stuff, but the structure has changed quite a bit. E.g. the address file is now a separate table called addr and can have a 1 to many with the edges table (what used to be complete chain)

3. This will all take a bit of work. Not sure if Steve Frost has started this work yet or not.

[sfr...@…]

I've already updated the geocoder itself to use the TIGER 2007/2008 shapefile data. I'm working with someone else on improving it further as well as improving the load scripts to build the whole system.

If you have git, you can pull the latest from here:  http://www.snowman.net/git/tiger_geocoder

Once it's working well with the import scripts, the 2008 data, etc, I'll be committing it back to the PostGIS SVN. I'm on the hook to complete this work before PgCon? 2009 (in May).

[pramsey]

Ping. Moving this to next release soon...

[robe]

I think we can get this into PostGIS 2.0.0. Steve let me know if you are still interested in working on this. If so I'd like it committed to trunk. If not I'll take what you have so far and merge in my corrections and commit to trunk.

[robe]

We've ammended the tiger geocoder to work with tiger 2010 and as far as we can tell, it works fine. At least for Massachusetts data anyway. We'll be stress testing it more with a consulting project we are currently working on and tweaking it as we go along, but we consider it more or less done. It is now also documented as part of the official PostGIS docs.  http://www.postgis.org/documentation/manual-svn/Extras.html#Tiger_Geocoder

We still haven't gotten around to testing the loader script on linux if anyone else wants to take a stab at that. Jan had tested it on Ubuntu for our book, and we incorporated his corrections as best we could.

We may add a reverse geocoder function to it before we are done since we'll need to build that anyway for work we are currently doing, but will add that as a separate task item.