revise loader to handle 2011 tiger data

[robe]

Tiger data 2011 recently came out.

 ftp://ftp2.census.gov/geo/tiger/TIGER2011

Might be trivial to upgrade the loader to handle it especially if they have state by state structure available similar to 2010

[robe]

I think as Steve Frost mentioned, the 2011 data has some new tables like addrfeat which look very useful and would simplify some queries. To use that would mean breaking away from current 2010 structure a bit and revising the queries accordingly.

Not something we can squeeze into a micro update. We'll start working on this but won't commit anything until 2.1 has become trunk or perhaps we can get our feet wet with git :).

[robe]

Okay I have this pretty much working. Just fine tuning the upgrade scripts. As it stands right now, its compatible with tiger2010 structure. I haven't changed any of the structures yet so you can have a mix of 2010 and 2011 data and things will work fine. I did add logic in the loader to load up the new addrfeat (which I'm in the middle of testing). I haven't done anything with this and actually not convinced it will help much since the way it is structured I predict it will make less efficient use of indexes (although it will save a join). Only testing will see.

Anyrate -- since I had to add in some helper functions (e.g. one that is a bit more respective of user search path during install) and addition of a nation load function (since the county and state are in 2011 just packaged as whole nation instead of individual state), I don't feel compfortable releasing this as a micro update. Will probably get shot down by folks since I can't introduce new functions in a micro. If you want to give it a try -- I have it in spike

http://trac.osgeo.org/postgis/browser/spike/robe/tiger_geocoder/tiger_2011

The only caveat -- which I was going to create a helper function do do is:

You need to drop all the state county and state tables -- e.g. ma_county, ma_state

And then right after, generate download script to download nation data:

 SELECT tiger.loader_generate_nation_script('windows');

or if on unix:

 SELECT tiger.loader_generate_nation_script('sh');

I've been testing only on windows. Won't be able to test on Linux for another week or so. However most of the changes I made were to the psql and wget calls which are the same regardless of if you are on windows or unix so I don't forsee any issues with the sh version working out of the box and if there are, they are probably minor.

The new routine is also less greedy about downloads. It won't download files we don't load up and also will delete each set of table files after use (instead of extracting a whole states worth in temp) so should minimize the issue people had with too many files generated on Mac.

[robe]

This is working for me and now part of 2.1.0. I haven't changed the underlying query logic so will work fine on 2.0.0 and 1.5 installs. Eventually I may take advantage of new changes in 2.0.0, but not at the moment. Looking forward to having testers. I still haven't tested on Linux.