[postgis-devel] postgis_tiger geocoding extension : France & temporal geocoding

[Regina Obe]

> On Thu, Jun 16, 2016 at 6:10 AM, Rémi Cura <remi.cura at gmail.com> wrote:

   >  Hey dear list,
  >  I'm currently considering extending postgis geocoder.

  >  I have two needs:
 >    - to work with French data
 >    - to perform temporal geocoding (you give an addresses and a date, you
get geographic position).

>    For the moment I'm trying to understand how the geocoder works with USA
data.

>    I imported Massachusetts for a test, and everything seems okay except
that my tiger.state table is empty, which prevents geocoding.

>    Notwithstanding the import step, and the adress normaliser, it seems to
boil down to a large query in the function tiger.geocode_address().


>    I would be delighted to have some pointers on the design of the
postgis_tiger_geocoder extension.
>    In perticular, I don't understand the address standardizer business at
all (yeah, I found some doc, no, I don't understand it well)


>    Many thanks for help,
>    Cheers,
>    Rémi-C

Just adding to what Steve has already said. 

Don't try to use the postgis tiger geocoder as a model of how to set up a
geocoder.  
The structure of it is very tied to how the US Census Tiger data is setup.
The irony is because it is setup like US Census, I think people try to use
it for more than a geocoder and I get bug reports on fields I ignored
because they really aren't relevant as far as geocoding is concerned or not
relevant to me (in other moneywise ways).  I can't blame them since I do use
it myself on other projects for more than just geocoding.  E.g. census tract
work or just joining with the states, counties etc when I need a county
list.

If I had started from scratch instead of inheriting the code this is not how
I would do it.  
But without enough traction of people willing to devote time/fund  a more
generic geocoder, the more globally useful solution will not go anywhere.


Thanks,
Regina