[Geocoding] make_standard_name() and Virginia

[Sarah Hoffmann]

Hi,

On Sun, May 06, 2012 at 10:25:09AM -0400, Brian DeRocher wrote:
> Hey there,
> 
> I'm trying to understand how the nominatim works and i'm perplexed
> by this issue, where the state abbreviations get mangled.
> 
> osm=# select make_standard_name( 'VA' );
>  make_standard_name
> --------------------
>  v
> 
> This happens for 5 US states.
> 
> osm=# select st, make_standard_name(st) from brian.states where st
> <> make_standard_name(st);
>  st | make_standard_name
> ----+--------------------
>  ga | g
>  il |
>  in | 1
>  la |
>  va | v
> (5 rows)
> 
> This appears to be coming from the tokenstringreplacements.inc.

Yes, that is were it comes from. The idea behind this is to catch
common abbriviations so that e.g. 'swan road' will be found even when you
are typing 'swan rd'. Problem is that this needs to work for all
major languages and that causes quite a few clashes. Above 'il' and
'la' are common articles in French/Italian/Spanish and are therefore
deleted (module/nominatim.c line 258 ff.), the others are in
tokenstringreplacements.inc and are legal words in a language I
don't know at the moment.

I was wondering if it would help to not do these string replacements
when they constitute a full name (as opposed to words as parts
of the full names). It would need some careful consideration 
of the current uses of make_standard_name, though,
because it does not yet know the difference between words and
names.

Sarah