[OSM-dev] Anyone with a speedy gazetteer

[Milo van der Linden]

Tom Hughes wrote:
> Milo van der Linden wrote:
>
>> I strongly suggest that you read the postgresql text search[1] 
>> chapter in depth. You will find that a lot of textual and 
>> multilingual confusions can be solved with that function set. the 
>> name "text search" is by far too simple for what it covers...
>
> I have read it, and it's not at all clear it can be made to work right.
>
> The basic problem is that it assumes that all text in a given field is 
> in the same language - you can give it a set of language specific 
> rules to use when parsing a the record data and when parsing the query 
> but you can't vary that on a record by record basis.
That is indeed a thing. It also shows that it might be a good option to 
extend the column remapping that for instance osm2pgsql uses to create 
new columns per language.

In the end, a well structured gazetteer that works on a steady system of 
internationalization
http://wiki.openstreetmap.org/wiki/Key:name

might give people a stimulance to not only create 
name:yourlanguageisocode tags, but actually see them in action..
>
> Though even if you could you wouldn't know what language the query was 
> in anyway, so you wouldn't know how to parse it...
Being it from a browser: You might know the browser language, or give 
the user a language selection option
And as a webservice: set a directive to a language iso code in the post 
or get variables.
>
> I have actually been playing with using Postgres for this, using 
> osm2pgsql to load planet data into a gazetteer with Postgres text 
> search for name matching and a spatial index for location matching. I 
> wound up sticking to using the "simple" text parsing ruleset in the end.
Could you be more specific on the "Why" for that? Because the two 
sentences above sound a bit like:
"I walked by the restaurant yesterday but ended up eating with my parents."

;-)
>
> Tom
>