[OSM-talk] TIGER / Mapnik suggestion

[Dave Hansen]

On Wed, 2007-09-05 at 23:46 -0400, Andrew MacKinnon wrote:
> On 9/5/07, Dave Hansen <dave at sr71.net> wrote:
> > I'd be happy to implement any sane suggestions for fixing this.  Any
> > ideas?
> 
> Perhaps "St" could be converted like this:
> 
> St at end of name = Street
> St at beginning of name = Saint
> 
> Most of the other abbreviations are unambiguous in the US. We should
> also add a tag like "tiger:original_name" which indicates the original
> name of the way in the TIGER data, so that we can correct errors
> caused by this conversion.

Have you actually gone and looked at the TIGER data?  Take a look at the
tags:

        <tag k="name" v="SE View Meadows Ln"/>
        <tag k="tiger:name_type" v="Ln"/>
        <tag k="tiger:name_base" v="View Meadows"/>
        <tag k="tiger:name_direction_prefix" v="SE"/>

The original name is just the three (or more) TIGER parts put together.
Go to down editing the "name" because it isn't really part of TIGER to
begin with.

Frankly, this is a much bigger problem than TIGER itself.  A nice,
robust, solution would do something along the lines of looking at the
planet dump or scanning the database and removing these abbreviations as
they pop up.  This isn't just a problem with TIGER, it is much more
widespread.  

If you want to send me patches to the TIGER converter program to do
this, and re-run the generation of the ~6GB of compressed .osm files
that it spit out of the entire country, I'll be happy to upload them.
Otherwise, I think this is a pretty minor nit.

Here's the source to the converter program:

	https://www.sr71.net/~dave/osm/tiger/tiger-to-osm-v0.7.tar.gz

It takes about 1-2 weeks of constant churning on my laptop with 2GB of
RAM to spit out the entire country.

-- Dave