[OSM-talk] TIGER 101

[Andy Robinson]

ok, understood. But can one rough rule be used for urban and another for
back country? A small import for a few random counties would give an idea of
the range required.

Cheers

Andy

Andy Robinson
Andy_J_Robinson at blueyonder.co.uk 

>-----Original Message-----
>From: talk-bounces at openstreetmap.org [mailto:talk-
>bounces at openstreetmap.org] On Behalf Of Ben Gimpert
>Sent: 30 November 2006 5:26 PM
>To: talk at openstreetmap.org
>Subject: Re: [OSM-talk] TIGER 101
>
>Hi Andy,
>
>As I explained in my previous email, the displacement technique does not
>scale.  Yes, "one meter" might work for San Francisco but it doesn't for
>rural Montana.  (This is not conjecture.  I've already written this
>code, and it doesn't work.)
>
>		Ben
>
>
>On Thu, 30 Nov 06 @04:55pm, Andy Robinson wrote:
>> Ben,
>>
>> Thanks for the update. I think the key one to get right is the node
>reuse.
>> Grouping individual street ways parts into a single street is less
>critical
>> because it's a fairly easy task to do by hand if you need it. It's not
>> feasible to merge nodes by hand though.
>>
>> As you say the data is coarse and therefore its not really a problem to
>> consider node combination similarly. However if I look at the San
>Francisco
>> data I don't see node displacement at junctions exceeding 1 metre (in
>fact
>> significantly smaller if not the same lat and lon) so I'm guessing that
>many
>> of the street junctions do have the same lat/lon or at least very close
>to
>> it. Its very unlikely that there is a problem combining points that are
>> within that sort of radius anyway.
>>
>> Cheers
>>
>> Andy
>>
>> Andy Robinson
>> Andy_J_Robinson at blueyonder.co.uk
>>
>> >-----Original Message-----
>> >From: talk-bounces at openstreetmap.org [mailto:talk-
>> >bounces at openstreetmap.org] On Behalf Of Ben Gimpert
>> >Sent: 30 November 2006 4:36 PM
>> >To: talk at openstreetmap.org
>> >Subject: Re: [OSM-talk] TIGER 101
>> >
>> >Hi Andy & Steve,
>> >
>> >Right, nodes are not being reused.  This is just because of how they are
>> >represrented they are represented in the TIGER data itself.  I've posted
>> >about this before -- how "5th Avenue" in Manhattan is actually N
>> >separate record chains in TIGER, though obviously it's a
>> >physically-contiguously street.  And that this occurs also at the node
>> >level, where there is no level of abstraction in the TIGER dataset for
>> >shared nodes.
>> >
>> >Take a look at:
>> >
>> >	http://svn.openstreetmap.org/utils/tiger_import/tiger/tiger.rb
>> >
>> >...and note how the TIGER files present (only) raw lat/long, at every
>> >scale (node, street, point-of-interest).
>> >
>> >We have to remember that the TIGER data is just a *very* rough first
>> >step to a usable streetmap of America.  I'm sure companies like MapQuest
>> >have had to spend enormous money and/or effort to clean things up for
>> >their driving directions.
>> >
>> >I wrote some code to intelligently try to merge roads with the same
>> >names that share (roughly) an end node-or-two.  This code didn't scale
>> >well outside of Manhattan, since FIPS counties can be very strangely
>> >shaped and sized.  (See Steve's 'blog post on Gerrymandering for a
>> >similar topic...  Heh.)
>> >
>> >As for re-using nodes, we face a similar problem of scaling:  At what
>> >lat/long precision do we consider two points the "same"?  (Say 0.00005
>> >of a degree, or what?)  Again, answering this question is hard across
>> >the entire (HUGE) country.
>> >
>> >We might define a formula based upon the smallest rectangle that can
>> >cover a county.  Say, (maxCountyLongitude - minCountyLongitude) / 10^5,
>> >but this, umm, doesn't work.  (I tried already.)
>> >
>> >Let me know via email if someone else wants to take a crack at writing
>> >code to "merge nodes" (and streets) in the TIGER data.  I myself won't
>> >be able to write any more code for OSM since I'm bogged down in other
>> >responsibilities.  Especially given OSM's wiki nature, I feel like a
>> >routing system will have to have some intelligence about assuming two
>> >"nearby" nodes are really the same intersection / bend / whatever.
>> >
>> >Hope this make sense!
>> >
>> >		Cheers,
>> >		Ben
>> >
>> >
>> >On Thu, 30 Nov 06 @03:20pm, Andy Robinson wrote:
>> >> Ben,
>> >>
>> >> Looking at the San Francisco data newly imported it still appears that
>> >ways
>> >> are being added without connection to adjacent ways, ie there is no
>> >sharing
>> >> of common nodes. See JOSM screen dump where I have selected and
>dragged a
>> >> way. It should have tugged the adjacent ways too.
>> >>
>> >> http://ajr.hopto.org/osm/tiger-nodes.png
>> >>
>> >> Cheers
>> >>
>> >> Andy
>> >>
>> >> Andy Robinson
>> >> Andy_J_Robinson at blueyonder.co.uk
>> >>
>> >> >-----Original Message-----
>> >> >From: talk-bounces at openstreetmap.org [mailto:talk-
>> >> >bounces at openstreetmap.org] On Behalf Of Ben Gimpert
>> >> >Sent: 30 November 2006 2:53 PM
>> >> >To: talk at openstreetmap.org
>> >> >Subject: [OSM-talk] TIGER 101
>> >> >
>> >> >Hi OSM,
>> >> >
>> >> >The TIGER -> OSM import is (again) kicked off and going.
>> >> >
>> >> >Since there is now disaster recovery logic atop a MySQL tracking
>> >> >database, a proper status report is possible.  Unless there are any
>> >> >objections, I intend to commit the status report to the OSM SVN
>> >> >repository every night at 3am:
>> >> >
>> >> >		http://svn.openstreetmap.org/utils/tiger_import/status
>> >> >
>> >> >Dig the tiny-but-increasing numbers for CAlifornia, around our first
>> >> >prioritized counties (for Mikel).
>> >> >
>> >> >		Cheers,
>> >> >		Ben
>> >> >
>> >> >
>> >> >_______________________________________________
>> >> >talk mailing list
>> >> >talk at openstreetmap.org
>> >> >http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>> >>
>> >
>> >_______________________________________________
>> >talk mailing list
>> >talk at openstreetmap.org
>> >http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk
>>
>
>_______________________________________________
>talk mailing list
>talk at openstreetmap.org
>http://lists.openstreetmap.org/cgi-bin/mailman/listinfo/talk