| 12:29:03 | bcrosby: | hidyho |
| 14:23:48 | darkblue_B: | bcrosby: ?? |
| 14:46:15 | landonf: | Anyone do a lot of geocoding and have a suggestion for free/cheap software/service that is very accepting of inconsistently formatted address data? = ) |
| 15:05:58 | darkblue_B: | hmmm.. what is "a lot" ?? |
| 15:11:04 | landonf: | Open to interpretation, but I've got to pipe this first batch of about 30,000 addresses through, hopefully normalize the addresses and get decent coordinates. |
| 15:12:09 | landonf: | I'm using the tiger geocoder implementation right now, but working with something written entirely in PL/PgSQL is pretty hard, both in terms of profiling/improving performance, as well as development to improve handling of parsing issues. |
| 15:13:42 | landonf: | It's taking from 150 ms to 1 second to geocode an address, and pulling 30 MB/s sustained from disk while doing a batch geocode of addresses in a single city, and I'm having a hard time figuring out why : ) |
| 15:14:23 | darkblue_B: | I believe google has changed their geocoding license such hat it is ony rate limited.. so if you have a few days, you can do all of them without a problem |
| 15:14:57 | darkblue_B: | yahoo has changed their license in turn, to substantially up the daily limit.. I think in the 25,0000 range, but I dont klnow |
| 15:15:24 | darkblue_B: | the bg problem for them is sites that dont save their results I thnk |
| 15:15:54 | darkblue_B: | then there is the OSS one, I thnk Rich Gibson is the contact for that. .though he may be getting ready to go to Burning Man now.. I |
| 15:16:16 | landonf: | Oh really? Only rate limited? That would be handy. |
| 15:16:28 | darkblue_B: | I assume that the large player have done extensive tweaks to their system |
| 15:16:41 | darkblue_B: | far morew than any small outfit would e likely to complete |
| 15:18:18 | landonf: | Yeah, especially for stuff that requires local knowledge, like "1 Ferry Building, San Francisco". It's a building, not a street, and the street is Embarcadero, but it has no street address. Google will map it fine. |
| 15:19:40 | iggy: | I wrote a script that was going through a list of 50k addresses I had and geocoding against geocoder.us, mapquest, yahoo, google in series |
| 15:21:57 | landonf: | "You may use the Maps API geocoder solely to obtain and display points on map images provided through the Service." <-- google terms of service :( |
| 15:23:02 | landonf: | Yeah, Yahoo has a similar restriction |
| 15:23:16 | landonf: | Was looking at this too: http://sourceforge.net/projects/jgeocoder |
| 15:23:17 | sigq: | Title: SourceForge.net: jgeocoder - Free Java Geocoder ( at sourceforge.net ) |
| 15:26:39 | iggy: | www.pagcgeo.org <--- see this |
| 15:30:03 | darkblue_B: | oh yeah.. there is a small python project that implements a geralized geocoding interface |
| 15:30:52 | darkblue_B: | I looked inside.. not much code.. good examples of hwo call quite a few services |
| 15:31:49 | darkblue_B: | geopy |
| 15:34:07 | darkblue_B: | geopy uses a pythno package called BeautifulSoup to work with malformed address input |
| 15:34:21 | darkblue_B: | .. not sure of how extensive that is.. but its there |
| 16:20:28 | jlivni: | i have found virtual earth has the best geocoding, but that was ~6 months ago and i think now google has modified their results a bit |
| 16:21:02 | jlivni: | you can purchase some commericial mapquest geocoding if you need, more easily than from the other commercial providers, again as far as i can tell |
| 16:21:40 | jlivni: | if you dont need crazy accuracy, what i usually do is send everything through my own tiger geocoder, and then the things that fail i put through a commercial service |
| 16:22:28 | jlivni: | on a cheap ~3 year old dell that does lots of other stuff, i get ~5/second on my geocoder which is basically just following the instructions from geocoder.us |
| 16:23:02 | jlivni: | er, i rememberd wrong. make that ~9/sec now that i just tested again ... |
| 16:47:59 | landonf: | Thanks |
| 16:54:27 | jlivni_: | not sure if anyone said anything while i was gone .. but if so, i didn't see it |
| 16:58:50 | landonf: | Just 'thanks' = ) |