On Mon, May 01, 2006 at 09:00:31PM -0400, Kent Hargesheimer wrote: > Wow Bill, you got that built fast! > > You are right, the core C++ engine is very fast...Ned writes EXTREMELY > efficient code. It will be interesting to see how this move by SRC impacts > the geocoder market. > > Please do keep us posted on access to the code and any future developments.
Well, I was inspired by having several hundred thousand of mostly rural addresses to geocode. I'm really cuurious to see how fast a native C++ tool that uses this engine can be once somebody gets such a tool available for the rest of us who don't know C++ yet. Even in MapBasic, it runs pretty well. The MapBasic code I wrote is hobbled by its interpretive nature, a slow "do" loop and the fact that I used passed variables rather than global ones to keep the code "cleaner" and easier to read. I am planning to release the source code, but that needs to wait for a change in the DLL so that it doesn't expose more than it should. As to the impact of this open source geocoder, I think the heavy users of products like MapInfo MapMarker and Group 1's geocoder are still going to pay the high cost to get the detail and accuracy these products provide. Not that Explorer geocoder is inaccurate; it's just that you can only go so far with TIGER-based data. There are still many rural roads in TIGER that don't have address ranges on them and there are no ZIP+4 boundaries in TIGER. When I geocoded 195,000 (mostly rural) addresses using Explorer geocoder, about 25% ended up being located at the ZIP code centroids because the streets they were on were missing address ranges. (But getting 146,250 rural address located at their street address wasn't too shabby.) But for the vast majority of small businesses, governments and experimenters, this tool is a windfall. There's a lot that it can provide for very little cost right now, and as people start working with the code, support will be added for non-US address information and better data sources like Tele Atlas'. Then the high-priced tools are going to have some serious competition! I think it will be a little while before we see the full impact of Explorer geocoder. It takes a little while for everyone who can work with C++ to first clear their plate from ongoing projects (people with these skills tend to stay busy), then a little while to read and understand the source code, and then a little while longer to build extensions and them get them back into the open source community. Right now, it's just a few rolling pebbles; the real landslide is still to come. But I noticed that Directions Magazine wants to know the impact right now. There's a little one-click opinon poll on this topic at http://www.allpointsblog.com/ halfway down the page on the right. - Bill Thoen _______________________________________________ MapInfo-L mailing list MapInfo-L@lists.directionsmag.com http://www.directionsmag.com/mailman/listinfo/mapinfo-l