John Sample

# re: Lax

Great to hear that you havn't dropped out of sight. I try to check your site about twice a week to see if you have anything new. Keep up the good work... I think that there are many that will be interested in your dll in the future.

I'm still waiting for some C# usage examples to be posted in the docs... although having the API reference should be enough to get most people going.

-Mark

# re: Lax

Excellent work John!
I hate using PERL ... especially on Windows ;)

2 Questions:
I noticed that 2 "TIGER" tables are created TIGER_01 & TIGER_02 even though no data is put in TIGER_02. Is this correct?

1 1/2 more question: I think the GeoCoder.us PERL code uses 5 Tiger files in each ZIP file... Will your code produce as robust and *COMPLETE* a dataset (as geocoder.us's)? If not, are there any improvements to the completeness of the data I could help with?

Thanks & Best Regards, Dan!

# re: Lax

The installer just does Type 1, mostly because Type 2 is overkill in all but a few circumstances.
Merging in the the other types isn't hard, I did it manually, but the installer won't do it yet. I'm going to release the source soon so any mods would be welcome.

# re: Lax

Hey John,
Out of a sample of 100's of addresses close to my house (in Colorado) I found the following 'conversion' or lookup rates:
Found using .NET Geocode.dll: 34%
Found using Perl Code: 88%
Not Found: 12%
(BTW, about 12-18% of the Addresses that were not found could be found in Google Maps... I want their GeoCode DB :P )

Can you point me in the right direction as to how I can import all the TIGER files ... I would love to get the SQL DB as complete, if not better than, the Perl/Berkley File DB code.

Thanks John!!

Best Regards, Dan!

# re: Lax

I believe the alternate street names are in type 5, its in the tiger reference but I don't have it handy at the moment. You would need to map these to the "official" name using the TLID column.

Yikes, 34% seems incredibly low. I'd be interested to know whether this is due to the regex not being able to parse the address (and therefore not even attempt a lookup). I don't think the type 5 records contain 60% of the street names, but I could be wrong.

If its due to the parsing you can override the regex using the API. See this post for more info:
http://www.johnsample.com/archive/2005/09/15/801.aspx#804

# re: Lax

Hi Dan:

If you eventually get your DB fully complete, would you (at that point) mind describing step-by-step the process you used to do it? I too would like to have a very complete and up-to-date database!!

Also, because I'm a newbie to geocoding, what are the relevance of the various "type" files such as Type-5 and Type-2? Wouldn't you naturally just want Types 1-5 so that you'ld have everything?

-Mark

# re: Lax

I'll post a better description soon, along with an announcement about some new developments I think you guys will like.

# re: Lax

For info about the TIGER/Line data, see here: http://www.census.gov/geo/www/tiger/tiger2004se/tl2004semeta.txt
Look for the section 3/4 down the page titled "Entity_and_Attribute_Overview."
Keep in mind not all the files are relavent for address/street geocoding.

Here is some other (more generic) info: http://www.census.gov/geo/www/tiger/tiger2004se/readme.txt
====
I will let everyone know if I can figure out an easy way to load the data - without reinventing the wheel ;) .

# re: Lax

John: By the way, I did some manual tweaking of the "bad" addresses ... no luck. I seems like your Reg. Exp. works quite well.