[postgis-devel] Tiger 2010 Geocoder Divide By Zero Error

[Paragon Corporation]

Emily,

Good catch.  I really should have revised the least_hn etc to handle
decimals instead of rejecting them.  I had run into a situation where tiger
data had a goofy
Address number of 1.0.0.0 so was breaking the index I was building   So I
had decided to just reject these things because I wasn't clear what a
decimal address would mean anyway.  But I forgot 
About that other piece of code that was making everything decimals.

Regarding why the .0 is there  -- it's to make it a numeric instead of an
integer (since parsed.address always returns an integer).

Unfortunately because least_hn/greatest_hn always return back integers that
hack wouldn't work anyway.

The code was right when it was there before I converted everything to using
least_hn/greatest_hn so I accidentally changed the meaning of it.


The reason why you want to divide by decimals in SQL is because

SELECT 1/2  => 0

But 

SELECT 1.0/2 => 0.5


Anyrate I've put this in as a bug and will fix in a bit.
http://trac.osgeo.org/postgis/ticket/1035

Thanks,
Regina
 

-----Original Message-----
From: postgis-devel-bounces at postgis.refractions.net
[mailto:postgis-devel-bounces at postgis.refractions.net] On Behalf Of Emily
Gouge
Sent: Friday, June 17, 2011 7:06 PM
To: postgis-devel at postgis.refractions.net
Subject: [postgis-devel] Tiger 2010 Geocoder Divide By Zero Error

I have recently been doing some geocoding testing using the Postgis geocoder
from 2.0.0 branch.  I started getting divide by zero errors today and was
looking through the source code at the geocode_address function.  After
figuring out the source of the error I looked in SVN and saw there was a fix
committed about a week ago (thanks!) but I'm still somewhat confused about
this part of the code.

The original divide by 0 error is caused by the following partial statement:

  ((1.0 - ' || '(least_hn(' || coalesce(quote_literal(parsed.address ||
'.0'),'NULL') || ',least_hn(b.fromhn,b.tohn)::text) /' || ' 
greatest_hn(' || coalesce(quote_literal(parsed.address || '.0'),'NULL') 
|| ',greatest_hn(b.fromhn,b.tohn)::text))' || ') * 5)::integer + 5'

My question is what is the purpose of the '.0' concatenation?

I looked at the greatest_hn function and if any of it's parameters 
contain characters that are not digits then it returns 0.  So by 
concatenating the '.0' I believe you are guaranteeing the first argument 
is going to be treated as 0?  So either the '.0' should not be there or 
the statement can be simplified.  If I'm missing something please 
correct me.

Thanks,
Emily
_______________________________________________
postgis-devel mailing list
postgis-devel at postgis.refractions.net
http://postgis.refractions.net/mailman/listinfo/postgis-devel