Exception: Error. Google Maps API return status: ZERO_RESULTS for geocode_address

[wolframteetz]

Every few hundred location calls I get this error.

import sys
reload(sys)
<module 'sys' (built-in)>
sys.setdefaultencoding('utf8')
origin = 'Martinsried'
duration = 90
number_of_angles=48
tolerance = 5
tolerance = 2
import isocronut
isochrone = isocronut.generate_isochrone_map(origin, duration, number_of_angles, tolerance)
Iteration 0
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Iteration 6
Traceback (most recent call last):
File "", line 1, in
File "isocronut.py", line 413, in generate_isochrone_map

File "isocronut.py", line 383, in get_isochrone
for i in range(number_of_angles):
File "isocronut.py", line 222, in geocode_address
if not d['status'] == 'OK':
Exception: Error. Google Maps API return status: ZERO_RESULTS

See related issues :

http://stackoverflow.com/questions/20664723/google-maps-api-v3-returning-zero-results-but-maps-google-com-showing-fine

I think one needs to provide a radius iteration to find a place with a name nearby sometimes. See e.g.

http://stackoverflow.com/questions/11754114/why-is-google-place-api-giving-me-zero-results

[wolframteetz]

Test with large radius

(---begin---)
# Parse the json to pull out the geocode
if not d['status'] == 'OK':
full_url = full_url + '&radius=5000'
print "Radius 5000"
# Request geocode from address
req = urllib2.Request(full_url)
opener = urllib2.build_opener()
f = opener.open(req)
d = simplejson.load(f)
if not d['status'] == 'OK':
raise Exception('Error. Google Maps API return status: {}'.format(d['status']))
geocode = [d['results'][0]['geometry']['location']['lat'],
d['results'][0]['geometry']['location']['lng']]
return geocode
(---)

Iteration 0
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Iteration 6
Radius 5000
Traceback (most recent call last):
File "", line 1, in
File "isocronut.py", line 424, in generate_isochrone_map
iso = get_isochrone(origin, duration, number_of_angles, tolerance, access_type, config_path)
File "isocronut.py", line 394, in get_isochrone
iso[i] = geocode_address(data[0][i], access_type, config_path)
File "isocronut.py", line 233, in geocode_address
raise Exception('Error. Google Maps API return status: {}'.format(d['status']))
Exception: Error. Google Maps API return status: ZERO_RESULTS

Failed, so it is some other issue

[wolframteetz]

Debug output appended

CALL

python
import sys
reload(sys)
sys.setdefaultencoding('utf8')
origin=[48.10892,11.451958]
duration = 90
number_of_angles=48
tolerance = 2
import isocronut
isochrone = isocronut.generate_isochrone_map(origin, duration, number_of_angles, tolerance)
from subprocess import call
call(["open", "isochrone.html"])

CODE (in get_isochrone)

for i in range(number_of_angles):
    print "Geocoding #",i
    print "Data ",data[0][i]
    iso[i] = geocode_address(data[0][i], access_type, config_path)
    time.sleep(0.5)

OUTPUT

isochrone = isocronut.generate_isochrone_map(origin, duration, number_of_angles, tolerance)
Iteration 0
Iteration 1
Iteration 2
Iteration 3
Iteration 4
Iteration 5
Iteration 6
Geocoding # 0
Data Ölschlag 3, 92334 Berching, Germany
Geocoding # 1
Data Schweinkofen 21, 92345 Dietfurt, Germany
Geocoding # 2
Data Essing, Germany
Geocoding # 3
Data Auf der Winzerer Höhe, 93059 Regensburg, Germany
Geocoding # 4
Data Unnamed Road, 93073 Neutraubling, Germany
Geocoding # 5
Data Holztraubacher Str. 15, 84066 Mallersdorf-Pfaffenberg, Germany
Geocoding # 6
Data SR65, 94339 Leiblfing, Germany
Geocoding # 7
Data Freundorf 2, 94569 Stephansposching, Germany
Geocoding # 8
Data Marklkofen, Germany
Geocoding # 9
Data Elsenbach 7, 84494 Neumarkt-Sankt Veit, Germany
Geocoding # 10
Data Hölzling 5, 84568 Pleiskirchen, Germany
Geocoding # 11
Data Kraiburg, Germany
Geocoding # 12
Data Aigner 16, 83342 Tacherting, Germany
Geocoding # 13
Data TS31, 83352 Altenmarkt, Germany
Geocoding # 14
Data Betonstraße 20, 83362 Surberg, Germany
Geocoding # 15
Data Unnamed Road, 83324 Ruhpolding, Germany
Geocoding # 16
Data Unterbergen 4, 83246 Unterwössen, Germany
Geocoding # 17
Data Kohlentalstraße 8, 6385 Schwendt, Austria
Geocoding # 18
Data Stadtberg, 6351 Kufstein, Austria
Geocoding # 19
Data Almen 4, 6335 Almen, Austria
Geocoding # 20
Data Unnamed Road, 83700 Rottach-Egern, Germany
Geocoding # 21
Data Unnamed Road, 83708 Kreuth, Germany
Geocoding # 22
Data Unnamed Road, 83661 Lenggries, Germany
Geocoding # 23
Data Unnamed Road, 83661 Lenggries, Germany
Geocoding # 24
Data Gemeinde Vomp, Austria
Geocoding # 25
Data Unnamed Road, 6108, Austria
Waiting 5s
Radius 5000
Traceback (most recent call last):
File "", line 1, in
File "isocronut.py", line 429, in generate_isochrone_map

File "isocronut.py", line 399, in get_isochrone
time.sleep(0.5)
File "isocronut.py", line 236, in geocode_address
geocode = [d['results'][0]['geometry']['location']['lat'],
Exception: Error. Google Maps API return status: ZERO_RESULTS

[wolframteetz]

Exception can be caught with s.th. like this, but should be failsafe for i=0. Better idea?

I don't understand why there have to be these addresses anyway ;) I think they are also the primary reason for the outliers - geocoding Unknown Road, can lead to surprising results of course...

for i in range(number_of_angles):
    print "Geocoding #",i
    print "Data ",data[0][i]
    try:
        iso[i] = geocode_address(data[0][i], access_type, config_path)
    except:
        if i > 0:
            print "Exception caught for isocoding"
            iso[i] = iso[i-1]           
    time.sleep(0.1)

[drewfustin]

I think the proper way to handle this is:

a) initialize iso with None:

iso = [None] * number_of_angles

b) don't bother resetting iso[i] for i that fails:

    try:
        iso[i] = geocode_address(data[0][i], access_type, config_path)
    except:
        print "Exception caught for isocoding"

c) then after the i for loop, throw out all angles that failed, and fail ultimately if none of them work at all:

iso = [x for x in iso if x is not None]
if len(iso) == 0:
    raise Exception('No valid points along the isochrone were found.')

Thoughts?

[wolframteetz]

Proper but still very messy because translating a known location to "unknown road, cleveland ohio" and back introduced terrible noise outside of city areas and quite some outliers.

Working on the mapbox solution instead, it's much nicer anyway.

[drewfustin]

Yeah, my other thought was: as a post-process, determine outlying radii using some outlier detection method and through those out.