def geocode(schema=None):
"""
Geocode NewsItems with null locations.
If ``schema`` is provided, only geocode NewsItems with that particular
schema slug.
"""
geocoder = SmartGeocoder()
qs = NewsItem.objects.filter(location__isnull=True).order_by('-id')
if schema is not None:
print "Geocoding %s..." % schema
qs = qs.filter(schema__slug=schema)
else:
print "Geocoding all ungeocoded newsitems..."
geocoded_count = 0
not_found_count = 0
ambiguous_count = 0
parsing_error_count = 0
invalid_block_count = 0
for ni in qs.iterator():
loc_name = ni.location_name
try:
add = geocoder.geocode(loc_name)
except InvalidBlockButValidStreet:
print ' invalid block but valid street: %s' % loc_name
invalid_block_count += 1
except AmbiguousResult, e:
print ' ambiguous: %s' % loc_name
ambiguous_count += 1
except GeocodingException, e:
print ' not found: %s' % loc_name
not_found_count += 1
class TestSmartGeocoder(django.test.TestCase):
fixtures = ["wabash.yaml"]
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
@mock.patch("ebpub.streets.models.get_metro")
def test_address_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {"city_name": "CHICAGO", "multiple_cities": False}
address = self.geocoder.geocode("200 S Wabash")
self.assertEqual(address["city"], "Chicago")
@mock.patch("ebpub.streets.models.get_metro")
def test_address_geocoder_ambiguous(self, mock_get_metro):
mock_get_metro.return_value = {"city_name": "CHICAGO", "multiple_cities": False}
self.assertRaises(AmbiguousResult, self.geocoder.geocode, "220 Wabash")
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet, self.geocoder.geocode, "100000 S Wabash")
@mock.patch("ebpub.streets.models.get_metro")
def test_block_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {"city_name": "CHICAGO", "multiple_cities": False}
address = self.geocoder.geocode("200 block of Wabash")
self.assertEqual(address["city"], "Chicago")
def test_intersection_geocoder(self):
address = self.geocoder.geocode("Wabash and Jackson")
self.assertEqual(address["city"], "CHICAGO")
def clean(self):
loc_info = self.cleaned_data.get('location')
if isinstance(loc_info, list):
# olwidget wraps geometries up as lists in case there's several per map
assert len(loc_info) == 1
loc_info = loc_info[0]
if not loc_info:
address = self.cleaned_data.get('address')
if not address:
self._append_error(
'location',
u'Either an address or a location must be specified.')
else:
# try to geocode the address...
try:
geocoder = SmartGeocoder()
addr = geocoder.geocode(address)
loc_info = addr['point']
except AmbiguousResult:
self._append_error(
'location',
u'Address is ambiguous, please specify a point directly.'
)
except GeocodingException:
self._append_error(
'location',
u'Unable to geocode address, please correct the address or specify a point directly.'
)
# Again, olwidget expects these to be lists...
loc_info = [loc_info]
self.cleaned_data['location'] = loc_info
return super(PlaceAdminForm, self).clean()
class BaseGeocoderTestCase(django.test.TestCase):
fixtures = ['wabash.yaml']
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
address = self.geocoder.geocode('200 S Wabash')
self.assertEqual(address['city'], 'Chicago')
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder_ambiguous(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
self.assertRaises(AmbiguousResult, self.geocoder.geocode, '220 Wabash')
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet, self.geocoder.geocode, '100000 S Wabash')
@mock.patch('ebpub.streets.models.get_metro')
def test_block_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
address = self.geocoder.geocode('200 block of Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_intersection_geocoder(self):
address = self.geocoder.geocode('Wabash and Jackson')
self.assertEqual(address['city'], 'CHICAGO')
def __init__(self, use_cache=True):
if not use_cache:
self.retriever = Retriever(cache=None, sleep=self.sleep, timeout=self.timeout)
else:
self.retriever = Retriever(sleep=self.sleep, timeout=self.timeout)
self.logger = logging.getLogger('eb.retrieval.%s' % self.logname)
self.start_time = datetime.datetime.now()
self._geocoder = SmartGeocoder()
def __init__(self, *args, **kwargs):
if self.logname is None:
self.logname = '%s.%s' % (settings.SHORT_NAME, self.schema_slugs[0])
super(NewsItemListDetailScraper, self).__init__(*args, **kwargs)
self._schema_cache = None
self._schemas_cache = None
self._lookups_cache = None
self._schema_fields_cache = None
self._schema_field_mapping_cache = None
self._geocoder = SmartGeocoder()
def full_geocode(query, search_places=True):
"""
Tries the full geocoding stack on the given query (a string):
* Normalizes whitespace/capitalization
* Searches the Misspelling table to corrects location misspellings
* Searches the Location table
* Failing that, searches the Place table (if search_places is True)
* Failing that, uses the given geocoder to parse this as an address
* Failing that, raises whichever error is raised by the geocoder --
except AmbiguousResult, in which case all possible results are
returned
Returns a dictionary of {type, result, ambiguous}, where ambiguous is True
or False and type can be:
* 'location' -- in which case result is a Location object.
* 'place' -- in which case result is a Place object. (This is only
possible if search_places is True.)
* 'address' -- in which case result is an Address object as returned
by geocoder.geocode().
* 'block' -- in which case result is a list of Block objects.
If ambiguous is True, result will be a list of objects.
"""
query = normalize(query)
# First, try correcting the spelling ("LAKEVIEW" -> "LAKE VIEW").
try:
miss = Misspelling.objects.get(incorrect=query)
except Misspelling.DoesNotExist:
pass
else:
query = miss.correct
# Search the Location table.
try:
loc = Location.objects.get(normalized_name=query)
except Location.DoesNotExist:
pass
else:
return {'type': 'location', 'result': loc, 'ambiguous': False}
# Search the Place table, for stuff like "Sears Tower".
if search_places:
places = Place.objects.filter(normalized_name=query)
if len(places) == 1:
return {'type': 'place', 'result': places[0], 'ambiguous': False}
elif len(places) > 1:
return {'type': 'place', 'result': places, 'ambiguous': True}
# Try geocoding this as an address.
geocoder = SmartGeocoder()
try:
result = geocoder.geocode(query)
except AmbiguousResult, e:
return {'type': 'address', 'result': e.choices, 'ambiguous': True}
def full_geocode(query, search_places=True):
"""
Tries the full geocoding stack on the given query (a string):
* Normalizes whitespace/capitalization
* Searches the Misspelling table to corrects location misspellings
* Searches the Location table
* Failing that, searches the Place table (if search_places is True)
* Failing that, uses the given geocoder to parse this as an address
* Failing that, raises whichever error is raised by the geocoder --
except AmbiguousResult, in which case all possible results are
returned
Returns a dictionary of {type, result, ambiguous}, where ambiguous is True
or False and type can be:
* 'location' -- in which case result is a Location object.
* 'place' -- in which case result is a Place object. (This is only
possible if search_places is True.)
* 'address' -- in which case result is an Address object as returned
by geocoder.geocode().
* 'block' -- in which case result is a list of Block objects.
If ambiguous is True, result will be a list of objects.
"""
query = normalize(query)
# First, try correcting the spelling ("LAKEVIEW" -> "LAKE VIEW").
try:
miss = Misspelling.objects.get(incorrect=query)
except Misspelling.DoesNotExist:
pass
else:
query = miss.correct
# Search the Location table.
try:
loc = Location.objects.get(normalized_name=query)
except Location.DoesNotExist:
pass
else:
return {'type': 'location', 'result': loc, 'ambiguous': False}
# Search the Place table, for stuff like "Sears Tower".
if search_places:
places = Place.objects.filter(normalized_name=query)
if len(places) == 1:
return {'type': 'place', 'result': places[0], 'ambiguous': False}
elif len(places) > 1:
return {'type': 'place', 'result': places, 'ambiguous': True}
# Try geocoding this as an address.
geocoder = SmartGeocoder()
try:
result = geocoder.geocode(query)
except AmbiguousResult, e:
return {'type': 'address', 'result': e.choices, 'ambiguous': True}
def geocode(schema=None):
"""
Geocode NewsItems with null locations.
If ``schema`` is provided, only geocode NewsItems with that particular
schema slug.
"""
geocoder = SmartGeocoder()
qs = NewsItem.objects.filter(location__isnull=True).order_by('-id')
if schema is not None:
print "Geocoding %s..." % schema
qs = qs.filter(schema__slug=schema)
else:
print "Geocoding all ungeocoded newsitems..."
geocoded_count = 0
not_found_count = 0
ambiguous_count = 0
parsing_error_count = 0
invalid_block_count = 0
for ni in qs.iterator():
loc_name = ni.location_name
try:
add = geocoder.geocode(loc_name)
except InvalidBlockButValidStreet:
print ' invalid block but valid street: %s' % loc_name
invalid_block_count += 1
except AmbiguousResult:
print ' ambiguous: %s' % loc_name
ambiguous_count += 1
except GeocodingException:
print ' not found: %s' % loc_name
not_found_count += 1
except ParsingError:
print ' parse error: %s' % loc_name
parsing_error_count += 1
except:
raise
else:
ni.location = add['point']
ni.block = add['block']
ni.save()
print '%s (%s)' % (loc_name, ni.item_url())
geocoded_count += 1
else:
print "No NewsItems with null locations found"
print "------------------------------------------------------------------"
print "Geocoded: %s" % geocoded_count
print "Not found: %s" % not_found_count
print "Ambiguous: %s" % ambiguous_count
print "Parse errors: %s" % parsing_error_count
print "Invalid blocks: %s" % invalid_block_count
def geocode(schema=None):
"""
Geocode NewsItems with null locations.
If ``schema`` is provided, only geocode NewsItems with that particular
schema slug.
"""
geocoder = SmartGeocoder()
qs = NewsItem.objects.filter(location__isnull=True).order_by('-id')
if schema is not None:
print "Geocoding %s..." % schema
qs = qs.filter(schema__slug=schema)
else:
print "Geocoding all ungeocoded newsitems..."
geocoded_count = 0
not_found_count = 0
ambiguous_count = 0
parsing_error_count = 0
invalid_block_count = 0
for ni in qs.iterator():
loc_name = ni.location_name
try:
add = geocoder.geocode(loc_name)
except InvalidBlockButValidStreet:
print ' invalid block but valid street: %s' % loc_name
invalid_block_count += 1
except AmbiguousResult:
print ' ambiguous: %s' % loc_name
ambiguous_count += 1
except GeocodingException:
print ' not found: %s' % loc_name
not_found_count += 1
except ParsingError:
print ' parse error: %s' % loc_name
parsing_error_count += 1
except:
raise
else:
ni.location = add['point']
ni.block = add['block']
ni.save()
print '%s (%s)' % (loc_name, ni.item_url())
geocoded_count += 1
else:
print "No NewsItems with null locations found"
print "------------------------------------------------------------------"
print "Geocoded: %s" % geocoded_count
print "Not found: %s" % not_found_count
print "Ambiguous: %s" % ambiguous_count
print "Parse errors: %s" % parsing_error_count
print "Invalid blocks: %s" % invalid_block_count
def full_geocode(query, search_places=True):
"""
Tries the full geocoding stack on the given query (a string):
* Normalizes whitespace/capitalization
* Searches the Misspelling table to corrects location misspellings
* Searches the Location table
* Failing that, searches the Place table (if search_places is True)
* Failing that, uses the given geocoder to parse this as an address
* Failing that, raises whichever error is raised by the geocoder --
except AmbiguousResult, in which case all possible results are
returned
Returns a dictionary of {type, result, ambiguous}, where ambiguous is True
or False and type can be:
* 'location' -- in which case result is a Location object.
* 'place' -- in which case result is a Place object. (This is only
possible if search_places is True.)
* 'address' -- in which case result is an Address object as returned
by geocoder.geocode().
* 'block' -- in which case result is a list of Block objects.
If ambiguous is True, result will be a list of objects.
"""
# Search the Location table.
try:
canonical_loc = LocationSynonym.objects.get_canonical(query)
loc = Location.objects.get(normalized_name=canonical_loc)
except Location.DoesNotExist:
pass
else:
logger.debug('geocoded %r to Location %s' % (query, loc))
return {'type': 'location', 'result': loc, 'ambiguous': False}
# Search the Place table, for stuff like "Sears Tower".
if search_places:
canonical_place = PlaceSynonym.objects.get_canonical(query)
places = Place.objects.filter(normalized_name=canonical_place)
if len(places) == 1:
logger.debug(u'geocoded %r to Place %s' % (query, places[0]))
return {'type': 'place', 'result': places[0], 'ambiguous': False}
elif len(places) > 1:
logger.debug(u'geocoded %r to multiple Places: %s' % (query, unicode(places)))
return {'type': 'place', 'result': places, 'ambiguous': True}
# Try geocoding this as an address.
geocoder = SmartGeocoder(use_cache=getattr(settings, 'EBPUB_CACHE_GEOCODER', False))
try:
result = geocoder.geocode(query)
except AmbiguousResult, e:
logger.debug('Multiple addresses for %r' % query)
return {'type': 'address', 'result': e.choices, 'ambiguous': True}
def add_newsitem(seed_url, seed_name, url, article_headline, article_date,
name_excerpts):
schema = Schema.objects.get(slug='news-articles')
geocoder = SmartGeocoder()
try:
s = Seed.objects.get(url=seed_url)
except Seed.DoesNotExist:
s = Seed.objects.create(
url=seed_url,
base_url=seed_url,
delay=0,
depth=0,
is_crawled=False,
is_rss_feed=False,
is_active='t',
rss_full_entry=False,
normalize_www=3,
pretty_name=seed_name,
schema=schema,
autodetect_locations=True,
guess_article_text=False,
strip_noise=False,
city='',
)
try:
p = Page.objects.get(url=url)
except Page.DoesNotExist:
html = UnicodeRetriever().fetch_data(url)
p = Page.objects.create(seed=s,
url=url,
scraped_url=url,
html=html,
when_crawled=datetime.datetime.now(),
is_article=True,
is_pdf=False,
is_printer_friendly=False,
article_headline=article_headline,
article_date=article_date,
has_addresses=None,
when_geocoded=None,
geocoded_by='',
times_skipped=0,
robot_report='')
data_tuples = []
for location_name, excerpt in name_excerpts:
point = geocoder.geocode(location_name) # Let exceptions bubble up.
data_tuples.append(
(location_name, point['point'], excerpt, point['block']))
return geotag_page(p.id, seed_name, schema, url, data_tuples,
article_headline, article_date)
def add_newsitem(seed_url, seed_name, url, article_headline, article_date, name_excerpts):
schema = Schema.objects.get(slug='news-articles')
geocoder = SmartGeocoder()
try:
s = Seed.objects.get(url=seed_url)
except Seed.DoesNotExist:
s = Seed.objects.create(
url=seed_url,
base_url=seed_url,
delay=0,
depth=0,
is_crawled=False,
is_rss_feed=False,
is_active='t',
rss_full_entry=False,
normalize_www=3,
pretty_name=seed_name,
schema=schema,
autodetect_locations=True,
guess_article_text=False,
strip_noise=False,
city='',
)
try:
p = Page.objects.get(url=url)
except Page.DoesNotExist:
html = UnicodeRetriever().get_html(url)
p = Page.objects.create(
seed=s,
url=url,
scraped_url=url,
html=html,
when_crawled=datetime.datetime.now(),
is_article=True,
is_pdf=False,
is_printer_friendly=False,
article_headline=article_headline,
article_date=article_date,
has_addresses=None,
when_geocoded=None,
geocoded_by='',
times_skipped=0,
robot_report=''
)
data_tuples = []
for location_name, excerpt in name_excerpts:
point = geocoder.geocode(location_name) # Let exceptions bubble up.
data_tuples.append((location_name, point['point'], excerpt, point['block']))
return geotag_page(p.id, seed_name, schema, url, data_tuples, article_headline, article_date)
def quick_dirty_fallback_geocode(addr, parse=True):
"""
Try to get SOME x,y even with bad blocks data,
by falling back to external geocoders.
"""
from ebdata.nlp.addresses import parse_addresses
from ebpub.geocoder import SmartGeocoder
if parse:
addrs = parse_addresses(addr)
else:
addrs = [addr]
for addr, unused in addrs:
try:
try:
result = SmartGeocoder().geocode(addr)
point = result['point']
logger.debug("internally geocoded %r" % addr)
return point.x, point.y
except GeocodingException:
logger.debug("internal geocoder failed on %r:\n" % addr)
log_exception(level=logging.DEBUG)
x, y = None, None
# XXX Don't bother, external geocoding rarely gives us
# anything inside Boston now that we have decent
# blocks data. But I want to preserve this script for
# now till we figure out what to do with geocoding
# more generally
continue
except:
logger.error('uncaught geocoder exception on %r\n' % addr)
log_exception()
return None, None
class TestSmartGeocoder(django.test.TestCase):
fixtures = ['wabash.yaml']
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
result = self.geocoder.geocode('200 S Wabash Ave')
self.assertEqual(result['city'], 'Chicago')
self.assertEqual(result['address'], '200 S Wabash Ave.')
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder__wrong_suffix_works(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
result = self.geocoder.geocode('220 S Wabash St')
self.assertEqual(result['address'], '220 S Wabash Ave.')
# Or none at all.
result = self.geocoder.geocode('220 S Wabash')
self.assertEqual(result['address'], '220 S Wabash Ave.')
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder_ambiguous(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
# Ambiguous because of missing pre_dir.
self.assertRaises(AmbiguousResult, self.geocoder.geocode, '220 Wabash')
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet,
self.geocoder.geocode, '100000 S Wabash')
@mock.patch('ebpub.streets.models.get_metro')
def test_block_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {'city_name': 'CHICAGO',
'multiple_cities': False}
address = self.geocoder.geocode('200 block of Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_intersection_geocoder(self):
address = self.geocoder.geocode('Wabash and Jackson')
self.assertEqual(address['city'], 'CHICAGO')
class BaseGeocoderTestCase(django.test.TestCase):
fixtures = ['wabash.yaml']
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {
'city_name': 'CHICAGO',
'multiple_cities': False
}
address = self.geocoder.geocode('200 S Wabash')
self.assertEqual(address['city'], 'Chicago')
@mock.patch('ebpub.streets.models.get_metro')
def test_address_geocoder_ambiguous(self, mock_get_metro):
mock_get_metro.return_value = {
'city_name': 'CHICAGO',
'multiple_cities': False
}
self.assertRaises(AmbiguousResult, self.geocoder.geocode, '220 Wabash')
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet, self.geocoder.geocode,
'100000 S Wabash')
@mock.patch('ebpub.streets.models.get_metro')
def test_block_geocoder(self, mock_get_metro):
mock_get_metro.return_value = {
'city_name': 'CHICAGO',
'multiple_cities': False
}
address = self.geocoder.geocode('200 block of Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_intersection_geocoder(self):
address = self.geocoder.geocode('Wabash and Jackson')
self.assertEqual(address['city'], 'CHICAGO')
class BaseGeocoderTestCase(unittest.TestCase):
fixtures = ['wabash.yaml']
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
def test_address_geocoder(self):
address = self.geocoder.geocode('200 S Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_address_geocoder_ambiguous(self):
self.assertRaises(AmbiguousResult, self.geocoder.geocode, '200 Wabash')
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet, self.geocoder.geocode, '100000 S Wabash')
def test_block_geocoder(self):
address = self.geocoder.geocode('200 block of Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_intersection_geocoder(self):
address = self.geocoder.geocode('Wabash and Jackson')
self.assertEqual(address['city'], 'CHICAGO')
class BaseGeocoderTestCase(unittest.TestCase):
fixtures = ['wabash.yaml']
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
def test_address_geocoder(self):
address = self.geocoder.geocode('200 S Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_address_geocoder_ambiguous(self):
self.assertRaises(AmbiguousResult, self.geocoder.geocode, '200 Wabash')
def test_address_geocoder_invalid_block(self):
self.assertRaises(InvalidBlockButValidStreet, self.geocoder.geocode,
'100000 S Wabash')
def test_block_geocoder(self):
address = self.geocoder.geocode('200 block of Wabash')
self.assertEqual(address['city'], 'Chicago')
def test_intersection_geocoder(self):
address = self.geocoder.geocode('Wabash and Jackson')
self.assertEqual(address['city'], 'CHICAGO')
def clean(self):
loc_info = self.cleaned_data.get('location')
if isinstance(loc_info, list):
# olwidget wraps geometries up as lists in case there's several per map
assert len(loc_info) == 1
loc_info = loc_info[0]
if not loc_info:
address = self.cleaned_data.get('address')
if not address:
self._append_error('location', u'Either an address or a location must be specified.')
else:
# try to geocode the address...
try:
geocoder = SmartGeocoder()
addr = geocoder.geocode(address)
loc_info = addr['point']
except AmbiguousResult:
self._append_error('location', u'Address is ambiguous, please specify a point directly.')
except GeocodingException:
self._append_error('location', u'Unable to geocode address, please correct the address or specify a point directly.')
# Again, olwidget expects these to be lists...
loc_info = [loc_info]
self.cleaned_data['location'] = loc_info
return super(PlaceAdminForm, self).clean()
def main(argv=None):
url = 'http://search.boston.com/search/api?q=*&sort=-articleprintpublicationdate&subject=boston&scope=bonzai'
schema = 'local-news'
try:
schema = Schema.objects.get(slug=schema)
except Schema.DoesNotExist:
print "Schema (%s): DoesNotExist" % schema
sys.exit(0)
f = feedparser.parse(url)
geocoder = SmartGeocoder()
for e in f.entries:
try:
item = NewsItem.objects.get(title=e.title,
description=e.description)
except NewsItem.DoesNotExist:
item = NewsItem()
item.schema = schema
item.title = e.title
item.description = e.description
item.url = e.link
#item.location_name = e['x-calconnect-street']
item.item_date = datetime.datetime(*e.updated_parsed[:6])
item.pub_date = datetime.datetime(*e.updated_parsed[:6])
try:
if 'point' in e:
x, y = e.point.split(' ')
else:
x, y = e.georss_point.split(' ')
item.location = Point((float(y), float(x)))
item.save()
except:
pass
print "Added: %s" % item.title
def fix_newsitem_coords(item, dry_run=True):
"""
Try to fix a (presumably bad) NewsItem geometry by reversing its
coordinates, or reverse-geocoding if it has a location name; use
whatever works.
If dry_run=False, the item will be saved.
"""
if item.location is not None:
loc = item.location.centroid
print "Found %r outside bounds at %s, %s" % (item.title, loc.x, loc.y)
else:
loc = None
print "NO location on %s: %s" % (item.schema.slug, item.title)
fixed = False
if item.location_name:
from ebpub.geocoder import SmartGeocoder, AmbiguousResult
try:
result = SmartGeocoder().geocode(item.location_name)
except AmbiguousResult, e:
print "...%d choices, picking the first one" % len(e.choices)
result = e.choices[0]
except:
def update_from_query_params(self, request):
"""
Update the filters based on query parameters.
After this is called, it's recommended to redirect
to a normalized form of the URL, eg. self.sort(); self.make_url()
This takes care to preserve query parameters that aren't used
by any of the NewsitemFilters.
"""
# Make a mutable copy so we can leave only the params that FilterChain
# doesn't know about.
params = request.GET.copy()
def pop_key(key):
# request.GET.pop() returns a sequence.
# We only want a single value, stripped.
val = params.pop(key, [''])[0]
return val.strip()
address = pop_key('address')
if address:
xy_radius, block_radius, cookies_to_set = block_radius_value(
request)
params.pop('radius', None)
result = None
try:
result = SmartGeocoder().geocode(address)
except AmbiguousResult, e:
raise BadAddressException(address,
block_radius,
address_choices=e.choices)
except (GeocodingException, ParsingError):
raise BadAddressException(address,
block_radius,
address_choices=())
#
# everyblock is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with everyblock. If not, see <http://www.gnu.org/licenses/>.
#
from ebdata.retrieval.utils import locations_are_close
from ebpub.db.models import NewsItem
from ebpub.geocoder import SmartGeocoder, ParsingError, GeocodingException
from django.contrib.gis.geos import Point
geocoder = SmartGeocoder()
THRESHOLD = 375
def fix_crime_geom():
qs = NewsItem.objects.filter(schema__slug='crime', location__isnull=False)
count = qs.count()
for i, ni in enumerate(qs.iterator()):
print '# => Checking %s of %s' % (i, count)
x, y = [float(n) for n in ni.attributes['xy'].split(';')]
pt = Point((x, y))
pt.srid = 4326
location_name = ni.location_name.replace('XX', '01')
try:
result = geocoder.geocode(location_name)
except (GeocodingException, ParsingError):
class NewsItemListDetailScraper(ListDetailScraper):
"""
A ListDetailScraper that saves its data into the NewsItem table.
Subclasses are required to set the `schema_slugs` attribute.
self.schemas lazily loads the list of Schema objects the first time it's
accessed. It is a dictionary in the format {slug: Schema}.
self.schema is available if schema_slugs has only one element. It's the
Schema object.
self.lookups lazily loads a dictionary of all SchemaFields with
lookup=True. The dictionary is in the format {name: schemafield}. If
schema_slug has more than one element, self.lookups is a dictionary in the
format {schema_slug: {name: schemafield}}.
self.schema_field_mapping lazily loads a dictionary of each SchemaField,
mapping the name to the real_name. If schema_slug has more than one element,
self.schema_field_mapping is a dictionary in the format
{schema_slug: {name: real_name}}.
"""
schema_slugs = None
logname = None
def __init__(self, *args, **kwargs):
if self.logname is None:
self.logname = '%s.%s' % (settings.SHORT_NAME, self.schema_slugs[0])
super(NewsItemListDetailScraper, self).__init__(*args, **kwargs)
self._schema_cache = None
self._schemas_cache = None
self._lookups_cache = None
self._schema_fields_cache = None
self._schema_field_mapping_cache = None
self._geocoder = SmartGeocoder()
# schemas, schema, lookups and schema_field_mapping are all lazily loaded
# so that this scraper can be run (in raw_data(), xml_data() or
# display_data()) without requiring a valid database to be set up.
def _get_schemas(self):
if self._schemas_cache is None:
self._schemas_cache = dict([(s, Schema.objects.get(slug=s)) for s in self.schema_slugs])
return self._schemas_cache
schemas = property(_get_schemas)
def _get_schema(self):
if self._schema_cache is None:
if len(self.schema_slugs) > 1:
raise AttributeError('self.schema is only available if len(schema_slugs) == 1')
self._schema_cache = self.schemas[self.schema_slugs[0]]
return self._schema_cache
schema = property(_get_schema)
def _get_lookups(self):
if self._lookups_cache is None:
lc = dict([(s.slug, dict([(sf.name, sf) for sf in s.schemafield_set.filter(is_lookup=True)])) for s in self.schemas.values()])
if len(self.schema_slugs) == 1:
lc = lc[self.schema_slugs[0]]
self._lookups_cache = lc
return self._lookups_cache
lookups = property(_get_lookups)
def _get_schema_fields(self):
if self._schema_fields_cache is None:
sfs = dict([(s.slug, dict([(sf.name, sf) for sf in s.schemafield_set.all()])) for s in self.schemas.values()])
if len(self.schema_slugs) == 1:
sfs = sfs[self.schema_slugs[0]]
self._schema_fields_cache = sfs
return self._schema_fields_cache
schema_fields = property(_get_schema_fields)
def _get_schema_field_mapping(self):
if self._schema_field_mapping_cache is None:
schema_objs = self.schemas.values()
mapping = field_mapping([s.id for s in schema_objs])
fm = dict([(s.slug, mapping[s.id]) for s in schema_objs])
if len(self.schema_slugs) == 1:
fm = fm[self.schema_slugs[0]]
self._schema_field_mapping_cache = fm
return self._schema_field_mapping_cache
schema_field_mapping = property(_get_schema_field_mapping)
def get_or_create_lookup(self, schema_field_name, name, code, description='', schema=None, make_text_slug=True):
"""
Returns the Lookup instance matching the given Schema slug, SchemaField
name and Lookup.code, creating it (with the given name/code/description)
if it doesn't already exist.
If make_text_slug is True, then a slug will be created from the given
name. If it's False, then the slug will be the Lookup's ID.
"""
if len(self.schema_slugs) > 1:
sf = self.lookups[schema][schema_field_name]
else:
sf = self.lookups[schema_field_name]
return Lookup.objects.get_or_create_lookup(sf, name, code, description, make_text_slug, self.logger)
@transaction.commit_on_success
def create_newsitem(self, attributes, **kwargs):
"""
Creates and saves a NewsItem with the given kwargs. Returns the new
NewsItem.
kwargs MUST have the following keys:
title
item_date
location_name
For any other kwargs whose values aren't provided, this will use
sensible defaults.
kwargs may optionally contain a 'convert_to_block' boolean. If True,
this will convert the given kwargs['location_name'] to a block level
but will use the real (non-block-level) address for geocoding and Block
association.
attributes is a dictionary to use to populate this NewsItem's Attribute
object.
"""
block = location = None
if 'location' not in kwargs:
location = self.geocode(kwargs['location_name'])
if location:
block = location['block']
location = location['point']
if kwargs.pop('convert_to_block', False):
kwargs['location_name'] = address_to_block(kwargs['location_name'])
# If the exact address couldn't be geocoded, try using the
# normalized location name.
if location is None:
location = self.geocode(kwargs['location_name'])
if location:
block = location['block']
location = location['point']
# Normally we'd just use "schema = kwargs.get('schema', self.schema)",
# but self.schema will be evaluated even if the key is found in
# kwargs, which raises an error when using multiple schemas.
schema = kwargs.get('schema', None)
schema = schema or self.schema
ni = NewsItem.objects.create(
schema=schema,
title=kwargs['title'],
description=kwargs.get('description', ''),
url=kwargs.get('url', ''),
pub_date=kwargs.get('pub_date', self.start_time),
item_date=kwargs['item_date'],
location=kwargs.get('location', location),
location_name=kwargs['location_name'],
location_object=kwargs.get('location_object', None),
block=kwargs.get('block', block),
)
ni.attributes = attributes
self.num_added += 1
self.logger.info(u'Created NewsItem %s (total created in this scrape: %s)', ni.id, self.num_added)
return ni
@transaction.commit_on_success
def update_existing(self, newsitem, new_values, new_attributes):
"""
Given an existing NewsItem and dictionaries new_values and
new_attributes, determines which values and attributes have changed
and saves the object and/or its attributes if necessary.
"""
newsitem_updated = False
# First, check the NewsItem's values.
for k, v in new_values.items():
if getattr(newsitem, k) != v:
self.logger.info('ID %s %s changed from %r to %r' % (newsitem.id, k, getattr(newsitem, k), v))
setattr(newsitem, k, v)
newsitem_updated = True
if newsitem_updated:
newsitem.save()
# Next, check the NewsItem's attributes.
for k, v in new_attributes.items():
if newsitem.attributes[k] != v:
self.logger.info('ID %s %s changed from %r to %r' % (newsitem.id, k, newsitem.attributes[k], v))
newsitem.attributes[k] = v
newsitem_updated = True
if newsitem_updated:
self.num_changed += 1
self.logger.debug('Total changed in this scrape: %s', self.num_changed)
else:
self.logger.debug('No changes to NewsItem %s detected', newsitem.id)
def update(self):
"""
Updates the Schema.last_updated fields after scraping is done.
"""
self.num_added = 0
self.num_changed = 0
update_start = datetime.datetime.now()
# We use a try/finally here so that the DataUpdate object is created
# regardless of whether the scraper raised an exception.
try:
got_error = True
super(NewsItemListDetailScraper, self).update()
got_error = False
finally:
# Rollback, in case the database is in an aborted transaction. his
# avoids the "psycopg2.ProgrammingError: current transaction is aborted,
# commands ignored until end of transaction block" error.
from django.db import connection
connection._rollback()
update_finish = datetime.datetime.now()
# Clear the Schema cache, in case the schemas have been updated in the
# database since we started the scrape.
self._schemas_cache = self._schema_cache = None
for s in self.schemas.values():
s.last_updated = datetime.date.today()
s.save()
DataUpdate.objects.create(
schema=s,
update_start=update_start,
update_finish=update_finish,
num_added=self.num_added,
num_changed=self.num_changed,
# None of our scrapers delete records yet, but we have the
# plumbing in place here in case future scrapers need to do
# that.
num_deleted=0,
num_skipped=self.num_skipped,
got_error=got_error,
)
def geocode(self, location_name):
"""
Tries to geocode the given location string, returning a Point object
or None.
"""
try:
return self._geocoder.geocode(location_name)
except (GeocodingException, ParsingError):
return None
def safe_location(self, location_name, geom, max_distance=200):
"""
Returns a location (geometry) to use, given a location_name and
geometry. This is used for data sources that publish both a geometry
and a location_name -- we double-check that the geometry is within
a certain `max_distance` from the geocoded location_name.
If there's a discrepancy or if the location_name can't be geocoded,
this returns None.
"""
location = self.geocode(location_name)
if location is None:
return None
location_point = location['point']
if not location_point:
return None
location_point.srid = 4326
is_close, distance = locations_are_close(location_point, geom, max_distance)
if not is_close:
return None
return geom
class BaseScraper(object):
"""
Base class for all scrapers in ebdata.retrieval.scrapers.
"""
logname = 'basescraper'
sleep = 0
timeout = 20
def __init__(self, use_cache=True):
if not use_cache:
self.retriever = Retriever(cache=None, sleep=self.sleep, timeout=self.timeout)
else:
self.retriever = Retriever(sleep=self.sleep, timeout=self.timeout)
self.logger = logging.getLogger('eb.retrieval.%s' % self.logname)
self.start_time = datetime.datetime.now()
self._geocoder = SmartGeocoder()
def geocode(self, location_name, zipcode=None):
"""
Tries to geocode the given location string, returning a Point object
or None.
"""
# Try to lookup the adress, if it is ambiguous, attempt to use
# any provided zipcode information to resolve the ambiguity.
# The zipcode is not included in the initial pass because it
# is often too picky yeilding no results when there is a
# legitimate nearby zipcode identified in either the address
# or street number data.
try:
return self._geocoder.geocode(location_name)
except AmbiguousResult as result:
# try to resolve based on zipcode...
if zipcode is None:
self.logger.info(
"Ambiguous results for address %s. (no zipcode to resolve dispute)" %
(location_name, ))
return None
in_zip = [r for r in result.choices if r['zip'] == zipcode]
if len(in_zip) == 0:
self.logger.info(
"Ambiguous results for address %s, but none in specified zipcode %s" %
(location_name, zipcode))
return None
elif len(in_zip) > 1:
self.logger.info(
"Ambiguous results for address %s in zipcode %s, guessing first." %
(location_name, zipcode))
return in_zip[0]
else:
return in_zip[0]
except (GeocodingException, ParsingError):
self.logger.info(
"Could not geocode location: %s: %s" %
(location_name, traceback.format_exc()))
return None
def update(self):
'Run the scraper.'
raise NotImplementedError()
def fetch_data(self, *args, **kwargs):
return self.retriever.fetch_data(*args, **kwargs)
def get_html(self, *args, **kwargs):
"""An alias for fetch_data().
For backward compatibility.
"""
return self.fetch_data(*args, **kwargs)
@classmethod
def parse_html(cls, html):
from lxml import etree
from cStringIO import StringIO
return etree.parse(StringIO(html), etree.HTMLParser())
@transaction.commit_on_success
def create_newsitem(self, attributes, **kwargs):
"""
Creates and saves a NewsItem with the given kwargs. Returns the new
NewsItem.
kwargs MUST have the following keys:
title
item_date
location_name
For any other kwargs whose values aren't provided, this will use
sensible defaults.
kwargs MAY have the following keys:
zipcode - used to disambiguate geocoded locations
kwargs may optionally contain a 'convert_to_block' boolean. If True,
this will convert the given kwargs['location_name'] to a block level
but will use the real (non-block-level) address for geocoding and Block
association.
attributes is a dictionary to use to populate this NewsItem's Attribute
objects.
"""
location = kwargs.get('location')
location_name = kwargs.get('location_name')
assert location or location_name, "At least one of location or location_name must be provided"
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
location = location['point']
if kwargs.pop('convert_to_block', False):
kwargs['location_name'] = address_to_block(kwargs['location_name'])
# If the exact address couldn't be geocoded, try using the
# normalized location name.
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
location = location['point']
# Normally we'd just use "schema = kwargs.get('schema', self.schema)",
# but self.schema will be evaluated even if the key is found in
# kwargs, which raises an error when using multiple schemas.
schema = kwargs.get('schema', None) or self.schema
ni = NewsItem.objects.create(
schema=schema,
title=kwargs['title'],
description=kwargs.get('description', ''),
url=kwargs.get('url', ''),
pub_date=kwargs.get('pub_date', self.start_time),
item_date=kwargs['item_date'],
location=location,
location_name=location_name,
location_object=kwargs.get('location_object', None),
)
if attributes is not None:
ni.attributes = attributes
self.num_added += 1
self.logger.info(u'Created NewsItem %s: %s (total created in this scrape: %s)', schema.slug, ni.id, self.num_added)
return ni
@transaction.commit_on_success
def update_existing(self, newsitem, new_values, new_attributes):
"""
Given an existing NewsItem and dictionaries new_values and
new_attributes, determines which values and attributes have changed
and saves the object and/or its attributes if necessary.
"""
newsitem_updated = False
# First, check the NewsItem's values.
for k, v in new_values.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if getattr(newsitem, k) != v:
self.logger.info('ID %s %s changed from %r to %r' % (newsitem.id, k, getattr(newsitem, k), v))
setattr(newsitem, k, v)
newsitem_updated = True
if newsitem_updated:
newsitem.save()
else:
self.logger.debug("No change to %s <%s>" % (newsitem.id, newsitem))
# Next, check the NewsItem's attributes.
for k, v in new_attributes.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if newsitem.attributes.get(k) == v:
continue
elif k not in newsitem.attributes:
self.logger.info('ID %s %s was missing, setting to %r' %
(newsitem.id, k, v))
elif newsitem.attributes.get(k) != v:
self.logger.info('ID %s %s changed from %r to %r' %
(newsitem.id, k, newsitem.attributes[k], v))
newsitem.attributes[k] = v
newsitem_updated = True
if newsitem_updated:
self.num_changed += 1
self.logger.debug('Total changed in this scrape: %s', self.num_changed)
else:
self.logger.debug('No changes to NewsItem %s detected', newsitem.id)
def create_or_update(self, old_record, attributes, **kwargs):
"""unified API for updating or creating a NewsItem.
"""
if old_record:
self.update_existing(old_record, kwargs, attributes or {})
else:
self.create_newsitem(attributes=attributes, **kwargs)
def update_from_request(self, filter_sf_dict):
"""Update the list of filters based on the request params.
After this is called, it's recommended to redirect to a
normalized form of the URL, which you can get via self.sort();
self.make_url()
``filter_sf_dict`` is a mapping of name -> SchemaField which have
either is_filter or is_searchable True. We remove
SchemaFields that we create filters for. (This is so that
templates can display input widgets for the ones we're not
already filtering by.)
TODO: This should not bail out on the first error,
it should do as much as possible and signal multiple errors.
(Use the forms framework?)
"""
request, context = self.request, self.context
qs = self.qs
params = request.GET.copy()
def pop_key(key, single=False):
"""
Pop the value(s) from params, treat it as a
comma-separated list of values, and split that into a
list. So ?foo=bar,baz is equivalent to ?foo=bar&foo=baz.
If single==True, return only the first one; in the example
we'd return 'bar'. Otherwise, by default, return the
list; in the example we'd return ['bar', 'baz']
"""
result = []
# Doesn't seem to be a way to get a list of values *and*
# remove it in one call; so use both getlist() and pop().
values = params.getlist(key) or [u'']
params.pop(key, None)
for value in values:
value = value.replace(u'+',
u' ') # XXX does django do this already?
values = [s.strip() for s in value.split(u',')]
result.extend(values)
result = [r for r in result if r]
if single:
return result[0] if result else u''
return result
# IDs.
ids = pop_key('id', single=False)
if ids:
self.replace('id', *ids)
# Address.
address = pop_key('address', single=True)
if address:
xy_radius, block_radius, cookies_to_set = block_radius_value(
request)
pop_key(
'radius') # Just to remove it, block_radius_value() used it.
result = None
try:
result = SmartGeocoder().geocode(address)
except AmbiguousResult, e:
raise BadAddressException(address,
block_radius,
address_choices=e.choices)
except (GeocodingException, ParsingError):
raise BadAddressException(address,
block_radius,
address_choices=())
def import_csv_view(self, request):
if not self.has_add_permission(request):
raise PermissionDenied
if request.method == 'GET':
import_form = PlaceImportForm()
if request.method == 'POST':
import_form = PlaceImportForm(request.POST, request.FILES)
if not import_form.is_bound or not import_form.is_valid():
return self._show_import_csv_form(request, import_form)
# csv fields:
# pretty_name, address, lat, lon, url, <synonym>, <synonym>, ...
context = dict(
errors=[],
actions_taken=[],
)
validated_rows = []
place_type = import_form.cleaned_data['place_type']
try:
csvfile = import_form.cleaned_data['csv_file']
rows = csv.reader(csvfile)
except:
message = "Unable to read the specified CSV file"
context['errors'].append(message)
return self._show_import_csv_results(request, context)
try:
for row in rows:
if len(row) < 2:
message = "Line %d: Missing required fields." % rows.line_num
context['errors'].append(message)
continue
synonyms = []
point = None
place_url = ''
pretty_name, address = [x.strip() for x in row[0:2]]
if pretty_name == '':
message = "Line %d: Empty name" % rows.line_num
context['errors'].append(message)
continue
if len(row) > 2:
try:
lat, lon = row[2:4]
if lat != '' or lon != '':
lat = float(lat.strip())
lon = float(lon.strip())
point = geos.Point(lon, lat)
if len(row) > 4:
place_url = row[4]
synonyms = [x.strip() for x in row[5:]]
except ValueError:
message = 'Line %d "%s": Invalid lat, lon' % (
rows.line_num, pretty_name)
context['errors'].append(message)
continue
if point is None:
if address == '':
message = 'Line %d "%s": Address and lat,lon are both empty.' % (
rows.line_num, pretty_name)
context['errors'].append(message)
continue
# try to geocode the address
try:
geocoder = SmartGeocoder()
addr = geocoder.geocode(address)
point = addr['point']
except AmbiguousResult:
message = 'Line %d "%s": Address "%s" is ambiguous, please specify a point directly.' % (
rows.line_num, pretty_name, address)
context['errors'].append(message)
continue
except GeocodingException:
message = 'Line %d "%s": Unable to geocode address "%s", please correct the address or specify a point directly.' % (
rows.line_num, pretty_name, address)
context['errors'].append(message)
continue
# phew!
validated_rows.append(
[pretty_name, address, point, place_url, synonyms])
except csv.Error, e:
message = "Stopped on line %d: %s" % (rows.line_num, e)
context['errors'].append(message)
return self._show_import_csv_results(request, context)
def import_csv_view(self, request):
if not self.has_add_permission(request):
raise PermissionDenied
if request.method == 'GET':
import_form = PlaceImportForm()
if request.method == 'POST':
import_form = PlaceImportForm(request.POST, request.FILES)
if not import_form.is_bound or not import_form.is_valid():
return self._show_import_csv_form(request, import_form)
# csv fields:
# pretty_name, address, lat, lon, url, <synonym>, <synonym>, ...
context = dict(
errors = [],
actions_taken = [],
)
validated_rows = []
place_type = import_form.cleaned_data['place_type']
try:
csvfile = import_form.cleaned_data['csv_file']
rows = csv.reader(csvfile)
except:
message = "Unable to read the specified CSV file"
context['errors'].append(message)
return self._show_import_csv_results(request, context)
try:
for row in rows:
if len(row) < 2:
message = "Line %d: Missing required fields." % rows.line_num
context['errors'].append(message)
continue
synonyms = []
point = None
place_url = ''
pretty_name, address = [x.strip() for x in row[0:2]]
if pretty_name == '':
message = "Line %d: Empty name" % rows.line_num
context['errors'].append(message)
continue
if len(row) > 2:
try:
lat, lon = row[2:4]
if lat != '' or lon != '':
lat = float(lat.strip())
lon = float(lon.strip())
point = geos.Point(lon, lat)
if len(row) > 4:
place_url = row[4]
synonyms = [x.strip() for x in row[5:]]
except ValueError:
message = 'Line %d "%s": Invalid lat, lon' % (rows.line_num, pretty_name)
context['errors'].append(message)
continue
if point is None:
if address == '':
message = 'Line %d "%s": Address and lat,lon are both empty.' % (rows.line_num, pretty_name)
context['errors'].append(message)
continue
# try to geocode the address
try:
geocoder = SmartGeocoder()
addr = geocoder.geocode(address)
point = addr['point']
except AmbiguousResult:
message = 'Line %d "%s": Address "%s" is ambiguous, please specify a point directly.' % (rows.line_num, pretty_name, address)
context['errors'].append(message)
continue
except GeocodingException:
message = 'Line %d "%s": Unable to geocode address "%s", please correct the address or specify a point directly.' % (rows.line_num, pretty_name, address)
context['errors'].append(message)
continue
# phew!
validated_rows.append([pretty_name, address, point, place_url, synonyms])
except csv.Error, e:
message = "Stopped on line %d: %s" % (rows.line_num, e)
context['errors'].append(message)
return self._show_import_csv_results(request, context)
class NewsItemListDetailScraper(ListDetailScraper):
"""
A ListDetailScraper that saves its data into the NewsItem table.
Subclasses are required to set the `schema_slugs` attribute.
Once you've set schema_slugs, there are a number of properties for
conveniently accessing the relevant Schemas and SchemaFields:
self.schemas lazily loads the list of Schema objects the first time it's
accessed. It is a dictionary in the format {slug: Schema}.
self.schema is available if schema_slugs has only one element. It's the
Schema object.
self.lookups lazily loads a dictionary of all SchemaFields with
lookup=True. The dictionary is in the format {name: schemafield}.
If schema_slugs has more than one element, self.lookups is a
dictionary in the format {schema_slug: {name: schemafield}}.
self.schema_fields lazily loads a dictionary of each SchemaField,
mapping the name to the SchemaField object.
If schema_slugs has more than one element, self.schema_fields is a
dictionary in the format {schema_slug: {name: schema_field}}.
self.schema_field_mapping lazily loads a dictionary of each
SchemaField, mapping the name to the real_name.
If schema_slugs has more than one element, self.schema_field_mapping
is a dictionary in the format {schema_slug: {name: real_name}}.
"""
schema_slugs = None
logname = None
def __init__(self, *args, **kwargs):
if self.logname is None:
self.logname = '%s.%s' % (settings.SHORT_NAME, self.schema_slugs[0])
super(NewsItemListDetailScraper, self).__init__(*args, **kwargs)
self._schema_cache = None
self._schemas_cache = None
self._lookups_cache = None
self._schema_fields_cache = None
self._schema_field_mapping_cache = None
self._geocoder = SmartGeocoder()
# schemas, schema, lookups and schema_field_mapping are all lazily loaded
# so that this scraper can be run (in raw_data(), xml_data() or
# display_data()) without requiring a valid database to be set up.
@property
def schemas(self):
if self._schemas_cache is None:
self._schemas_cache = dict([(s, Schema.objects.get(slug=s)) for s in self.schema_slugs])
return self._schemas_cache
@property
def schema(self):
if self._schema_cache is None:
if len(self.schema_slugs) > 1:
raise AttributeError('self.schema is only available if len(schema_slugs) == 1')
self._schema_cache = self.schemas[self.schema_slugs[0]]
return self._schema_cache
@property
def lookups(self):
if self._lookups_cache is None:
lc = dict([(s.slug, dict([(sf.name, sf) for sf in s.schemafield_set.filter(is_lookup=True)])) for s in self.schemas.values()])
if len(self.schema_slugs) == 1:
lc = lc[self.schema_slugs[0]]
self._lookups_cache = lc
return self._lookups_cache
@property
def schema_fields(self):
if self._schema_fields_cache is None:
sfs = dict([(s.slug, dict([(sf.name, sf)
for sf in s.schemafield_set.all()]))
for s in self.schemas.values()])
if len(self.schema_slugs) == 1:
sfs = sfs[self.schema_slugs[0]]
self._schema_fields_cache = sfs
return self._schema_fields_cache
@property
def schema_field_mapping(self):
if self._schema_field_mapping_cache is None:
schema_objs = self.schemas.values()
mapping = field_mapping([s.id for s in schema_objs])
fm = dict([(s.slug, mapping[s.id]) for s in schema_objs])
if len(self.schema_slugs) == 1:
fm = fm[self.schema_slugs[0]]
self._schema_field_mapping_cache = fm
return self._schema_field_mapping_cache
def get_or_create_lookup(self, schema_field_name, name, code, description='', schema=None, make_text_slug=True):
"""
Returns the Lookup instance matching the given Schema slug, SchemaField
name and Lookup.code, creating it (with the given name/code/description)
if it doesn't already exist.
If make_text_slug is True, then a slug will be created from the given
name. If it's False, then the slug will be the Lookup's ID.
"""
if len(self.schema_slugs) > 1:
sf = self.lookups[schema][schema_field_name]
else:
sf = self.lookups[schema_field_name]
return Lookup.objects.get_or_create_lookup(sf, name, code, description, make_text_slug, self.logger)
@transaction.commit_on_success
def create_newsitem(self, attributes, **kwargs):
"""
Creates and saves a NewsItem with the given kwargs. Returns the new
NewsItem.
kwargs MUST have the following keys:
title
item_date
location_name
For any other kwargs whose values aren't provided, this will use
sensible defaults.
kwargs MAY have the following keys:
zipcode - used to disambiguate geocoded locations
kwargs may optionally contain a 'convert_to_block' boolean. If True,
this will convert the given kwargs['location_name'] to a block level
but will use the real (non-block-level) address for geocoding and Block
association.
attributes is a dictionary to use to populate this NewsItem's Attribute
objects.
"""
block = kwargs.get('block')
location = kwargs.get('location')
location_name = kwargs.get('location_name')
assert location or location_name, "At least one of location or location_name must be provided"
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
block = location['block']
location = location['point']
if kwargs.pop('convert_to_block', False):
kwargs['location_name'] = address_to_block(kwargs['location_name'])
# If the exact address couldn't be geocoded, try using the
# normalized location name.
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
block = location['block']
location = location['point']
# Normally we'd just use "schema = kwargs.get('schema', self.schema)",
# but self.schema will be evaluated even if the key is found in
# kwargs, which raises an error when using multiple schemas.
schema = kwargs.get('schema', None) or self.schema
ni = NewsItem.objects.create(
schema=schema,
title=kwargs['title'],
description=kwargs.get('description', ''),
url=kwargs.get('url', ''),
pub_date=kwargs.get('pub_date', self.start_time),
item_date=kwargs['item_date'],
location=location,
location_name=location_name,
location_object=kwargs.get('location_object', None),
block=block,
)
if attributes is not None:
ni.attributes = attributes
self.num_added += 1
self.logger.info(u'Created NewsItem %s: %s (total created in this scrape: %s)', schema.slug, ni.id, self.num_added)
return ni
@transaction.commit_on_success
def update_existing(self, newsitem, new_values, new_attributes):
"""
Given an existing NewsItem and dictionaries new_values and
new_attributes, determines which values and attributes have changed
and saves the object and/or its attributes if necessary.
"""
newsitem_updated = False
# First, check the NewsItem's values.
for k, v in new_values.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if getattr(newsitem, k) != v:
self.logger.info('ID %s %s changed from %r to %r' % (newsitem.id, k, getattr(newsitem, k), v))
setattr(newsitem, k, v)
newsitem_updated = True
if newsitem_updated:
newsitem.save()
else:
self.logger.debug("No change to %s <%s>" % (newsitem.id, newsitem))
# Next, check the NewsItem's attributes.
for k, v in new_attributes.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if newsitem.attributes.get(k) == v:
continue
elif k not in newsitem.attributes:
self.logger.info('ID %s %s was missing, setting to %r' %
(newsitem.id, k, v))
elif newsitem.attributes.get(k) != v:
self.logger.info('ID %s %s changed from %r to %r' %
(newsitem.id, k, newsitem.attributes[k], v))
newsitem.attributes[k] = v
newsitem_updated = True
if newsitem_updated:
self.num_changed += 1
self.logger.debug('Total changed in this scrape: %s', self.num_changed)
else:
self.logger.debug('No changes to NewsItem %s detected', newsitem.id)
def create_or_update(self, old_record, attributes, **kwargs):
"""unified API for updating or creating a NewsItem.
"""
if old_record:
self.update_existing(old_record, kwargs, attributes or {})
else:
self.create_newsitem(attributes=attributes, **kwargs)
def update(self):
"""
Updates the Schema.last_updated fields after scraping is done.
"""
self.num_added = 0
self.num_changed = 0
update_start = datetime.datetime.now()
# We use a try/finally here so that the DataUpdate object is created
# regardless of whether the scraper raised an exception.
try:
got_error = True
super(NewsItemListDetailScraper, self).update()
got_error = False
finally:
# Rollback, in case the database is in an aborted
# transaction. This avoids the "psycopg2.ProgrammingError:
# current transaction is aborted, commands ignored until
# end of transaction block" error.
from django.db import connection
connection._rollback()
update_finish = datetime.datetime.now()
# Clear the Schema cache, in case the schemas have been
# updated in the database since we started the scrape.
self._schemas_cache = self._schema_cache = None
for s in self.schemas.values():
s.last_updated = datetime.date.today()
s.save()
DataUpdate.objects.create(
schema=s,
update_start=update_start,
update_finish=update_finish,
num_added=self.num_added,
num_changed=self.num_changed,
# None of our scrapers delete records yet, but we have the
# plumbing in place here in case future scrapers need to do
# that.
num_deleted=0,
num_skipped=self.num_skipped,
got_error=got_error,
)
def geocode(self, location_name, zipcode=None):
"""
Tries to geocode the given location string, returning a Point object
or None.
"""
# Try to lookup the adress, if it is ambiguous, attempt to use
# any provided zipcode information to resolve the ambiguity.
# The zipcode is not included in the initial pass because it
# is often too picky yeilding no results when there is a
# legitimate nearby zipcode identified in either the address
# or street number data.
try:
return self._geocoder.geocode(location_name)
except AmbiguousResult as result:
# try to resolve based on zipcode...
if zipcode is None:
self.logger.warning("Ambiguous results for address %s. (no zipcode to resolve dispute)" % (location_name, ))
return None
in_zip = [r for r in result.choices if r['zip'] == zipcode]
if len(in_zip) == 0:
self.logger.warning("Ambiguous results for address %s, but none in specified zipcode %s" % (location_name, zipcode))
return None
if len(in_zip) > 1:
self.logger.warning("Ambiguous results for address %s in zipcode %s, guessing first." % (location_name, zipcode))
return in_zip[0]
else:
return in_zip[0]
except (GeocodingException, ParsingError):
self.logger.warning("Could not geocode location: %s: %s" % (location_name, traceback.format_exc()))
return None
def safe_location(self, location_name, geom, max_distance=200):
"""
Returns a location (geometry) to use, given a location_name and
geometry. This is used for data sources that publish both a geometry
and a location_name -- we double-check that the geometry is within
a certain `max_distance` from the geocoded location_name.
If there's a discrepancy or if the location_name can't be geocoded,
this returns None.
"""
location = self.geocode(location_name)
if location is None:
return None
location_point = location['point']
if not location_point:
return None
location_point.srid = 4326
is_close, distance = locations_are_close(location_point, geom, max_distance)
if not is_close:
return None
return geom
def last_updated_time(self, schema=None):
"""
Returns a DateTime representing the last time we started
scraping our schema(s). (We use start time rather than end
time on the assumption that a few overlaps are preferable to
missing updates.)
"""
schema = schema or self.schema
try:
last_update = DataUpdate.objects.order_by('update_start')[0]
return last_update.update_start
except IndexError:
# Use the unix epoch (1970) as a stand-in for "never updated".
return datetime.datetime.fromtimestamp(0)
def save(self, old_record, list_record, detail_record):
# TODO: move some of this to clean_list_record?
date = datetime.date(*list_record['updated_parsed'][:3])
# Get the precinct from the tags.
precincts = [
'A1', 'A7', 'B2', 'B3', 'C11', 'C6', 'D14', 'D4', 'E13', 'E18',
'E5'
]
precinct = None
tags = [t['term'] for t in list_record['tags']]
if not tags:
return
for precinct in tags:
if precinct in precincts:
# TODO: we need a LocationType for precincts, and shapes; and
# then we could set newsitem.location_object to the Location
# for this precinct.
break
if not precinct:
self.logger.debug("no precinct found in tags %r" % tags)
description = list_record['summary']
full_description = list_record['content'][0]['value']
full_description = text_from_html(full_description)
addrs = parse_addresses(full_description)
if not addrs:
self.logger.info("no addresses found in %r %r" %
(list_record['title'], list_record['link']))
return
location = None
location_name = u''
block = None
# This feed doesn't provide geographic data; we'll try to
# extract addresses from the text, and stop on the first
# one that successfully geocodes.
for addr, unused in addrs:
addr = addr.strip()
try:
location = SmartGeocoder().geocode(addr)
except GeocodingException:
log_exception(level=logging.DEBUG)
continue
location_name = location['address']
block = location['block']
location = location['point']
break
else:
self.logger.info("no addresses geocoded in %r" %
list_record['title'])
return
kwargs = dict(
item_date=date,
location=location,
location_name=location_name,
title=list_record['title'],
description=description,
url=list_record['link'],
)
attributes = None
self.create_or_update(old_record, attributes, **kwargs)
def auto_locations(paragraph_list, default_city=''):
"""
Given a list of strings, detects all valid, unique addresses and returns a
tuple (result, report), where result is a list of tuples in the format
(address, wkt, excerpt, block) and report is a string of what happened.
If default_city is given, it will be used in the geocoding for detected
addresses that don't specify a city.
"""
result, report = [], []
addresses_seen = set()
geocoder = SmartGeocoder()
for para in paragraph_list:
for addy, city in parse_addresses(para):
# Skip addresses if they have a city that's a known suburb.
if city and Suburb.objects.filter(
normalized_name=normalize(city)).count():
report.append('got suburb "%s, %s"' % (addy, city))
continue
# Try geocoding the address. If a city was provided, first try
# geocoding with the city, then fall back to just the address
# (without the city).
point = None
attempts = [addy]
if default_city:
attempts.insert(0, '%s, %s' % (addy, default_city))
if city and city.lower() != default_city.lower():
attempts.insert(0, '%s, %s' % (addy, city))
for attempt in attempts:
try:
point = geocoder.geocode(attempt)
break
except AmbiguousResult:
report.append('got ambiguous address "%s"' % attempt)
# Don't try any other address attempts, because they only
# get *more* ambiguous. Plus, the subsequent attempts could
# be incorrect. For example, with this:
# addy = '100 Broadway'
# city = 'Manhattan'
# default_city = 'Brooklyn'
# There are multiple "100 Broadway" addresses in Manhattan,
# so geocoding should fail at this point. It should not
# roll back to try the default_city (Brooklyn).
break
except (DoesNotExist, InvalidBlockButValidStreet):
report.append('got nonexistent address "%s"' % attempt)
except ParsingError:
report.append('got parsing error "%s"' % attempt)
if point is None:
continue # This address could not be geocoded.
if point['address'] in addresses_seen:
continue
if len(para) > 300:
try:
excerpt = smart_excerpt(para, addy)
except ValueError:
excerpt = para
else:
excerpt = para
result.append((addy, point['point'], excerpt, point['block']))
addresses_seen.add(point['address'])
return (result, '; '.join(report))
class BaseScraper(object):
"""
Base class for all scrapers in ebdata.retrieval.scrapers.
"""
logname = 'basescraper'
sleep = 0
timeout = 20
def __init__(self, use_cache=True):
if not use_cache:
self.retriever = Retriever(cache=None, sleep=self.sleep, timeout=self.timeout)
else:
self.retriever = Retriever(sleep=self.sleep, timeout=self.timeout)
self.logger = logging.getLogger('eb.retrieval.%s' % self.logname)
self.start_time = datetime.datetime.now()
self._geocoder = SmartGeocoder()
self.num_added = 0
self.num_changed = 0
def geocode(self, location_name, zipcode=None):
"""
Tries to geocode the given location string, returning a Point object
or None.
"""
# Try to lookup the adress, if it is ambiguous, attempt to use
# any provided zipcode information to resolve the ambiguity.
# The zipcode is not included in the initial pass because it
# is often too picky yeilding no results when there is a
# legitimate nearby zipcode identified in either the address
# or street number data.
try:
return self._geocoder.geocode(location_name)
except AmbiguousResult as result:
# try to resolve based on zipcode...
if zipcode is None:
self.logger.info(
"Ambiguous results for address %s. (no zipcode to resolve dispute)" %
(location_name, ))
return None
in_zip = [r for r in result.choices if r['zip'] == zipcode]
if len(in_zip) == 0:
self.logger.info(
"Ambiguous results for address %s, but none in specified zipcode %s" %
(location_name, zipcode))
return None
elif len(in_zip) > 1:
self.logger.info(
"Ambiguous results for address %s in zipcode %s, guessing first." %
(location_name, zipcode))
return in_zip[0]
else:
return in_zip[0]
except (GeocodingException, ParsingError):
self.logger.info(
"Could not geocode location: %s: %s" %
(location_name, traceback.format_exc()))
return None
def update(self):
'Run the scraper.'
raise NotImplementedError()
def fetch_data(self, *args, **kwargs):
return self.retriever.fetch_data(*args, **kwargs)
def get_html(self, *args, **kwargs):
"""An alias for fetch_data().
For backward compatibility.
"""
return self.fetch_data(*args, **kwargs)
@classmethod
def parse_html(cls, html):
from lxml import etree
from cStringIO import StringIO
return etree.parse(StringIO(html), etree.HTMLParser())
@transaction.commit_on_success
def create_newsitem(self, attributes, **kwargs):
"""
Creates and saves a NewsItem with the given kwargs. Returns the new
NewsItem.
kwargs MUST have the following keys:
title
item_date
location_name
For any other kwargs whose values aren't provided, this will use
sensible defaults.
kwargs MAY have the following keys:
zipcode - used to disambiguate geocoded locations
kwargs may optionally contain a 'convert_to_block' boolean. If True,
this will convert the given kwargs['location_name'] to a block level
but will use the real (non-block-level) address for geocoding and Block
association.
attributes is a dictionary to use to populate this NewsItem's Attribute
objects.
"""
location = kwargs.get('location')
location_name = kwargs.get('location_name')
assert location or location_name, "At least one of location or location_name must be provided"
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
location = location['point']
if kwargs.pop('convert_to_block', False):
kwargs['location_name'] = address_to_block(kwargs['location_name'])
# If the exact address couldn't be geocoded, try using the
# normalized location name.
if location is None:
location = self.geocode(kwargs['location_name'], zipcode=kwargs.get('zipcode'))
if location:
location = location['point']
# Normally we'd just use "schema = kwargs.get('schema', self.schema)",
# but self.schema will be evaluated even if the key is found in
# kwargs, which raises an error when using multiple schemas.
schema = kwargs.get('schema', None) or self.schema
ni = NewsItem.objects.create(
schema=schema,
title=kwargs['title'],
description=kwargs.get('description', ''),
url=kwargs.get('url', ''),
pub_date=kwargs.get('pub_date', self.start_time),
item_date=kwargs['item_date'],
location=location,
location_name=location_name,
location_object=kwargs.get('location_object', None),
)
if attributes is not None:
ni.attributes = attributes
self.num_added += 1
self.logger.info(u'Created NewsItem %s: %s (total created in this scrape: %s)', schema.slug, ni.id, self.num_added)
return ni
@transaction.commit_on_success
def update_existing(self, newsitem, new_values, new_attributes):
"""
Given an existing NewsItem and dictionaries new_values and
new_attributes, determines which values and attributes have changed
and saves the object and/or its attributes if necessary.
"""
newsitem_updated = False
# First, check the NewsItem's values.
for k, v in new_values.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if getattr(newsitem, k) != v:
self.logger.info('ID %s %s changed from %r to %r' % (newsitem.id, k, getattr(newsitem, k), v))
setattr(newsitem, k, v)
newsitem_updated = True
if newsitem_updated:
newsitem.save()
else:
self.logger.debug("No change to %s <%s>" % (newsitem.id, newsitem))
# Next, check the NewsItem's attributes.
for k, v in new_attributes.items():
if isinstance(v, datetime.datetime) and v.tzinfo is not None:
# Django datetime fields are not timezone-aware, so we
# can't compare them without stripping the zone.
v = v.astimezone(local_tz).replace(tzinfo=None)
if newsitem.attributes.get(k) == v:
continue
elif k not in newsitem.attributes:
self.logger.info('ID %s %s was missing, setting to %r' %
(newsitem.id, k, v))
elif newsitem.attributes.get(k) != v:
self.logger.info('ID %s %s changed from %r to %r' %
(newsitem.id, k, newsitem.attributes[k], v))
newsitem.attributes[k] = v
newsitem_updated = True
if newsitem_updated:
self.num_changed += 1
self.logger.debug('Total changed in this scrape: %s', self.num_changed)
else:
self.logger.debug('No changes to NewsItem %s detected', newsitem.id)
def create_or_update(self, old_record, attributes, **kwargs):
"""unified API for updating or creating a NewsItem.
"""
if old_record:
self.update_existing(old_record, kwargs, attributes or {})
else:
self.create_newsitem(attributes=attributes, **kwargs)
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
def setUp(self):
self.geocoder = SmartGeocoder(use_cache=False)
def auto_locations(paragraph_list, default_city=''):
"""
Given a list of strings, detects all valid, unique addresses and returns a
tuple (result, report), where result is a list of tuples in the format
(address, wkt, excerpt, block) and report is a string of what happened.
If default_city is given, it will be used in the geocoding for detected
addresses that don't specify a city.
"""
result, report = [], []
addresses_seen = set()
geocoder = SmartGeocoder()
for para in paragraph_list:
for addy, city in parse_addresses(para):
# Skip addresses if they have a city that's a known suburb.
if city and Suburb.objects.filter(normalized_name=normalize(city)).count():
report.append('got suburb "%s, %s"' % (addy, city))
continue
# Try geocoding the address. If a city was provided, first try
# geocoding with the city, then fall back to just the address
# (without the city).
point = None
attempts = [addy]
if default_city:
attempts.insert(0, '%s, %s' % (addy, default_city))
if city and city.lower() != default_city.lower():
attempts.insert(0, '%s, %s' % (addy, city))
for attempt in attempts:
try:
point = geocoder.geocode(attempt)
break
except AmbiguousResult:
report.append('got ambiguous address "%s"' % attempt)
# Don't try any other address attempts, because they only
# get *more* ambiguous. Plus, the subsequent attempts could
# be incorrect. For example, with this:
# addy = '100 Broadway'
# city = 'Manhattan'
# default_city = 'Brooklyn'
# There are multiple "100 Broadway" addresses in Manhattan,
# so geocoding should fail at this point. It should not
# roll back to try the default_city (Brooklyn).
break
except (DoesNotExist, InvalidBlockButValidStreet):
report.append('got nonexistent address "%s"' % attempt)
except ParsingError:
report.append('got parsing error "%s"' % attempt)
if point is None:
continue # This address could not be geocoded.
if point['address'] in addresses_seen:
continue
if len(para) > 300:
try:
excerpt = smart_excerpt(para, addy)
except ValueError:
excerpt = para
else:
excerpt = para
result.append((addy, point['point'], excerpt, point['block']))
addresses_seen.add(point['address'])
return (result, '; '.join(report))
def update(self):
logger.info("Starting LocalNewsScraper update %s" % self.url)
try:
schema = Schema.objects.get(slug=self.schema_slug)
except Schema.DoesNotExist:
logger.error("Schema (%s): DoesNotExist" % self.schema_slug)
return 1
response, content = self.http.request(self.url)
if response.fromcache:
logger.info("Feed is unchanged since last update (cached)")
return
f = feedparser.parse(content)
addcount = updatecount = 0
for entry in f.entries:
title = convert_entities(entry.title)
description = convert_entities(entry.description)
if entry.id.startswith('http'):
item_url = entry.id
else:
item_url = entry.link
try:
item = NewsItem.objects.get(schema__id=schema.id,
title=title,
description=description)
#url=item_url)
status = 'updated'
except NewsItem.DoesNotExist:
item = NewsItem()
status = 'added'
except NewsItem.MultipleObjectsReturned:
# Seen some where we get the same story with multiple URLs. Why?
logger.warn(
"Multiple entries matched title %r and description %r. Expected unique!"
% (title, description))
continue
try:
item.title = title
item.schema = schema
item.description = description
item.url = item_url
# Support both georss and xcal for getting the location name.
# TODO: should also support ev:location per http://web.resource.org/rss/1.0/modules/event/
item.location_name = entry.get(
'xCal_x-calconnect-street') or entry.get(
'x-calconnect-street') or entry.get(
'georss_featurename') or entry.get('featurename')
item.item_date = datetime.datetime(*entry.updated_parsed[:6])
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
_short_title = item.title[:30] + '...'
# feedparser bug: depending on which parser it magically uses,
# we either get the xml namespace in the key name, or we don't.
point = entry.get('georss_point') or entry.get('point')
x, y = None, None
if point:
# GeoRSS puts latitude (Y) first.
y, x = point.split(' ')
else:
if item.location_name:
text = item.location_name
else:
# Geocode whatever we can find.
text = item.title + ' ' + item.description
logger.debug("...Falling back on geocoding from %r..." %
text[:50])
addrs = parse_addresses(text)
for addr, unused in addrs:
try:
result = SmartGeocoder().geocode(addr)
point = result['point']
logger.debug("internally geocoded %r" % addr)
x, y = point.x, point.y
if not item.location_name:
item.location_name = result['address']
break
except GeocodingException:
logger.debug("Geocoding exception on %r:" % text)
log_exception(level=logging.DEBUG)
continue
except:
logger.error(
'uncaught geocoder exception on %r\n' % addr)
log_exception()
if None in (x, y):
logger.debug(
"Skip, couldn't geocode any addresses in item '%s...'"
% _short_title)
continue
item.location = Point((float(x), float(y)))
if not intersects_metro_bbox(item.location):
reversed_loc = Point((float(y), float(x)))
if intersects_metro_bbox(reversed_loc):
logger.info(
"Got points in apparently reverse order, flipping them"
)
item.location = reversed_loc
else:
logger.info("Skipping %r as %s,%s is out of bounds" %
(_short_title, y, x))
continue
if not item.location_name:
# Fall back to reverse-geocoding.
from ebpub.geocoder import reverse
try:
block, distance = reverse.reverse_geocode(
item.location)
logger.debug(" Reverse-geocoded point to %r" %
block.pretty_name)
item.location_name = block.pretty_name
except reverse.ReverseGeocodeError:
logger.info(
" Skip, failed to reverse geocode %s for %r" %
(item.location.wkt, _short_title))
continue
item.save()
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, _short_title))
except:
logger.error("Warning: couldn't save %r. Traceback:" %
_short_title)
log_exception()
logger.info("Finished LocalNewsScraper update: %d added, %d updated" %
(addcount, updatecount))
def update(argv=None):
logger.info("Starting add_news")
if argv:
url = argv[0]
else:
url = 'http://search.boston.com/search/api?q=*&sort=-articleprintpublicationdate&subject=massachusetts&scope=bonzai'
schema_slug = 'local-news'
try:
schema = Schema.objects.get(slug=schema_slug)
except Schema.DoesNotExist:
logger.error("Schema (%s): DoesNotExist" % schema_slug)
sys.exit(1)
f = feedparser.parse(url)
addcount = updatecount = 0
for entry in f.entries:
title = convert_entities(entry.title)
description = convert_entities(entry.description)
if entry.id.startswith('http'):
item_url = entry.id
else:
item_url = entry.link
try:
item = NewsItem.objects.get(schema__id=schema.id,
title=title,
description=description)
#url=item_url)
status = 'updated'
except NewsItem.DoesNotExist:
item = NewsItem()
status = 'added'
except NewsItem.MultipleObjectsReturned:
# Seen some where we get the same story with multiple URLs. Why?
logger.warn(
"Multiple entries matched title %r and description %r. Expected unique!"
% (title, description))
continue
try:
item.title = title
item.schema = schema
item.description = description
item.url = item_url
item.location_name = entry.get('x-calconnect-street') or entry.get(
'georss_featurename')
item.item_date = datetime.datetime(*entry.updated_parsed[:6])
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
# feedparser bug: depending on which parser it magically uses,
# we either get the xml namespace in the key name, or we don't.
point = entry.get('georss_point') or entry.get('point')
x, y = None, None
if point:
x, y = point.split(' ')
if True:
# Fall back on geocoding.
text = item.title + ' ' + item.description
addrs = parse_addresses(text)
for addr, unused in addrs:
try:
result = SmartGeocoder().geocode(addr)
point = result['point']
logger.debug("internally geocoded %r" % addr)
x, y = point.x, point.y
break
except (GeocodingException, ParsingError):
logger.debug("Geocoding exception on %r:" % text)
log_exception(level=logging.DEBUG)
continue
except:
logger.error('uncaught geocoder exception on %r\n' %
addr)
log_exception()
if None in (x, y):
logger.info("couldn't geocode '%s...'" % item.title[:30])
continue
item.location = Point((float(y), float(x)))
if item.location.x == 0.0 and item.location.y == 0.0:
# There's a lot of these. Maybe attempt to
# parse and geocode if we haven't already?
logger.info("Skipping %r as it has bad location 0,0" %
item.title)
continue
if not item.location_name:
# Fall back to reverse-geocoding.
from ebpub.geocoder import reverse
try:
block, distance = reverse.reverse_geocode(item.location)
logger.debug(" Reverse-geocoded point to %r" %
block.pretty_name)
item.location_name = block.pretty_name
except reverse.ReverseGeocodeError:
logger.debug(" Failed to reverse geocode %s for %r" %
(item.location.wkt, item.title))
item.location_name = u''
item.save()
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except:
logger.error("Warning: couldn't save %r. Traceback:" % item.title)
log_exception()
logger.info("Finished add_news: %d added, %d updated" %
(addcount, updatecount))
