def clean_list_record(self, record):
record.title = convert_entities(record['title'])
record.description = convert_entities(record['description'])
# Don't know why, but some feeds have 'id' *instead* of 'link'.
if record.get('id', '').startswith('http'):
record['link'] = record['id']
# This tries GeoRSS, RDF Geo, xCal, ...
point, location_name = self.get_point_and_location_name(record)
_short_title = record['title'][:30] + '...'
if not point:
raise SkipRecord("couldn't geocode any addresses in item '%s...'"
% _short_title)
if not location_name:
raise SkipRecord(
"Skip, no location name and failed to reverse geocode %s for %r" % (point.wkt, _short_title))
if not intersects_metro_bbox(point):
# Check if latitude, longitude seem to be reversed; I've
# seen that in some bad feeds!
reversed_loc = Point(point.y, point.x)
if intersects_metro_bbox(reversed_loc):
self.logger.info(
"Got points in apparently reverse order, flipping them")
point = reversed_loc
else:
raise SkipRecord("Skipping %r as %s,%s is out of bounds" %
(_short_title, point.y, point.x))
record['location_name'] = location_name
record['location'] = point
return record
def main():
""" Download Calendar RSS feed and update database """
url = """http://calendar.boston.com/search?acat=&cat=&commit=Search\
&new=n&rss=1&search=true&sort=0&srad=20&srss=50&ssrss=5&st=event\
&st_select=any&svt=text&swhat=&swhen=today&swhere=&trim=1"""
schema = 'events'
parser = OptionParser()
parser.add_option('-q', '--quiet', action="store_true", dest="quiet",
default=False, help="no output")
(options, args) = parser.parse_args()
if len(args) > 0:
return parser.error('script does not take any arguments')
try:
schema = Schema.objects.get(slug=schema)
except Schema.DoesNotExist:
print "Schema (%s): DoesNotExist" % schema
sys.exit(1)
feed = feedparser.parse(url)
for entry in feed.entries:
try:
item = NewsItem.objects.get(title=entry.title,
description=entry.description)
status = "Updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "Added"
try:
item.location_name = entry.get('xcal_x-calconnect-street')
item.schema = schema
item.title = convert_entities(entry.title)
item.description = convert_entities(entry.description)
item.url = entry.link
item.item_date = datetime.datetime(*entry.updated_parsed[:6])
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.location = Point((float(entry['geo_long']),
float(entry['geo_lat'])))
if (item.location.x, item.location.y) == (0.0, 0.0):
print "Skipping %r, bad location 0,0" % item.title
continue
item.save()
if not options.quiet:
print "%s: %s" % (status, item.title)
except ValueError:
if not options.quiet:
print "unexpected error:", sys.exc_info()[1]
def text_from_html(html):
"""Remove ALL tags and return all plain text.
"""
text = preprocess_to_string(html, drop_tags=_html_droptags,
drop_trees=_html_droptrees)
text = convert_entities(text)
return text
def clean_detail_record(self, record):
if record == {}:
# Parsing the detail page failed.
return None
if record['violation_points'].startswith('Not Available'):
record['violation_points'] = 'N/A'
record['followup_inspection'] = False
else:
if not record['violation_points'].isdigit():
raise ScraperBroken('Got odd violation points value %r' %
record['violation_points'])
record['followup_inspection'] = int(
record['violation_points']) > 27
# Parse the violations from the HTML chunk. When we're done,
# record['violation_list'] will be a (possibly empty) list of strings.
vio_chunk = record.pop('violations')
if vio_chunk == '':
record['violation_list'] = []
else:
vios = violation_list_re.findall(vio_chunk)
if not vios:
raise ScraperBroken(
"Violation data not found for restaurant %s",
record['restaurant_name'])
record['violation_list'] = [
strip_tags(convert_entities(v.strip())) for v in vios
]
# Remove the ZIP code from the address, as it complicates geocoding.
record['address'] = re.sub(r'\s*\d{5}\s*$', '', record['address'])
# Strip extra internal whitespace.
record['address'] = re.sub(r'\s+', ' ', record['address'])
return record
def clean_detail_record(self, record):
if record == {}:
# Parsing the detail page failed.
return None
if record['violation_points'].startswith('Not Available'):
record['violation_points'] = 'N/A'
record['followup_inspection'] = False
else:
if not record['violation_points'].isdigit():
raise ScraperBroken('Got odd violation points value %r' % record['violation_points'])
record['followup_inspection'] = int(record['violation_points']) > 27
# Parse the violations from the HTML chunk. When we're done,
# record['violation_list'] will be a (possibly empty) list of strings.
vio_chunk = record.pop('violations')
if vio_chunk == '':
record['violation_list'] = []
else:
vios = violation_list_re.findall(vio_chunk)
if not vios:
raise ScraperBroken("Violation data not found for restaurant %s", record['restaurant_name'])
record['violation_list'] = [strip_tags(convert_entities(v.strip())) for v in vios]
# Remove the ZIP code from the address, as it complicates geocoding.
record['address'] = re.sub(r'\s*\d{5}\s*$', '', record['address'])
# Strip extra internal whitespace.
record['address'] = re.sub(r'\s+', ' ', record['address'])
return record
def text_from_html(html):
"""Remove ALL tags and return all plain text.
"""
text = preprocess_to_string(html,
drop_tags=_html_droptags,
drop_trees=_html_droptrees)
text = convert_entities(text)
return text
def clean_list_record(self, record):
# clean up a record dict
# Item date, in timezone of the photo owner.
# Not sure how to determine what that is, so we'll leave it.
cleaned = {}
cleaned['item_date'] = datetime.datetime.strptime(record['datetaken'],
'%Y-%m-%d %H:%M:%S')
cleaned['item_date'] = cleaned['item_date'].date()
# Posted date, UTC timestamp.
pub_date = datetime.datetime.fromtimestamp(
float(record['dateupload']), utc)
cleaned['pub_date'] = pub_date.astimezone(local_tz)
description = record['description']['_content']
cleaned['description'] = convert_entities(description.strip())
cleaned['title'] = convert_entities(record['title'])
x, y = record['longitude'], record['latitude']
cleaned['location'] = Point((float(x), float(y)))
# Possibly we could figure out flickr's geo API and resolve
# the photo's place_id and/or woeid to the place name? But
# those are probably not specific enough; reverse-geocode
# instead.
try:
block, distance = reverse_geocode(cleaned['location'])
cleaned['location_name'] = block.pretty_name
except ReverseGeocodeError:
raise SkipRecord("Could not geocode location %s, %s" % (x, y))
# Don't think any of the urls returned by the API's "extras"
# correspond to the page? not sure.
cleaned['url'] = 'http://www.flickr.com/photos/%(owner)s/%(id)s' % record
attributes = {}
attributes['sourcename'] = 'Flickr'
#attributes['photo_id'] = record['id']
attributes['user_id'] = record['owner']
attributes['username'] = record['ownername']
# Thumbnail. 'Small square' photos are 75x75.
attributes['photo_href'] = record['url_m']
cleaned['_attributes'] = attributes
return cleaned
def clean_list_record(self, record):
# clean up a record dict
# Item date, in timezone of the photo owner.
# Not sure how to determine what that is, so we'll leave it.
cleaned = {}
cleaned['item_date'] = datetime.datetime.strptime(
record['datetaken'], '%Y-%m-%d %H:%M:%S')
cleaned['item_date'] = cleaned['item_date'].date()
# Posted date, UTC timestamp.
pub_date = datetime.datetime.fromtimestamp(float(record['dateupload']),
utc)
cleaned['pub_date'] = pub_date.astimezone(local_tz)
description = record['description']['_content']
cleaned['description'] = convert_entities(description.strip())
cleaned['title'] = convert_entities(record['title'])
x, y = record['longitude'], record['latitude']
cleaned['location'] = Point((float(x), float(y)))
# Possibly we could figure out flickr's geo API and resolve
# the photo's place_id and/or woeid to the place name? But
# those are probably not specific enough; reverse-geocode
# instead.
try:
block, distance = reverse_geocode(cleaned['location'])
cleaned['location_name'] = block.pretty_name
except ReverseGeocodeError:
raise SkipRecord("Could not geocode location %s, %s" % (x, y))
# Don't think any of the urls returned by the API's "extras"
# correspond to the page? not sure.
cleaned[
'url'] = 'http://www.flickr.com/photos/%(owner)s/%(id)s' % record
attributes = {}
attributes['sourcename'] = 'Flickr'
#attributes['photo_id'] = record['id']
attributes['user_id'] = record['owner']
attributes['username'] = record['ownername']
# Thumbnail. 'Small square' photos are 75x75.
attributes['photo_href'] = record['url_sq']
cleaned['_attributes'] = attributes
return cleaned
def html_to_paragraph_list(tree):
"""
Given an HTML tree, removes HTML tags and returns a list of strings, with
each string representing a paragraph/block.
"""
block_tags = set([
'blockquote', 'dd', 'div', 'dt', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
'h7', 'h8', 'li', 'p', 'td', 'th', 'tr'
])
drop_tags_only = set([
'a', 'abbr', 'acronym', 'b', 'center', 'dir', 'dl', 'em', 'font',
'form', 'hr', 'i', 'label', 'menu', 'ol', 'pre', 'small', 'span',
'strong', 'sub', 'sup', 'table', 'tbody', 'tfoot', 'thead', 'topic',
'u', 'ul', 'wbr'
])
drop_tags_and_contents = set([
'applet', 'area', 'button', 'embed', 'img', 'iframe', 'head', 'input',
'link', 'map', 'meta', 'noscript', 'object', 'option', 'script',
'select', 'spacer', 'style', 'textarea', 'title'
])
elements_to_drop = []
for element in tree.getiterator():
if not isinstance(element.tag, basestring): # If it's a comment...
element.drop_tag()
continue
if element.text and '\n' in element.text:
element.text = element.text.replace('\n', ' ')
if element.tail and '\n' in element.tail:
element.tail = element.tail.replace('\n', ' ')
if element.tag in block_tags:
element.text = '\n' + (element.text or '')
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag == 'br':
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag in drop_tags_only:
element.drop_tag()
elif element.tag in drop_tags_and_contents:
elements_to_drop.append(element)
elif element.tag not in ('html', 'body'): # Unknown tag!
element.drop_tag()
for e in elements_to_drop:
e.drop_tree()
try:
tree.body
except IndexError:
return ''
else:
new_html = etree.tostring(
tree.body, method='html')[6:-7] # strip <body> and </body>
new_html = convert_entities(new_html)
return re.split(r'\s*\n+\s*', new_html.strip())
def text_from_html(html):
"""Remove ALL tags and return all plain text.
"""
text = preprocess_to_string(html, drop_tags=_html_droptags,
drop_trees=_html_droptrees)
if not text:
# Maybe there was something there but not really HTML.
if text and not isinstance(text, unicode):
text = UnicodeDammit(html, isHTML=False).unicode.strip()
else:
text = u''
text = convert_entities(text)
return text
def text_from_html(html):
"""Remove ALL tags and return all plain text.
"""
text = preprocess_to_string(html,
drop_tags=_html_droptags,
drop_trees=_html_droptrees)
if not text:
# Maybe there was something there but not really HTML.
if text and not isinstance(text, unicode):
text = UnicodeDammit(html, isHTML=False).unicode.strip()
else:
text = u''
text = convert_entities(text)
return text
def geocode_if_needed(self,
point,
location_name,
address_text='',
**kwargs):
"""
If either ``point`` or ``location_name`` is not set, try to
geocode / reverse-geocode as needed to derive one from the
other. Returns (point, location_name).
If neither one is set, try to parse addresses out of
``address_text`` and derive both.
Either value may be None if it can't be determined.
Any other keyword args are passed to ``full_geocode()``.
"""
if not point:
text = convert_entities(location_name or address_text)
self.logger.debug("...Falling back on geocoding from '%s...'" %
text[:50])
addrs = parse_addresses(text)
for addr, unused in addrs:
try:
result = self.geocode(addr, **kwargs)
if result is not None:
point = result['point']
self.logger.debug("internally geocoded %r" % addr)
# TODO: what if it's a Place?
if not location_name:
location_name = result['address']
break
except:
self.logger.exception(
'uncaught geocoder exception on %r\n' % addr)
continue
if point and not location_name:
# Fall back to reverse-geocoding.
from ebpub.geocoder import reverse
try:
block, distance = reverse.reverse_geocode(point)
self.logger.debug(" Reverse-geocoded point to %r" %
block.pretty_name)
location_name = block.pretty_name
except reverse.ReverseGeocodeError:
location_name = None
return (point, location_name)
def clean_list_record(self, record):
record.title = convert_entities(record['title'])
record.description = convert_entities(record['description'])
# Don't know why, but some feeds have 'id' *instead* of 'link'.
if record.get('id', '').startswith('http'):
record['link'] = record['id']
# This tries GeoRSS, RDF Geo, xCal, ...
point, location_name = self.get_point_and_location_name(record)
_short_title = record['title'][:30] + '...'
if not point:
raise SkipRecord("couldn't geocode any addresses in item '%s...'" %
_short_title)
if not location_name:
raise SkipRecord(
"Skip, no location name and failed to reverse geocode %s for %r"
% (point.wkt, _short_title))
if not intersects_metro_bbox(point):
# Check if latitude, longitude seem to be reversed; I've
# seen that in some bad feeds!
reversed_loc = Point(point.y, point.x)
if intersects_metro_bbox(reversed_loc):
self.logger.info(
"Got points in apparently reverse order, flipping them")
point = reversed_loc
else:
raise SkipRecord("Skipping %r as %s,%s is out of bounds" %
(_short_title, point.y, point.x))
record['location_name'] = location_name
record['location'] = point
return record
def update(self):
""" Download Calendar RSS feed and update database """
logger.info("Starting ObituaryScraper")
feed = feedparser.parse(self.url)
total_created = 0
for entry in feed.entries:
title = convert_entities(entry.title)
try:
created = self.parse_entry(entry, title)
if created:
total_created += 1
except:
logger.error("unexpected error:", sys.exc_info()[1])
log_exception()
break
logger.info("Created %d of %d total" % (created, len(feed.entries)))
def geocode_if_needed(self, point, location_name, address_text='',
**kwargs):
"""
If either ``point`` or ``location_name`` is not set, try to
geocode / reverse-geocode as needed to derive one from the
other. Returns (point, location_name).
If neither one is set, try to parse addresses out of
``address_text`` and derive both.
Either value may be None if it can't be determined.
Any other keyword args are passed to ``full_geocode()``.
"""
if not point:
text = convert_entities(location_name or address_text)
self.logger.debug("...Falling back on geocoding from '%s...'" % text[:50])
addrs = parse_addresses(text)
for addr, unused in addrs:
try:
result = self.geocode(addr, **kwargs)
if result is not None:
point = result['point']
self.logger.debug("internally geocoded %r" % addr)
# TODO: what if it's a Place?
if not location_name:
location_name = result['address']
break
except:
self.logger.exception('uncaught geocoder exception on %r\n' % addr)
continue
if point and not location_name:
# Fall back to reverse-geocoding.
from ebpub.geocoder import reverse
try:
block, distance = reverse.reverse_geocode(point)
self.logger.debug(" Reverse-geocoded point to %r" % block.pretty_name)
location_name = block.pretty_name
except reverse.ReverseGeocodeError:
location_name = None
return (point, location_name)
def html_to_paragraph_list(tree):
"""
Given an HTML tree, removes HTML tags and returns a list of strings, with
each string representing a paragraph/block.
"""
block_tags = set(['blockquote', 'dd', 'div', 'dt', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'li', 'p', 'td', 'th', 'tr'])
drop_tags_only = set(['a', 'abbr', 'acronym', 'b', 'center', 'dir', 'dl', 'em', 'font', 'form', 'hr', 'i', 'label', 'menu', 'ol', 'pre', 'small', 'span', 'strong', 'sub', 'sup', 'table', 'tbody', 'tfoot', 'thead', 'topic', 'u', 'ul', 'wbr'])
drop_tags_and_contents = set(['applet', 'area', 'button', 'embed', 'img', 'iframe', 'head', 'input', 'link', 'map', 'meta', 'noscript', 'object', 'option', 'script', 'select', 'spacer', 'style', 'textarea', 'title'])
elements_to_drop = []
for element in tree.getiterator():
if not isinstance(element.tag, basestring): # If it's a comment...
element.drop_tag()
continue
if element.text and '\n' in element.text:
element.text = element.text.replace('\n', ' ')
if element.tail and '\n' in element.tail:
element.tail = element.tail.replace('\n', ' ')
if element.tag in block_tags:
element.text = '\n' + (element.text or '')
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag == 'br':
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag in drop_tags_only:
element.drop_tag()
elif element.tag in drop_tags_and_contents:
elements_to_drop.append(element)
elif element.tag not in ('html', 'body'): # Unknown tag!
element.drop_tag()
for e in elements_to_drop:
e.drop_tree()
try:
tree.body
except IndexError:
return ''
else:
new_html = etree.tostring(tree.body, method='html')[6:-7] # strip <body> and </body>
new_html = convert_entities(new_html)
return re.split(r'\s*\n+\s*', new_html.strip())
def parse_entry(self, entry, title):
try:
item = NewsItem.objects.get(title=title, schema__id=self.schema.id)
except NewsItem.DoesNotExist:
item = NewsItem(title=title, schema=self.schema)
description = convert_entities(entry.description)
try:
location, description = description.split(' -- ', 1)
except ValueError:
logger.error("Unable to parse description: %s", description)
return
item.url = entry.link
item.description = description
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
try:
item.location_name = self.geocoder.geocode(location)
except geocoder.DoesNotExist:
logger.error("Failed to geocode %s" % location)
item.location_name = location
created = item.pk is not None
item.save()
return created
def update():
""" Download Calendar RSS feed and update database """
logger.info("Starting add_events")
url = """http://calendar.boston.com/search?acat=&cat=&commit=Search\
&new=n&rss=1&search=true&sort=0&srad=20&srss=50&ssrss=5&st=event\
&st_select=any&svt=text&swhat=&swhen=today&swhere=&trim=1"""
schema = 'events'
try:
schema = Schema.objects.get(slug=schema)
except Schema.DoesNotExist:
logger.error("Schema (%s): DoesNotExist" % schema)
sys.exit(1)
feed = feedparser.parse(url)
addcount = updatecount = 0
for entry in feed.entries:
title = convert_entities(entry.title).strip()
# Putting 'event' in the title is redundant, ticket #227
if title.lower().startswith('event: '):
title = title[7:]
try:
item = NewsItem.objects.get(title=title,
schema__id=schema.id)
status = "updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "added"
except NewsItem.MultipleObjectsReturned:
logger.warn("Multiple entries matched title %r, event titles are not unique?" % title)
continue
try:
item.location_name = entry.get('xcal_x-calconnect-street') or entry.get('x-calconnect-street') or u''
item.schema = schema
item.title = title
item.description = convert_entities(entry.description)
item.url = entry.link
item.item_date = datetime.datetime(*entry.updated_parsed[:6])
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.location = Point((float(entry['geo_long']),
float(entry['geo_lat'])))
if (item.location.x, item.location.y) == (0.0, 0.0):
logger.warn("Skipping %r, 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.info(" 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.exception("unexpected error:", sys.exc_info()[1])
logger.info("add_events finished: %d added, %d updated" % (addcount, updatecount))
def unique_fields(self, list_record):
# not necessarily primary key, but for this script's purposes
# these are the fields that in combination uniquely idenfity
# an article.
date = datetime.date(*list_record['updated_parsed'][:3])
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 can set newsitem.location_object to the Location
# for this precinct.
break
if not precinct:
self.logger.debug("no precinct found in tags %r" % tags)
if 'Boston 24' in tags:
# TODO: the 'Boston 24' tag indicates posts with aggregate
# daily stats. Make a separate schema for aggregates,
# with attributes like those used in
# everyblock/everyblock/cities/nyc/crime_aggregate/retrieval.py.
# These are citywide though, not by precinct.
# So what would be the Location? Whole city??
self.logger.info("boston daily crime stats, we don't know how to "
"handle these yet")
description = list_record['content'][0]['value']
# TODO: we should have a stock 'clean up html' function.
description = preprocess_to_string(
description,
drop_tags=('a', 'area', 'b', 'center', 'font', 'form', 'img', 'input', 'p', 'strong', 'map', 'small', 'span', 'sub', 'sup', 'topic', 'u'),
drop_trees=('applet', 'button', 'embed', 'iframe', 'object', 'select', 'textarea'),
drop_attrs=('background', 'border', 'cellpadding', 'cellspacing', 'class', 'clear', 'id', 'rel', 'style', 'target'))
from ebdata.retrieval.utils import convert_entities
description = convert_entities(description)
#description = description.replace(' ', ' ').replace(' ', ' ')
addrs = parse_addresses(description)
if not addrs:
self.logger.info("no addresses found in %r" % list_record['title'])
location = None
location_name = u''
for addr, unused in addrs:
addr = addr.strip()
try:
from geocoder_hack import quick_dirty_fallback_geocode
x, y = quick_dirty_fallback_geocode(addr)
if (x, y) != (None, None):
location = Point((float(x), float(y)))
location_name = addr.title()
except:
print "ugh, %r" % addr
# XXX log something
return dict(item_date=date,
location=location,
location_name=location_name,
title=list_record['title'],
description=description,
)
def update(url):
logger.info("Scraping police reports")
schema_slug = 'police'
try:
schema = Schema.objects.get(slug=schema_slug)
except Schema.DoesNotExist:
logger.error( "Schema (%s): DoesNotExist" % schema_slug)
sys.exit(1)
incident_type_field = SchemaField.objects.get(schema=schema, name='incident_type')
f = feedparser.parse(url)
addcount = updatecount = 0
for entry in f.entries:
title = convert_entities(entry.title).strip()
# The title will be used as the incident type.
if title in SKIP_TYPES:
logger.info("Skipping entry of type %s" % title)
description = convert_entities(entry.summary)
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.pub_date = datetime(*entry.updated_parsed[:6])
item.location = Point((float(entry.geo_long), float(entry.geo_lat)))
item.location_name = get_element(entry, 'address')
# parse call time
ct = datetime.strptime(get_element(entry, 'calldatetime'),
r"%m/%d/%Y %I:%M:%S %p")
#ct = datetime(ct.year, ct.month, ct.day, ct.hour, ct.minute, ct.second, tzinfo=tzlocal())
#ct = ct.astimezone(tzutc())
item.item_date = ct
item.save()
# extra attributes
try:
item.attributes['calldatetime'] = ct
except:
pass
try:
item.attributes['innum'] = int(get_element(entry, 'innum'))
except:
pass
for k in ['disp', 'aptlot', 'address']:
try:
item.attributes[k] = get_element(entry, k)
except:
pass
# create a lookup based on the title, this is the closest thing to
# a category that is available in the data.
lu = Lookup.objects.get_or_create_lookup(incident_type_field, title, title, "", False)
item.attributes['incident_type'] = lu.id
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except:
logger.error("Warning: couldn't save %r. Traceback: %s" % (item.title, traceback.format_exc()))
logger.info("Finished scraping police reports: %d added, %d updated" % (addcount, updatecount))
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']
item.block = result['block']
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
item.block = block
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(self):
""" Download Calendar RSS feed and update database """
logger.info("Starting EventsCalendarScraper")
feed = feedparser.parse(self.url)
seencount = addcount = updatecount = 0
for entry in feed.entries:
def ns_get(element):
# work around feedparser unpredictability.
namespace, element = element.split(':')
result = entry.get('%s_%s' % (namespace, element))
if result is None:
result = entry.get(element)
return result
seencount += 1
title = convert_entities(entry.title)
try:
item = NewsItem.objects.get(title=title,
schema__id=self.schema.id)
status = "updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "added"
except NewsItem.MultipleObjectsReturned:
logger.warn("Multiple entries matched title %r, event titles are not unique?" % title)
continue
try:
item.location_name = '%s %s' % (ns_get('xcal:x-calconnect-venue-name'),
ns_get('xcal:x-calconnect-street'))
item.location_name = item.location_name.strip()
item.schema = self.schema
item.title = title
item.description = convert_entities(entry.description)
item.url = entry.link
start_dt = ns_get('xcal:dtstart')
start_dt = dateutil.parser.parse(start_dt)
# Upstream bug: They provide a UTC offset of +0000 which
# means times in UTC, but they're actually times in
# US/Eastern, so do *not* fix the zone.
#start_dt = start_dt.astimezone(local_tz)
item.item_date = start_dt.date()
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.location = Point((float(ns_get('geo:long')),
float(ns_get('geo:lat'))))
if (item.location.x, item.location.y) == (0.0, 0.0):
logger.warn("Skipping %r, 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.info(" Reverse-geocoded point to %r" % block.pretty_name)
item.location_name = block.pretty_name
item.block = block
except reverse.ReverseGeocodeError:
logger.debug(" Failed to reverse geocode %s for %r" % (item.location.wkt, item.title))
item.location_name = u''
item.save()
item.attributes['start_time'] = start_dt.time()
end_dt = ns_get('xcal:dtend') or u''
if end_dt.strip():
end_dt = dateutil.parser.parse(end_dt.strip())
#end_dt = end_dt.astimezone(local_tz)
item.attributes['end_time'] = end_dt.time()
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except:
logger.error("unexpected error:", sys.exc_info()[1])
log_exception()
logger.info("EventsCalendarScraper finished: %d added, %d updated of %s total" % (addcount, updatecount, seencount))
def main(argv=None):
if argv:
url = argv[0]
else:
url = 'http://search.boston.com/search/api?q=*&sort=-articleprintpublicationdate&subject=massachusetts&scope=bonzai'
schema = 'local-news'
try:
schema = Schema.objects.get(slug=schema)
except Schema.DoesNotExist:
print "Schema (%s): DoesNotExist" % schema
sys.exit(1)
f = feedparser.parse(url)
for e in f.entries:
try:
item = NewsItem.objects.get(title=e.title, description=e.description)
print "Already have %r (id %d)" % (item.title, item.id)
except NewsItem.DoesNotExist:
item = NewsItem()
try:
item.schema = schema
item.title = convert_entities(e.title)
item.description = convert_entities(e.description)
item.url = e.link
item.location_name = e.get('x-calconnect-street') or e.get('georss_featurename')
item.item_date = datetime.datetime(*e.updated_parsed[:6])
item.pub_date = datetime.datetime(*e.updated_parsed[:6])
if 'point' in e:
x,y = e.point.split(' ')
elif 'georss_point' in e:
x,y = e.georss_point.split(' ')
else:
text = item.title + ' ' + item.description
from geocoder_hack import quick_dirty_fallback_geocode
x, y = quick_dirty_fallback_geocode(text, parse=True)
if None in (x, y):
print " 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?
print "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)
print " Reverse-geocoded point to %r" % block.pretty_name
item.location_name = block.pretty_name
except reverse.ReverseGeocodeError:
print " Failed to reverse geocode %s for %r" % (item.location.wkt, item.title)
item.location_name = u''
item.save()
print "Added: %s" % item.title
except:
print "Warning: couldn't save %r. Traceback:" % item.title
import cStringIO, traceback
f = cStringIO.StringIO()
traceback.print_exc(file=f)
msg = f.getvalue()
print msg
def update(url):
schema_slug = 'sheriff'
try:
schema = Schema.objects.get(slug=schema_slug)
except Schema.DoesNotExist:
logger.error( "Schema (%s): DoesNotExist" % schema_slug)
sys.exit(1)
incident_type_field = SchemaField.objects.get(schema=schema, name='incident_type')
try:
innum_field = SchemaField.objects.get(schema=schema, name='innum')
except SchemaField.DoesNotExist:
logger.error( "SchemaField innum Does Not Exist for %s" % schema_slug)
sys.exit(1)
logger.info("Scraping %s" % schema.name)
f = feedparser.parse(url)
addcount = updatecount = 0
for entry in f.entries:
innum = int(get_element(entry, 'innum'))
title = convert_entities(entry.title)
description = convert_entities(entry.summary)
try:
item = NewsItem.objects.filter(schema=schema).by_attribute(innum_field, innum)[0]
#url=item_url)
status = 'updated'
except IndexError:
item = NewsItem()
status = 'added'
try:
item.title = title
item.schema = schema
item.description = description
try:
item.location = Point((float(entry.geo_long), float(entry.geo_lat)))
except:
logger.info("Skipping item %s with no location information" % innum)
item.location_name = get_element(entry, 'address')
# this feed uses an invalidly formatted pubDate which
# appears to be intended to express the time of the
# incident, used for publication time as well.
# 24 hour time.
ct = datetime.strptime(entry.updated, r"%m/%d/%Y %H:%M:%S")
#ct = datetime(ct.year, ct.month, ct.day, ct.hour, ct.minute, ct.second, tzinfo=tzlocal())
#ct = ct.astimezone(tzutc())
item.item_date = ct
item.pub_date = ct.date()
item.save()
# extra attributes
item.attributes['innum'] = innum
for k in ['address']:
try:
item.attributes[k] = get_element(entry, k)
except:
pass
# create a lookup based on the title, this is the closest thing to
# a category that is available in the data.
lu = Lookup.objects.get_or_create_lookup(incident_type_field, title, title, "", False)
item.attributes['incident_type'] = lu.id
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except:
logger.error("Warning: couldn't save %r. Traceback: %s" % (item.title, traceback.format_exc()))
logger.info("Finished scraping %s: %d added, %d updated" % (schema.name, addcount, updatecount))
def update():
""" Download Calendar RSS feed and update database """
logger.info("Starting add_events")
url = """http://calendar.boston.com/search?acat=&cat=&commit=Search\
&new=n&rss=1&search=true&sort=0&srad=20&srss=50&ssrss=5&st=event\
&st_select=any&svt=text&swhat=&swhen=today&swhere=&trim=1"""
schema = 'events'
try:
schema = Schema.objects.get(slug=schema)
except Schema.DoesNotExist:
logger.error("Schema (%s): DoesNotExist" % schema)
sys.exit(1)
feed = feedparser.parse(url)
addcount = updatecount = 0
for entry in feed.entries:
title = convert_entities(entry.title).strip()
# Putting 'event' in the title is redundant, ticket #227
if title.lower().startswith('event: '):
title = title[7:]
try:
item = NewsItem.objects.get(title=title, schema__id=schema.id)
status = "updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "added"
except NewsItem.MultipleObjectsReturned:
logger.warn(
"Multiple entries matched title %r, event titles are not unique?"
% title)
continue
try:
item.location_name = entry.get(
'xcal_x-calconnect-street') or entry.get(
'x-calconnect-street') or u''
item.schema = schema
item.title = title
item.description = convert_entities(entry.description)
item.url = entry.link
item.item_date = datetime.datetime(*entry.updated_parsed[:6])
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.location = Point(
(float(entry['geo_long']), float(entry['geo_lat'])))
if (item.location.x, item.location.y) == (0.0, 0.0):
logger.warn("Skipping %r, 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.info(" 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.exception("unexpected error:", sys.exc_info()[1])
logger.info("add_events finished: %d added, %d updated" %
(addcount, updatecount))
def article_text_sections(tree):
"""
Given an HTML tree of a news article (or blog entry permalink), deduces
which part of it is text and returns a list of lists of strings, with each
string representing a paragraph and each list of strings representing a
"section" of the page.
"""
# The basic algorithm here is to combine all text within the same block
# (e.g., a <div>).
MIN_NUM_PARAGRAPHS = 3
MIN_NUM_PUNCTUATED = 3
# In order for a paragraph to be counted toward MIN_NUM_PUNCTUATED, it must
# have this number of characters.
MIN_CHARS_IN_PARAGRAPH = 30
# If this many paragraphs with MIN_CHARS_IN_PARAGRAPH are included in the
# section, then the section will be included, regardless of failing
# MIN_PERCENTAGE_PUNCTUATED.
NUM_PARAGRAPHS_SAFE_GUESS = 6
# In order for a section to be included in the result, at least this
# percentage of paragraphs in the section must be punctuated.
MIN_PERCENTAGE_PUNCTUATED = decimal.Decimal(".5")
block_tags = set(
["blockquote", "dd", "div", "dt", "h1", "h2", "h3", "h4", "h5", "h6", "h7", "h8", "li", "p", "td", "th", "tr"]
)
drop_tags_only = set(
[
"a",
"abbr",
"acronym",
"b",
"center",
"dir",
"dl",
"em",
"font",
"form",
"hr",
"i",
"label",
"menu",
"ol",
"pre",
"small",
"span",
"strong",
"sub",
"sup",
"table",
"tbody",
"tfoot",
"thead",
"topic",
"u",
"ul",
"wbr",
]
)
drop_tags_and_contents = set(
[
"applet",
"area",
"button",
"embed",
"img",
"iframe",
"head",
"input",
"link",
"map",
"meta",
"noscript",
"object",
"option",
"script",
"select",
"spacer",
"style",
"textarea",
"title",
]
)
layout_tags = set(["div", "h1", "h2", "h3", "h4", "h5", "h6", "h7", "h8", "td", "th", "tr"])
is_open_tag = re.compile("^<[^/][^>]+>$").search
is_close_tag = re.compile("^</[^>]+>$").search
ignored_paragraphs = set(
[
"del.icio.us",
"digg",
"email",
"e-mail editor",
"e-mail story",
"no comments",
"print",
"print article",
"printer-friendly",
"printer version",
"reprints",
]
)
elements_to_drop = []
for element in tree.getiterator():
if not isinstance(element.tag, basestring): # If it's a comment...
element.drop_tag()
continue
if element.text and "\n" in element.text:
element.text = element.text.replace("\n", " ")
if element.tail and "\n" in element.tail:
element.tail = element.tail.replace("\n", " ")
if element.tag in block_tags:
element.text = "\n" + (element.text or "")
element.tail = "\n" + (element.tail or "")
elif element.tag == "br":
element.tail = "\n" + (element.tail or "")
element.drop_tag()
elif element.tag in drop_tags_only:
element.drop_tag()
elif element.tag in drop_tags_and_contents:
elements_to_drop.append(element)
elif element.tag not in ("html", "body"): # Unknown tag!
element.drop_tag()
for e in elements_to_drop:
e.drop_tree()
for element in tree.getiterator():
if element.tag in block_tags:
if element.tag in layout_tags:
element.text = "\n<%s>\n%s\n" % (element.tag, (element.text or ""))
element.tail = "\n</%s>\n%s\n" % (element.tag, (element.tail or ""))
element.drop_tag()
try:
tree.body
except IndexError:
# In some cases, the article is missing a <body> tag, and tree.body
# will result in an IndexError. Just skip these.
return []
new_html = etree.tostring(tree.body, method="html")
new_html = convert_entities(new_html)
lines = re.split(r"\s*\n+\s*", new_html.strip())
result = []
sections = []
for line in lines:
if is_open_tag(line):
result.append([])
elif is_close_tag(line):
last_bit = result.pop()
if len(last_bit) >= MIN_NUM_PARAGRAPHS:
sections.append(last_bit)
else: # It's text, not a tag.
try:
result[-1].append(line)
except IndexError: # No tags seen yet.
result.append([line])
# Cut out the sections that don't contain enough punctuated sentences.
final_sections = []
for section in sections:
count = 0
to_delete = []
for i, paragraph in enumerate(section):
if paragraph.lower() in ignored_paragraphs:
to_delete.append(i)
elif is_punctuated(paragraph) and len(paragraph) >= MIN_CHARS_IN_PARAGRAPH:
count += 1
percent_punctuated = decimal.Decimal(count) / decimal.Decimal(len(section))
if count >= NUM_PARAGRAPHS_SAFE_GUESS or (
count >= MIN_NUM_PUNCTUATED and percent_punctuated >= MIN_PERCENTAGE_PUNCTUATED
):
for i in reversed(to_delete): # Delete in reverse so that index order is preserved.
del section[i]
final_sections.append(section)
return final_sections
def article_text_sections(tree):
"""
Given an HTML tree of a news article (or blog entry permalink), deduces
which part of it is text and returns a list of lists of strings, with each
string representing a paragraph and each list of strings representing a
"section" of the page.
"""
# The basic algorithm here is to combine all text within the same block
# (e.g., a <div>).
MIN_NUM_PARAGRAPHS = 3
MIN_NUM_PUNCTUATED = 3
# In order for a paragraph to be counted toward MIN_NUM_PUNCTUATED, it must
# have this number of characters.
MIN_CHARS_IN_PARAGRAPH = 30
# If this many paragraphs with MIN_CHARS_IN_PARAGRAPH are included in the
# section, then the section will be included, regardless of failing
# MIN_PERCENTAGE_PUNCTUATED.
NUM_PARAGRAPHS_SAFE_GUESS = 6
# In order for a section to be included in the result, at least this
# percentage of paragraphs in the section must be punctuated.
MIN_PERCENTAGE_PUNCTUATED = decimal.Decimal('.5')
block_tags = set(['blockquote', 'dd', 'div', 'dt', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'li', 'p', 'td', 'th', 'tr'])
drop_tags_only = set(['a', 'abbr', 'acronym', 'b', 'center', 'dir', 'dl', 'em', 'font', 'form', 'hr', 'i', 'label', 'menu', 'ol', 'pre', 'small', 'span', 'strong', 'sub', 'sup', 'table', 'tbody', 'tfoot', 'thead', 'topic', 'u', 'ul', 'wbr'])
drop_tags_and_contents = set(['applet', 'area', 'button', 'embed', 'img', 'iframe', 'head', 'input', 'link', 'map', 'meta', 'noscript', 'object', 'option', 'script', 'select', 'spacer', 'style', 'textarea', 'title'])
layout_tags = set(['div', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'td', 'th', 'tr'])
is_open_tag = re.compile('^<[^/][^>]+>$').search
is_close_tag = re.compile('^</[^>]+>$').search
ignored_paragraphs = set(['del.icio.us', 'digg', 'email', 'e-mail editor', 'e-mail story', 'no comments', 'print', 'print article', 'printer-friendly', 'printer version', 'reprints'])
elements_to_drop = []
for element in tree.getiterator():
if not isinstance(element.tag, basestring): # If it's a comment...
element.drop_tag()
continue
if element.text and '\n' in element.text:
element.text = element.text.replace('\n', ' ')
if element.tail and '\n' in element.tail:
element.tail = element.tail.replace('\n', ' ')
if element.tag in block_tags:
element.text = '\n' + (element.text or '')
element.tail = '\n' + (element.tail or '')
elif element.tag == 'br':
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag in drop_tags_only:
element.drop_tag()
elif element.tag in drop_tags_and_contents:
elements_to_drop.append(element)
elif element.tag not in ('html', 'body'): # Unknown tag!
element.drop_tag()
for e in elements_to_drop:
e.drop_tree()
for element in tree.getiterator():
if element.tag in block_tags:
if element.tag in layout_tags:
element.text = '\n<%s>\n%s\n' % (element.tag, (element.text or ''))
element.tail = '\n</%s>\n%s\n' % (element.tag, (element.tail or ''))
element.drop_tag()
try:
tree.body
except IndexError:
# In some cases, the article is missing a <body> tag, and tree.body
# will result in an IndexError. Just skip these.
return []
new_html = etree.tostring(tree.body, method='html')
new_html = convert_entities(new_html)
lines = re.split(r'\s*\n+\s*', new_html.strip())
result = []
sections = []
for line in lines:
if is_open_tag(line):
result.append([])
elif is_close_tag(line):
last_bit = result.pop()
if len(last_bit) >= MIN_NUM_PARAGRAPHS:
sections.append(last_bit)
else: # It's text, not a tag.
try:
result[-1].append(line)
except IndexError: # No tags seen yet.
result.append([line])
# Cut out the sections that don't contain enough punctuated sentences.
final_sections = []
for section in sections:
count = 0
to_delete = []
for i, paragraph in enumerate(section):
if paragraph.lower() in ignored_paragraphs:
to_delete.append(i)
elif is_punctuated(paragraph) and len(paragraph) >= MIN_CHARS_IN_PARAGRAPH:
count += 1
percent_punctuated = decimal.Decimal(count) / decimal.Decimal(len(section))
if count >= NUM_PARAGRAPHS_SAFE_GUESS or (count >= MIN_NUM_PUNCTUATED and percent_punctuated >= MIN_PERCENTAGE_PUNCTUATED):
for i in reversed(to_delete): # Delete in reverse so that index order is preserved.
del section[i]
final_sections.append(section)
return final_sections
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(self):
#
#
# Download Calendar RSS feed and update database
#
#
logger.info("Starting KSUStudentProgrammingScraper")
feed = feedparser.parse(self.url)
seencount = addcount = updatecount = 0
for entry in feed.entries:
seencount += 1
title = convert_entities(entry.title)
title = foo(title, '', ' (')
try:
item = NewsItem.objects.get(title=title,
schema__id=self.schema.id)
status = "updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "added"
except NewsItem.MultipleObjectsReturned:
logger.warn("Multiple entries matched title %r, event titles are not unique?" % title)
continue
try:
#
#
# The actual rss feed elements are grabbed here
#
#
itm_description = entry.description
soup = BeautifulSoup(foo(itm_description,"</table><br />","<br /><br />"))
locations = soup.findAll(text=True)
location = locations[0].strip()
place_grabber = places.place_grabber()
grab_results = place_grabber(location)
try:
item.location = Place.objects.get(pretty_name=grab_results[0][2]).location
item.location_name = Place.objects.get(pretty_name=grab_results[0][2]).pretty_name
except:
item.location = PlaceSynonym.objects.get(pretty_name=grab_results[0][2]).place.location
item.location_name = PlaceSynonym.objects.get(pretty_name=grab_results[0][2]).place.pretty_name
try:
item.attributes['room'] = locations[1].strip().replace("Room: ","")
except Exception as e:
logger.info("Tried saving item.room, error: %s" % e)
item.schema = self.schema
item.title = title
soup = BeautifulSoup(foo(itm_description,"<br /><br />","</td></tr>"))
item.description = soup.findAll(text=True)
item.description = item.description[0].strip()
item.url = entry.link
start_t = foo(itm_description,"Start Time:</b> </td><td>","</td>")
start_t = dateutil.parser.parse(start_t)
end_t = foo(itm_description,"End Time:</b> </td><td>","</td>")
end_t = dateutil.parser.parse(end_t)
end_dt = foo(itm_description,"End Date:</b> </td><td>","</td>")
end_dt = dateutil.parser.parse(end_dt)
item.item_date = dateutil.parser.parse(entry.category)
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.attributes['start-time'] = start_t.time()
item.attributes['end-time'] = end_t.time()
item.save()
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except Exception as e:
logger.exception("unexpected error: %s" % e)
logger.info("KSUStudentProgrammingScraper finished: %d added, %d updated of %s total" % (addcount, updatecount, seencount))
def update(self):
""" Download Calendar RSS feed and update database """
logger.info("Starting EventsCalendarScraper")
feed = feedparser.parse(self.url)
seencount = addcount = updatecount = 0
for entry in feed.entries:
def ns_get(element):
# work around feedparser unpredictability.
namespace, element = element.split(':')
result = entry.get('%s_%s' % (namespace, element))
if result is None:
result = entry.get(element)
return result
seencount += 1
title = convert_entities(entry.title)
try:
item = NewsItem.objects.get(title=title,
schema__id=self.schema.id)
status = "updated"
except NewsItem.DoesNotExist:
item = NewsItem()
status = "added"
except NewsItem.MultipleObjectsReturned:
logger.warn(
"Multiple entries matched title %r, event titles are not unique?"
% title)
continue
try:
item.location_name = '%s %s' % (
ns_get('xcal:x-calconnect-venue-name'),
ns_get('xcal:x-calconnect-street'))
item.location_name = item.location_name.strip()
item.schema = self.schema
item.title = title
item.description = convert_entities(entry.description)
item.url = entry.link
start_dt = ns_get('xcal:dtstart')
start_dt = dateutil.parser.parse(start_dt)
# Upstream bug: They provide a UTC offset of +0000 which
# means times in UTC, but they're actually times in
# US/Eastern, so do *not* fix the zone.
#start_dt = start_dt.astimezone(local_tz)
item.item_date = start_dt.date()
item.pub_date = datetime.datetime(*entry.updated_parsed[:6])
item.location = Point(
(float(ns_get('geo:long')), float(ns_get('geo:lat'))))
if (item.location.x, item.location.y) == (0.0, 0.0):
logger.warn("Skipping %r, 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.info(" 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()
item.attributes['start_time'] = start_dt.time()
end_dt = ns_get('xcal:dtend') or u''
if end_dt.strip():
end_dt = dateutil.parser.parse(end_dt.strip())
#end_dt = end_dt.astimezone(local_tz)
item.attributes['end_time'] = end_dt.time()
if status == 'added':
addcount += 1
else:
updatecount += 1
logger.info("%s: %s" % (status, item.title))
except Exception as e:
logger.exception("unexpected error: %s" % e)
logger.info(
"EventsCalendarScraper finished: %d added, %d updated of %s total"
% (addcount, updatecount, seencount))
def main(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')
if point:
x, y = point.split(' ')
else:
# Fall back on geocoding.
text = item.title + ' ' + item.description
try:
x, y = quick_dirty_fallback_geocode(text, parse=True)
except GeocodingException:
logger.debug("Geocoding exception on %r:" % text)
log_exception(level=logging.DEBUG)
continue
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
item.block = block
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))
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,
exc_info=True)
continue
except:
logger.exception('uncaught geocoder exception on %r\n' % addr)
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.exception("Warning: couldn't save %r. Traceback:" % item.title)
logger.info("Finished add_news: %d added, %d updated" % (addcount, updatecount))
def article_text_sections(tree):
"""
Given an HTML tree of a news article (or blog entry permalink), deduces
which part of it is text and returns a list of lists of strings, with each
string representing a paragraph and each list of strings representing a
"section" of the page.
"""
# The basic algorithm here is to combine all text within the same block
# (e.g., a <div>).
MIN_NUM_PARAGRAPHS = 3
MIN_NUM_PUNCTUATED = 3
# In order for a paragraph to be counted toward MIN_NUM_PUNCTUATED, it must
# have this number of characters.
MIN_CHARS_IN_PARAGRAPH = 30
# If this many paragraphs with MIN_CHARS_IN_PARAGRAPH are included in the
# section, then the section will be included, regardless of failing
# MIN_PERCENTAGE_PUNCTUATED.
NUM_PARAGRAPHS_SAFE_GUESS = 6
# In order for a section to be included in the result, at least this
# percentage of paragraphs in the section must be punctuated.
MIN_PERCENTAGE_PUNCTUATED = decimal.Decimal('.5')
block_tags = set([
'blockquote', 'dd', 'div', 'dt', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
'h7', 'h8', 'li', 'p', 'td', 'th', 'tr'
])
drop_tags_only = set([
'a', 'abbr', 'acronym', 'b', 'center', 'dir', 'dl', 'em', 'font',
'form', 'hr', 'i', 'label', 'menu', 'ol', 'pre', 'small', 'span',
'strong', 'sub', 'sup', 'table', 'tbody', 'tfoot', 'thead', 'topic',
'u', 'ul', 'wbr'
])
drop_tags_and_contents = set([
'applet', 'area', 'button', 'embed', 'img', 'iframe', 'head', 'input',
'link', 'map', 'meta', 'noscript', 'object', 'option', 'script',
'select', 'spacer', 'style', 'textarea', 'title'
])
layout_tags = set([
'div', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'h7', 'h8', 'td', 'th', 'tr'
])
is_open_tag = re.compile('^<[^/][^>]+>$').search
is_close_tag = re.compile('^</[^>]+>$').search
ignored_paragraphs = set([
'del.icio.us', 'digg', 'email', 'e-mail editor', 'e-mail story',
'no comments', 'print', 'print article', 'printer-friendly',
'printer version', 'reprints'
])
elements_to_drop = []
for element in tree.getiterator():
if not isinstance(element.tag, basestring): # If it's a comment...
element.drop_tag()
continue
if element.text and '\n' in element.text:
element.text = element.text.replace('\n', ' ')
if element.tail and '\n' in element.tail:
element.tail = element.tail.replace('\n', ' ')
if element.tag in block_tags:
element.text = '\n' + (element.text or '')
element.tail = '\n' + (element.tail or '')
elif element.tag == 'br':
element.tail = '\n' + (element.tail or '')
element.drop_tag()
elif element.tag in drop_tags_only:
element.drop_tag()
elif element.tag in drop_tags_and_contents:
elements_to_drop.append(element)
elif element.tag not in ('html', 'body'): # Unknown tag!
element.drop_tag()
for e in elements_to_drop:
e.drop_tree()
for element in tree.getiterator():
if element.tag in block_tags:
if element.tag in layout_tags:
element.text = '\n<%s>\n%s\n' % (element.tag,
(element.text or ''))
element.tail = '\n</%s>\n%s\n' % (element.tag,
(element.tail or ''))
element.drop_tag()
try:
tree.body
except IndexError:
# In some cases, the article is missing a <body> tag, and tree.body
# will result in an IndexError. Just skip these.
return []
new_html = etree.tostring(tree.body, method='html')
new_html = convert_entities(new_html)
lines = re.split(r'\s*\n+\s*', new_html.strip())
result = []
sections = []
for line in lines:
if is_open_tag(line):
result.append([])
elif is_close_tag(line):
last_bit = result.pop()
if len(last_bit) >= MIN_NUM_PARAGRAPHS:
sections.append(last_bit)
else: # It's text, not a tag.
try:
result[-1].append(line)
except IndexError: # No tags seen yet.
result.append([line])
# Cut out the sections that don't contain enough punctuated sentences.
final_sections = []
for section in sections:
count = 0
to_delete = []
for i, paragraph in enumerate(section):
# Quotes seem to be causing the is_punctuated to fail
paragraph = paragraph.replace('"', '').strip()
paragraph = paragraph.replace("'", '').strip()
if paragraph.lower() in ignored_paragraphs:
to_delete.append(i)
elif is_punctuated(
paragraph) and len(paragraph) >= MIN_CHARS_IN_PARAGRAPH:
count += 1
percent_punctuated = decimal.Decimal(count) / decimal.Decimal(
len(section))
if count >= NUM_PARAGRAPHS_SAFE_GUESS or (
count >= MIN_NUM_PUNCTUATED
and percent_punctuated >= MIN_PERCENTAGE_PUNCTUATED):
for i in reversed(
to_delete
): # Delete in reverse so that index order is preserved.
del section[i]
final_sections.append(section)
return final_sections
