def get_groundtruth(lang):
"""Get actual counts of top articles and their pageviews from a Wikipedia from yesterday."""
p = PageviewsClient(
user_agent="[email protected] -- diff private toolforge")
try:
groundtruth = p.top_articles(
project='{0}.wikipedia'.format(lang),
access='all-access',
year=None,
month=None,
day=None, # defaults to yesterday
limit=50)
except Exception:
two_days_ago = date.today() - timedelta(days=2)
groundtruth = p.top_articles(project='{0}.wikipedia'.format(lang),
access='all-access',
year=two_days_ago.year,
month=two_days_ago.month,
day=two_days_ago.day,
limit=50)
return {
r['article']: {
'gt-rank': r['rank'],
'gt-views': r['views']
}
for r in groundtruth
}
class WikiIngest(object):
def __init__(self):
self.db_connection = DBConnection()
self.logger = logging.getLogger(__name__)
self.api = PageviewsClient(
"Mozilla/5.0 (X11; Linux x86_64)"
" AppleWebKit/537.36 (KHTML, like Gecko) Chrome/60.0.3112.113 Safari/537.36"
)
def get_top_articles(self, time_collect=None, historic=False):
if not historic:
time_collect = datetime.now() - timedelta(days=1)
results = self.api.top_articles(project=WIKI_SOURCES.ENGLISH_WIKIPEDIA,
year=time_collect.year,
month=time_collect.month,
day=time_collect.day)
timestamp = calendar.timegm(time_collect.timetuple())
articles_to_insert = []
bulk_op = None
if historic:
bulk_op = self.db_connection.start_bulk_upsert(
collection=DB.WIKI_TRENDS)
for result in results:
name = result["article"]
if "_" in name:
name = name.replace("_", " ")
doc = {
WIKI_TREND.NAME: name,
WIKI_TREND.RANK: int(result["rank"]),
WIKI_TREND.VIEWS: int(result["views"]),
WIKI_TREND.TIMESTAMP: timestamp,
WIKI_TREND.DATE_OBJECT: time_collect,
WIKI_TREND.DATE_STRING: time_collect.strftime("%A %B %d %Y"),
WIKI_TREND.MONTH: time_collect.strftime("%B").lower(),
WIKI_TREND.WEEKDAY: time_collect.strftime("%A").lower(),
WIKI_TREND.MONTH_DAY: int(time_collect.strftime("%d")),
WIKI_TREND.YEAR: time_collect.strftime("%Y")
}
if historic:
self.db_connection.add_to_bulk_upsert(query={
"$and": [{
WIKI_TREND.NAME: name
}, {
WIKI_TREND.DATE_STRING:
time_collect.strftime("%A %B %d %Y")
}]
},
data=doc,
bulk_op=bulk_op)
else:
articles_to_insert.append(doc)
if historic:
self.db_connection.end_bulk_upsert(bulk_op=bulk_op)
else:
self.db_connection.bulk_insert(data=articles_to_insert,
collection=DB.WIKI_TRENDS)
def download_pageviews(entities=None, start='20150701', end=None, access='desktop', agent='user', limit=10000):
"""
Download pageviews from Wikipedia
:param entities: A list of entities (Wikipedia pages) to get pageview data for
:param start: The start date of the range over which to collect data;
2015-07-01 is the earliest supported by the API
:param end: The end date of the range, or None for today
:param access: The method by which Wikipedia was accessed (default: desktop)
:param agent: The user agent accessing Wikipedia (default: user)
:param limit: The number of most-trafficked entities to return data for, if no entities are specified in the call
:return: A DataFrame of entities x pageviews by day
"""
if end is None:
end = datetime.date.today().strftime('%Y%m%d')
p = PageviewsClient()
dates = pd.date_range(start=start, end=end)
#str -> list
if type(entities) is str:
entities = [entities]
# if entities aren't passed in, get the daily top entities for the period
if entities is None:
df_pvs = None
for d in dates:
try:
df = pd.DataFrame(p.top_articles('en.wikipedia', year=d.year, month=d.month,\
day=d.day, limit=limit, access=access))
except:
continue
df = df.set_index('article').rename(columns={'views': d})[[d]]
if df_pvs is None:
df_pvs = df
else:
df_pvs = df_pvs.join(df, how='outer')
entities = df_pvs.index.values.tolist()
for i in range(len(entities)):
try:
entities[i] = unidecode(wikipedia.page(entities[i]).title)
except wikipedia.exceptions.DisambiguationError as e:
print 'I could not understand that, please check your spelling or be more specific'
print 'Error: {0}'.format(e)
avere = pd.DataFrame(columns=['NONE'])
return avere
except wikipedia.exceptions.PageError as e:
print 'I could not understand that, please check your spelling or be more specific'
print 'Error: {0}'.format(e)
avere = pd.DataFrame(columns=['NONE'])
return avere
search = p.article_views('en.wikipedia', entities, start=start, end=end, access=access, agent=agent)
df = pd.DataFrame.from_dict(search, orient='index')
return df
