--- a/ckanext/ga_report/controller.py
+++ b/ckanext/ga_report/controller.py
@@ -1,6 +1,7 @@
import re
import csv
import sys
+import json
import logging
import operator
import collections
@@ -9,10 +10,11 @@
import sqlalchemy
from sqlalchemy import func, cast, Integer
import ckan.model as model
-from ga_model import GA_Url, GA_Stat, GA_ReferralStat
+from ga_model import GA_Url, GA_Stat, GA_ReferralStat, GA_Publisher
log = logging.getLogger('ckanext.ga-report')
+DOWNLOADS_AVAILABLE_FROM = '2012-12'
def _get_month_name(strdate):
import calendar
@@ -20,13 +22,39 @@
d = strptime(strdate, '%Y-%m')
return '%s %s' % (calendar.month_name[d.tm_mon], d.tm_year)
-
-def _month_details(cls):
+def _get_unix_epoch(strdate):
+ from time import strptime,mktime
+ d = strptime(strdate, '%Y-%m')
+ return int(mktime(d))
+
+def _month_details(cls, stat_key=None):
+ '''
+ Returns a list of all the periods for which we have data, unfortunately
+ knows too much about the type of the cls being passed as GA_Url has a
+ more complex query
+
+ This may need extending if we add a period_name to the stats
+ '''
months = []
- vals = model.Session.query(cls.period_name).distinct().all()
+ day = None
+
+ q = model.Session.query(cls.period_name,cls.period_complete_day)\
+ .filter(cls.period_name!='All').distinct(cls.period_name)
+ if stat_key:
+ q= q.filter(cls.stat_name==stat_key)
+
+ vals = q.order_by("period_name desc").all()
+
+ if vals and vals[0][1]:
+ day = int(vals[0][1])
+ ordinal = 'th' if 11 <= day <= 13 \
+ else {1:'st',2:'nd',3:'rd'}.get(day % 10, 'th')
+ day = "{day}{ordinal}".format(day=day, ordinal=ordinal)
+
for m in vals:
months.append( (m[0], _get_month_name(m[0])))
- return sorted(months, key=operator.itemgetter(0), reverse=True)
+
+ return months, day
class GaReport(BaseController):
@@ -34,7 +62,7 @@
def csv(self, month):
import csv
- q = model.Session.query(GA_Stat)
+ q = model.Session.query(GA_Stat).filter(GA_Stat.stat_name!='Downloads')
if month != 'all':
q = q.filter(GA_Stat.period_name==month)
entries = q.order_by('GA_Stat.period_name, GA_Stat.stat_name, GA_Stat.key').all()
@@ -51,11 +79,12 @@
entry.key.encode('utf-8'),
entry.value.encode('utf-8')])
+
def index(self):
# Get the month details by fetching distinct values and determining the
# month names from the values.
- c.months = _month_details(GA_Stat)
+ c.months, c.day = _month_details(GA_Stat)
# Work out which month to show, based on query params of the first item
c.month_desc = 'all months'
@@ -70,24 +99,39 @@
entries = q.order_by('ga_stat.key').all()
def clean_key(key, val):
- if key in ['Average time on site', 'Pages per visit', 'New visits', 'Bounces']:
+ if key in ['Average time on site', 'Pages per visit', 'New visits', 'Bounce rate (home page)']:
val = "%.2f" % round(float(val), 2)
if key == 'Average time on site':
mins, secs = divmod(float(val), 60)
hours, mins = divmod(mins, 60)
val = '%02d:%02d:%02d (%s seconds) ' % (hours, mins, secs, val)
- if key in ['New visits','Bounces']:
+ if key in ['New visits','Bounce rate (home page)']:
val = "%s%%" % val
if key in ['Total page views', 'Total visits']:
val = int(val)
return key, val
+
+ # Query historic values for sparkline rendering
+ sparkline_query = model.Session.query(GA_Stat)\
+ .filter(GA_Stat.stat_name=='Totals')\
+ .order_by(GA_Stat.period_name)
+ sparkline_data = {}
+ for x in sparkline_query:
+ sparkline_data[x.key] = sparkline_data.get(x.key,[])
+ key, val = clean_key(x.key,float(x.value))
+ tooltip = '%s: %s' % (_get_month_name(x.period_name), val)
+ sparkline_data[x.key].append( (tooltip,x.value) )
+ # Trim the latest month, as it looks like a huge dropoff
+ for key in sparkline_data:
+ sparkline_data[key] = sparkline_data[key][:-1]
c.global_totals = []
if c.month:
for e in entries:
key, val = clean_key(e.key, e.value)
- c.global_totals.append((key, val))
+ sparkline = sparkline_data[e.key]
+ c.global_totals.append((key, val, sparkline))
else:
d = collections.defaultdict(list)
for e in entries:
@@ -96,11 +140,19 @@
if k in ['Total page views', 'Total visits']:
v = sum(v)
else:
- v = float(sum(v))/len(v)
+ v = float(sum(v))/float(len(v))
+ sparkline = sparkline_data[k]
key, val = clean_key(k,v)
- c.global_totals.append((key, val))
- c.global_totals = sorted(c.global_totals, key=operator.itemgetter(0))
+ c.global_totals.append((key, val, sparkline))
+ # Sort the global totals into a more pleasant order
+ def sort_func(x):
+ key = x[0]
+ total_order = ['Total page views','Total visits','Pages per visit']
+ if key in total_order:
+ return total_order.index(key)
+ return 999
+ c.global_totals = sorted(c.global_totals, key=sort_func)
keys = {
'Browser versions': 'browser_versions',
@@ -137,12 +189,13 @@
for k, v in keys.iteritems():
q = model.Session.query(GA_Stat).\
- filter(GA_Stat.stat_name==k)
+ filter(GA_Stat.stat_name==k).\
+ order_by(GA_Stat.period_name)
+ # Buffer the tabular data
if c.month:
entries = []
q = q.filter(GA_Stat.period_name==c.month).\
order_by('ga_stat.value::int desc')
-
d = collections.defaultdict(int)
for e in q.all():
d[e.key] += int(e.value)
@@ -151,10 +204,27 @@
entries.append((key,val,))
entries = sorted(entries, key=operator.itemgetter(1), reverse=True)
+ # Run a query on all months to gather graph data
+ graph_query = model.Session.query(GA_Stat).\
+ filter(GA_Stat.stat_name==k).\
+ order_by(GA_Stat.period_name)
+ graph_dict = {}
+ for stat in graph_query:
+ graph_dict[ stat.key ] = graph_dict.get(stat.key,{
+ 'name':stat.key,
+ 'raw': {}
+ })
+ graph_dict[ stat.key ]['raw'][stat.period_name] = float(stat.value)
+ stats_in_table = [x[0] for x in entries]
+ stats_not_in_table = set(graph_dict.keys()) - set(stats_in_table)
+ stats = stats_in_table + sorted(list(stats_not_in_table))
+ graph = [graph_dict[x] for x in stats]
+ setattr(c, v+'_graph', json.dumps( _to_rickshaw(graph,percentageMode=True) ))
+
# Get the total for each set of values and then set the value as
# a percentage of the total
if k == 'Social sources':
- total = sum([x for n,x in c.global_totals if n == 'Total visits'])
+ total = sum([x for n,x,graph in c.global_totals if n == 'Total visits'])
else:
total = sum([num for _,num in entries])
setattr(c, v, [(k,_percent(v,total)) for k,v in entries ])
@@ -179,7 +249,9 @@
writer = csv.writer(response)
writer.writerow(["Publisher Title", "Publisher Name", "Views", "Visits", "Period Name"])
- for publisher,view,visit in _get_top_publishers(None):
+ top_publishers = _get_top_publishers(limit=None)
+
+ for publisher,view,visit in top_publishers:
writer.writerow([publisher.title.encode('utf-8'),
publisher.name.encode('utf-8'),
view,
@@ -199,19 +271,20 @@
if not c.publisher:
abort(404, 'A publisher with that name could not be found')
- packages = self._get_packages(c.publisher)
+ packages = self._get_packages(publisher=c.publisher, month=c.month)
response.headers['Content-Type'] = "text/csv; charset=utf-8"
response.headers['Content-Disposition'] = \
str('attachment; filename=datasets_%s_%s.csv' % (c.publisher_name, month,))
writer = csv.writer(response)
- writer.writerow(["Dataset Title", "Dataset Name", "Views", "Visits", "Period Name"])
-
- for package,view,visit in packages:
+ writer.writerow(["Dataset Title", "Dataset Name", "Views", "Visits", "Resource downloads", "Period Name"])
+
+ for package,view,visit,downloads in packages:
writer.writerow([package.title.encode('utf-8'),
package.name.encode('utf-8'),
view,
visit,
+ downloads,
month])
def publishers(self):
@@ -219,7 +292,7 @@
# Get the month details by fetching distinct values and determining the
# month names from the values.
- c.months = _month_details(GA_Url)
+ c.months, c.day = _month_details(GA_Url)
# Work out which month to show, based on query params of the first item
c.month = request.params.get('month', '')
@@ -228,49 +301,51 @@
c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
c.top_publishers = _get_top_publishers()
-
- return render('ga_report/publisher/index.html')
-
- def _get_packages(self, publisher=None, count=-1):
- '''Returns the datasets in order of visits'''
- if count == -1:
- count = sys.maxint
-
- q = model.Session.query(GA_Url)\
+ graph_data = _get_top_publishers_graph()
+ c.top_publishers_graph = json.dumps( _to_rickshaw(graph_data) )
+
+ x = render('ga_report/publisher/index.html')
+
+ return x
+
+ def _get_packages(self, publisher=None, month='', count=-1):
+ '''Returns the datasets in order of views'''
+ have_download_data = True
+ month = month or 'All'
+ if month != 'All':
+ have_download_data = month >= DOWNLOADS_AVAILABLE_FROM
+
+ q = model.Session.query(GA_Url,model.Package)\
+ .filter(model.Package.name==GA_Url.package_id)\
.filter(GA_Url.url.like('/dataset/%'))
if publisher:
q = q.filter(GA_Url.department_id==publisher.name)
- if c.month:
- q = q.filter(GA_Url.period_name==c.month)
- q = q.order_by('ga_url.visitors::int desc')
-
- if c.month:
- top_packages = []
- for entry in q.limit(count):
- package_name = entry.url[len('/dataset/'):]
- p = model.Package.get(package_name)
- if p:
- top_packages.append((p, entry.pageviews, entry.visitors))
+ q = q.filter(GA_Url.period_name==month)
+ q = q.order_by('ga_url.pageviews::int desc')
+ top_packages = []
+ if count == -1:
+ entries = q.all()
+ else:
+ entries = q.limit(count)
+
+ for entry,package in entries:
+ if package:
+ # Downloads ....
+ if have_download_data:
+ dls = model.Session.query(GA_Stat).\
+ filter(GA_Stat.stat_name=='Downloads').\
+ filter(GA_Stat.key==package.name)
+ if month != 'All': # Fetch everything unless the month is specific
+ dls = dls.filter(GA_Stat.period_name==month)
+ downloads = 0
+ for x in dls:
+ downloads += int(x.value)
else:
- log.warning('Could not find package "%s"', package_name)
- else:
- ds = {}
- for entry in q.limit(count):
- package_name = entry.url[len('/dataset/'):]
- p = model.Package.get(package_name)
- if p:
- if not p in ds:
- ds[p] = {'views': 0, 'visits': 0}
- ds[p]['views'] = ds[p]['views'] + int(entry.pageviews)
- ds[p]['visits'] = ds[p]['visits'] + int(entry.visitors)
- else:
- log.warning('Could not find package "%s"', package_name)
-
- results = []
- for k, v in ds.iteritems():
- results.append((k,v['views'],v['visits']))
-
- top_packages = sorted(results, key=operator.itemgetter(1), reverse=True)
+ downloads = 'No data'
+ top_packages.append((package, entry.pageviews, entry.visits, downloads))
+ else:
+ log.warning('Could not find package associated package')
+
return top_packages
def read(self):
@@ -297,7 +372,7 @@
# Get the month details by fetching distinct values and determining the
# month names from the values.
- c.months = _month_details(GA_Url)
+ c.months, c.day = _month_details(GA_Url)
# Work out which month to show, based on query params of the first item
c.month = request.params.get('month', '')
@@ -306,61 +381,157 @@
else:
c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
+ month = c.month or 'All'
c.publisher_page_views = 0
q = model.Session.query(GA_Url).\
filter(GA_Url.url=='/publisher/%s' % c.publisher_name)
- if c.month:
- entry = q.filter(GA_Url.period_name==c.month).first()
- c.publisher_page_views = entry.pageviews if entry else 0
- else:
- for e in q.all():
- c.publisher_page_views = c.publisher_page_views + int(e.pageviews)
-
- c.top_packages = self._get_packages(c.publisher, 20)
+ entry = q.filter(GA_Url.period_name==c.month).first()
+ c.publisher_page_views = entry.pageviews if entry else 0
+
+ c.top_packages = self._get_packages(publisher=c.publisher, count=20, month=c.month)
+
+ # Graph query
+ top_packages_all_time = self._get_packages(publisher=c.publisher, count=20, month='All')
+ top_package_names = [ x[0].name for x in top_packages_all_time ]
+ graph_query = model.Session.query(GA_Url,model.Package)\
+ .filter(model.Package.name==GA_Url.package_id)\
+ .filter(GA_Url.url.like('/dataset/%'))\
+ .filter(GA_Url.package_id.in_(top_package_names))
+ all_series = {}
+ for entry,package in graph_query:
+ if not package: continue
+ if entry.period_name=='All': continue
+ all_series[package.name] = all_series.get(package.name,{
+ 'name':package.title,
+ 'raw': {}
+ })
+ all_series[package.name]['raw'][entry.period_name] = int(entry.pageviews)
+ graph = [ all_series[series_name] for series_name in top_package_names ]
+ c.graph_data = json.dumps( _to_rickshaw(graph) )
return render('ga_report/publisher/read.html')
+
+def _to_rickshaw(data, percentageMode=False):
+ if data==[]:
+ return data
+ # x-axis is every month in c.months. Note that data might not exist
+ # for entire history, eg. for recently-added datasets
+ x_axis = [x[0] for x in c.months]
+ x_axis.reverse() # Ascending order
+ x_axis = x_axis[:-1] # Remove latest month
+ totals = {}
+ for series in data:
+ series['data'] = []
+ for x_string in x_axis:
+ x = _get_unix_epoch( x_string )
+ y = series['raw'].get(x_string,0)
+ series['data'].append({'x':x,'y':y})
+ totals[x] = totals.get(x,0)+y
+ if not percentageMode:
+ return data
+ # Turn all data into percentages
+ # Roll insignificant series into a catch-all
+ THRESHOLD = 1
+ raw_data = data
+ data = []
+ for series in raw_data:
+ for point in series['data']:
+ percentage = (100*float(point['y'])) / totals[point['x']]
+ if not (series in data) and percentage>THRESHOLD:
+ data.append(series)
+ point['y'] = percentage
+ others = [ x for x in raw_data if not (x in data) ]
+ if len(others):
+ data_other = []
+ for i in range(len(x_axis)):
+ x = _get_unix_epoch(x_axis[i])
+ y = 0
+ for series in others:
+ y += series['data'][i]['y']
+ data_other.append({'x':x,'y':y})
+ data.append({
+ 'name':'Other',
+ 'data': data_other
+ })
+ return data
+
def _get_top_publishers(limit=20):
'''
Returns a list of the top 20 publishers by dataset visits.
(The number to show can be varied with 'limit')
'''
+ month = c.month or 'All'
connection = model.Session.connection()
q = """
- select department_id, sum(pageviews::int) views, sum(visitors::int) visits
+ select department_id, sum(pageviews::int) views, sum(visits::int) visits
from ga_url
- where department_id <> ''"""
- if c.month:
- q = q + """
- and period_name=%s
- """
- q = q + """
- group by department_id order by visits desc
+ where department_id <> ''
+ and package_id <> ''
+ and url like '/dataset/%%'
+ and period_name=%s
+ group by department_id order by views desc
"""
if limit:
q = q + " limit %s;" % (limit)
- # Add this back (before and period_name =%s) if you want to ignore publisher
- # homepage views
- # and not url like '/publisher/%%'
-
top_publishers = []
- res = connection.execute(q, c.month)
-
+ res = connection.execute(q, month)
for row in res:
g = model.Group.get(row[0])
if g:
top_publishers.append((g, row[1], row[2]))
return top_publishers
+
+def _get_top_publishers_graph(limit=20):
+ '''
+ Returns a list of the top 20 publishers by dataset visits.
+ (The number to show can be varied with 'limit')
+ '''
+ connection = model.Session.connection()
+ q = """
+ select department_id, sum(pageviews::int) views
+ from ga_url
+ where department_id <> ''
+ and package_id <> ''
+ and url like '/dataset/%%'
+ and period_name='All'
+ group by department_id order by views desc
+ """
+ if limit:
+ q = q + " limit %s;" % (limit)
+
+ res = connection.execute(q)
+ department_ids = [ row[0] for row in res ]
+
+ # Query for a history graph of these department ids
+ q = model.Session.query(
+ GA_Url.department_id,
+ GA_Url.period_name,
+ func.sum(cast(GA_Url.pageviews,sqlalchemy.types.INT)))\
+ .filter( GA_Url.department_id.in_(department_ids) )\
+ .filter( GA_Url.url.like('/dataset/%') )\
+ .filter( GA_Url.package_id!='' )\
+ .group_by( GA_Url.department_id, GA_Url.period_name )
+ graph_dict = {}
+ for dept_id,period_name,views in q:
+ graph_dict[dept_id] = graph_dict.get( dept_id, {
+ 'name' : model.Group.get(dept_id).title,
+ 'raw' : {}
+ })
+ graph_dict[dept_id]['raw'][period_name] = views
+ return [ graph_dict[id] for id in department_ids ]
+
+
def _get_publishers():
'''
Returns a list of all publishers. Each item is a tuple:
- (names, title)
+ (name, title)
'''
publishers = []
for pub in model.Session.query(model.Group).\
- filter(model.Group.type=='publisher').\
+ filter(model.Group.type=='organization').\
filter(model.Group.state=='active').\
order_by(model.Group.name):
publishers.append((pub.name, pub.title))
--- a/ckanext/ga_report/download_analytics.py
+++ b/ckanext/ga_report/download_analytics.py
@@ -1,9 +1,12 @@
import os
import logging
import datetime
+import httplib
import collections
+import requests
+import json
from pylons import config
-
+from ga_model import _normalize_url
import ga_model
#from ga_client import GA
@@ -13,15 +16,19 @@
FORMAT_MONTH = '%Y-%m'
MIN_VIEWS = 50
MIN_VISITS = 20
+MIN_DOWNLOADS = 10
class DownloadAnalytics(object):
'''Downloads and stores analytics info'''
- def __init__(self, service=None, profile_id=None, delete_first=False):
+ def __init__(self, service=None, token=None, profile_id=None, delete_first=False,
+ skip_url_stats=False):
self.period = config['ga-report.period']
self.service = service
self.profile_id = profile_id
self.delete_first = delete_first
+ self.skip_url_stats = skip_url_stats
+ self.token = token
def specific_month(self, date):
import calendar
@@ -29,6 +36,11 @@
first_of_this_month = datetime.datetime(date.year, date.month, 1)
_, last_day_of_month = calendar.monthrange(int(date.year), int(date.month))
last_of_this_month = datetime.datetime(date.year, date.month, last_day_of_month)
+ # if this is the latest month, note that it is only up until today
+ now = datetime.datetime.now()
+ if now.year == date.year and now.month == date.month:
+ last_day_of_month = now.day
+ last_of_this_month = now
periods = ((date.strftime(FORMAT_MONTH),
last_day_of_month,
first_of_this_month, last_of_this_month),)
@@ -92,29 +104,47 @@
def download_and_store(self, periods):
for period_name, period_complete_day, start_date, end_date in periods:
+ log.info('Period "%s" (%s - %s)',
+ self.get_full_period_name(period_name, period_complete_day),
+ start_date.strftime('%Y-%m-%d'),
+ end_date.strftime('%Y-%m-%d'))
+
if self.delete_first:
- log.info('Deleting existing Analytics for period "%s"',
+ log.info('Deleting existing Analytics for this period "%s"',
period_name)
ga_model.delete(period_name)
- log.info('Downloading Analytics for period "%s" (%s - %s)',
- self.get_full_period_name(period_name, period_complete_day),
- start_date.strftime('%Y %m %d'),
- end_date.strftime('%Y %m %d'))
-
- data = self.download(start_date, end_date, '~/dataset/[a-z0-9-_]+')
- log.info('Storing Dataset Analytics for period "%s"',
- self.get_full_period_name(period_name, period_complete_day))
- self.store(period_name, period_complete_day, data, )
-
- data = self.download(start_date, end_date, '~/publisher/[a-z0-9-_]+')
- log.info('Storing Publisher Analytics for period "%s"',
- self.get_full_period_name(period_name, period_complete_day))
- self.store(period_name, period_complete_day, data,)
-
- ga_model.update_publisher_stats(period_name) # about 30 seconds.
- self.sitewide_stats( period_name )
-
+
+ if not self.skip_url_stats:
+ # Clean out old url data before storing the new
+ ga_model.pre_update_url_stats(period_name)
+
+ accountName = config.get('googleanalytics.account')
+
+ log.info('Downloading analytics for dataset views')
+ data = self.download(start_date, end_date, '~/%s/dataset/[a-z0-9-_]+' % accountName)
+
+ log.info('Storing dataset views (%i rows)', len(data.get('url')))
+ self.store(period_name, period_complete_day, data, )
+
+ log.info('Downloading analytics for publisher views')
+ data = self.download(start_date, end_date, '~/%s/publisher/[a-z0-9-_]+' % accountName)
+
+ log.info('Storing publisher views (%i rows)', len(data.get('url')))
+ self.store(period_name, period_complete_day, data,)
+
+ # Make sure the All records are correct.
+ ga_model.post_update_url_stats()
+
+ log.info('Associating datasets with their publisher')
+ ga_model.update_publisher_stats(period_name) # about 30 seconds.
+
+
+ log.info('Downloading and storing analytics for site-wide stats')
+ self.sitewide_stats( period_name, period_complete_day )
+
+ log.info('Downloading and storing analytics for social networks')
self.update_social_info(period_name, start_date, end_date)
+
def update_social_info(self, period_name, start_date, end_date):
start_date = start_date.strftime('%Y-%m-%d')
@@ -123,69 +153,102 @@
metrics = 'ga:entrances'
sort = '-ga:entrances'
- # Supported query params at
- # https://developers.google.com/analytics/devguides/reporting/core/v3/reference
- results = self.service.data().ga().get(
- ids='ga:' + self.profile_id,
- filters=query,
- start_date=start_date,
- metrics=metrics,
- sort=sort,
- dimensions="ga:landingPagePath,ga:socialNetwork",
- max_results=10000,
- end_date=end_date).execute()
+ try:
+ # Because of issues of invalid responses, we are going to make these requests
+ # ourselves.
+ headers = {'authorization': 'Bearer ' + self.token}
+
+ args = dict(ids='ga:' + self.profile_id,
+ filters=query,
+ metrics=metrics,
+ sort=sort,
+ dimensions="ga:landingPagePath,ga:socialNetwork",
+ max_results=10000)
+
+ args['st