Handling duplicate urls in ga_url processing
Handling duplicate urls in ga_url processing

file:a/README.rst -> file:b/README.rst
--- a/README.rst
+++ b/README.rst
@@ -31,10 +31,12 @@
 2. Ensure you development.ini (or similar) contains the info about your Google Analytics account and configuration::
 
       googleanalytics.id = UA-1010101-1
-      googleanalytics.account = Account name (i.e. data.gov.uk, see top level item at https://www.google.com/analytics)
+      googleanalytics.account = Account name (e.g. data.gov.uk, see top level item at https://www.google.com/analytics)
       ga-report.period = monthly
+      ga-report.bounce_url = /data
 
-   Note that your credentials will be readable by system administrators on your server. Rather than use sensitive account details, it is suggested you give access to the GA account to a new Google account that you create just for this purpose.
+   The ga-report.bounce_url specifies the path to use when calculating bounces. For DGU this is /data
+   but you may want to set this to /.
 
 3. Set up this extension's database tables using a paster command. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file)::
 
@@ -43,6 +45,12 @@
 4. Enable the extension in your CKAN config file by adding it to ``ckan.plugins``::
 
     ckan.plugins = ga-report
+
+Problem shooting
+----------------
+
+* ``(ProgrammingError) relation "ga_url" does not exist``
+  This means that the ``paster initdb`` step has not been run successfully. Refer to the installation instructions for this extension.
 
 
 Authorization
@@ -79,7 +87,7 @@
 Tutorial
 --------
 
-Download some GA data and store it in CKAN's db. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file) and specifying the name of your auth file (token.dat by default) from the previous step::
+Download some GA data and store it in CKAN's database. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file) and specifying the name of your auth file (token.dat by default) from the previous step::
 
     $ paster loadanalytics token.dat latest --config=../ckan/development.ini
 

--- a/ckanext/ga_report/command.py
+++ b/ckanext/ga_report/command.py
@@ -73,6 +73,14 @@
     max_args = 2
     min_args = 1
 
+    def __init__(self, name):
+        super(LoadAnalytics, self).__init__(name)
+        self.parser.add_option('-d', '--delete-first',
+                               action='store_true',
+                               default=False,
+                               dest='delete_first',
+                               help='Delete data for the period first')
+
     def command(self):
         self._load_config()
 
@@ -83,10 +91,11 @@
             svc = init_service(self.args[0], None)
         except TypeError:
             print ('Have you correctly run the getauthtoken task and '
-                   'specified the correct file here')
+                   'specified the correct token file?')
             return
 
-        downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc))
+        downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc),
+                                       delete_first=self.options.delete_first)
 
         time_period = self.args[1] if self.args and len(self.args) > 1 \
             else 'latest'

--- a/ckanext/ga_report/controller.py
+++ b/ckanext/ga_report/controller.py
@@ -1,11 +1,15 @@
+import re
+import csv
+import sys
 import logging
 import operator
-from ckan.lib.base import BaseController, c, render, request, response, abort
+import collections
+from ckan.lib.base import (BaseController, c, g, render, request, response, abort)
 
 import sqlalchemy
 from sqlalchemy import func, cast, Integer
 import ckan.model as model
-from ga_model import GA_Url, GA_Stat
+from ga_model import GA_Url, GA_Stat, GA_ReferralStat, GA_Publisher
 
 log = logging.getLogger('ckanext.ga-report')
 
@@ -18,8 +22,9 @@
 
 
 def _month_details(cls):
+    '''Returns a list of all the month names'''
     months = []
-    vals = model.Session.query(cls.period_name).distinct().all()
+    vals = model.Session.query(cls.period_name).filter(cls.period_name!='All').distinct().all()
     for m in vals:
         months.append( (m[0], _get_month_name(m[0])))
     return sorted(months, key=operator.itemgetter(0), reverse=True)
@@ -30,11 +35,13 @@
     def csv(self, month):
         import csv
 
-        entries = model.Session.query(GA_Stat).\
-            filter(GA_Stat.period_name==month).\
-            order_by('GA_Stat.stat_name, GA_Stat.key').all()
+        q = model.Session.query(GA_Stat)
+        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()
 
         response.headers['Content-Type'] = "text/csv; charset=utf-8"
+        response.headers['Content-Disposition'] = str('attachment; filename=stats_%s.csv' % (month,))
 
         writer = csv.writer(response)
         writer.writerow(["Period", "Statistic", "Key", "Value"])
@@ -52,87 +59,221 @@
         c.months = _month_details(GA_Stat)
 
         # Work out which month to show, based on query params of the first item
-        c.month = request.params.get('month', c.months[0][0] if c.months else '')
-        c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
-        entries = model.Session.query(GA_Stat).\
-            filter(GA_Stat.stat_name=='Totals').\
-            filter(GA_Stat.period_name==c.month).\
-            order_by('ga_stat.key').all()
-
-        c.global_totals = []
-        for e in entries:
-            val = e.value
-            if e.key in ['Average time on site', 'Pages per visit', 'Percent new visits']:
-                val =  "%.2f" % round(float(e.value), 2)
-                if e.key == 'Average time on site':
+        c.month_desc = 'all months'
+        c.month = request.params.get('month', '')
+        if c.month:
+            c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
+
+        q = model.Session.query(GA_Stat).\
+            filter(GA_Stat.stat_name=='Totals')
+        if c.month:
+            q = q.filter(GA_Stat.period_name==c.month)
+        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']:
+                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)
-                e.key = '%s *' % e.key
-            c.global_totals.append((e.key, val))
-
+                if key in ['New visits','Bounces']:
+                    val = "%s%%" % val
+            if key in ['Total page views', 'Total visits']:
+                val = int(val)
+
+            return key, val
+
+        c.global_totals = []
+        if c.month:
+            for e in entries:
+                key, val = clean_key(e.key, e.value)
+                c.global_totals.append((key, val))
+        else:
+            d = collections.defaultdict(list)
+            for e in entries:
+                d[e.key].append(float(e.value))
+            for k, v in d.iteritems():
+                if k in ['Total page views', 'Total visits']:
+                    v = sum(v)
+                else:
+                    v = float(sum(v))/len(v)
+                key, val = clean_key(k,v)
+
+                c.global_totals.append((key, val))
+                c.global_totals = sorted(c.global_totals, key=operator.itemgetter(0))
 
         keys = {
-            'Browser versions': 'browsers',
-            'Operating Systems versions': 'os',
+            'Browser versions': 'browser_versions',
+            'Browsers': 'browsers',
+            'Operating Systems versions': 'os_versions',
+            'Operating Systems': 'os',
             'Social sources': 'social_networks',
             'Languages': 'languages',
             'Country': 'country'
         }
 
+        def shorten_name(name, length=60):
+            return (name[:length] + '..') if len(name) > 60 else name
+
+        def fill_out_url(url):
+            import urlparse
+            return urlparse.urljoin(g.site_url, url)
+
+        c.social_referrer_totals, c.social_referrers = [], []
+        q = model.Session.query(GA_ReferralStat)
+        q = q.filter(GA_ReferralStat.period_name==c.month) if c.month else q
+        q = q.order_by('ga_referrer.count::int desc')
+        for entry in q.all():
+            c.social_referrers.append((shorten_name(entry.url), fill_out_url(entry.url),
+                                       entry.source,entry.count))
+
+        q = model.Session.query(GA_ReferralStat.url,
+                                func.sum(GA_ReferralStat.count).label('count'))
+        q = q.filter(GA_ReferralStat.period_name==c.month) if c.month else q
+        q = q.order_by('count desc').group_by(GA_ReferralStat.url)
+        for entry in q.all():
+            c.social_referrer_totals.append((shorten_name(entry[0]), fill_out_url(entry[0]),'',
+                                            entry[1]))
+
         for k, v in keys.iteritems():
-            entries = model.Session.query(GA_Stat).\
-                filter(GA_Stat.stat_name==k).\
-                filter(GA_Stat.period_name==c.month).\
-                order_by('ga_stat.value::int desc').all()
-            setattr(c, v, [(s.key, s.value) for s in entries ])
-
+            q = model.Session.query(GA_Stat).\
+                filter(GA_Stat.stat_name==k)
+            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)
+            entries = []
+            for key, val in d.iteritems():
+                entries.append((key,val,))
+            entries = sorted(entries, key=operator.itemgetter(1), reverse=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'])
+            else:
+                total = sum([num for _,num in entries])
+            setattr(c, v, [(k,_percent(v,total)) for k,v in entries ])
 
         return render('ga_report/site/index.html')
 
 
-class GaPublisherReport(BaseController):
+class GaDatasetReport(BaseController):
     """
-    Displays the pageview and visit count for specific publishers based on
-    the datasets associated with the publisher.
+    Displays the pageview and visit count for datasets
+    with options to filter by publisher and time period.
     """
-
-    def index(self):
+    def publisher_csv(self, month):
+        '''
+        Returns a CSV of each publisher with the total number of dataset
+        views & visits.
+        '''
+        c.month = month if not month == 'all' else ''
+        response.headers['Content-Type'] = "text/csv; charset=utf-8"
+        response.headers['Content-Disposition'] = str('attachment; filename=publishers_%s.csv' % (month,))
+
+        writer = csv.writer(response)
+        writer.writerow(["Publisher Title", "Publisher Name", "Views", "Visits", "Period Name"])
+
+        for publisher,view,visit in _get_top_publishers(None):
+            writer.writerow([publisher.title.encode('utf-8'),
+                             publisher.name.encode('utf-8'),
+                             view,
+                             visit,
+                             month])
+
+    def dataset_csv(self, id='all', month='all'):
+        '''
+        Returns a CSV with the number of views & visits for each dataset.
+
+        :param id: A Publisher ID or None if you want for all
+        :param month: The time period, or 'all'
+        '''
+        c.month = month if not month == 'all' else ''
+        if id != 'all':
+            c.publisher = model.Group.get(id)
+            if not c.publisher:
+                abort(404, 'A publisher with that name could not be found')
+
+        packages = self._get_packages(c.publisher)
+        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([package.title.encode('utf-8'),
+                             package.name.encode('utf-8'),
+                             view,
+                             visit,
+                             month])
+
+    def publishers(self):
+        '''A list of publishers and the number of views/visits for each'''
 
         # Get the month details by fetching distinct values and determining the
         # month names from the values.
         c.months = _month_details(GA_Url)
 
         # Work out which month to show, based on query params of the first item
-        c.month = request.params.get('month', c.months[0][0] if c.months else '')
-        c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
-        connection = model.Session.connection()
-        q = """
-            select department_id, sum(pageviews::int) views, sum(visitors::int) visits
-            from ga_url
-            where department_id <> ''
-                and period_name=%s
-            group by department_id order by views desc limit 20;
-        """
-        # Add this back (before and period_name =%s) if you want to ignore publisher
-        # homepage views
-        # and not url like '/publisher/%%'
-
-        c.top_publishers = []
-        res = connection.execute(q, c.month)
-        for row in res:
-            c.top_publishers.append((model.Group.get(row[0]), row[1], row[2]))
-
+        c.month = request.params.get('month', '')
+        c.month_desc = 'all months'
+        if c.month:
+            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 read(self, id):
-
-        c.publisher = model.Group.get(id)
-        if not c.publisher:
-            abort(404, 'A publisher with that name could not be found')
+    def _get_packages(self, publisher=None, count=-1):
+        '''Returns the datasets in order of visits'''
+        if count == -1:
+            count = sys.maxint
+
+        month = c.month or 'All'
+
+        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)
+        q = q.filter(GA_Url.period_name==month)
+        q = q.order_by('ga_url.visitors::int desc')
+        top_packages = []
+        for entry,package in q.limit(count):
+            if package:
+                top_packages.append((package, entry.pageviews, entry.visitors))
+            else:
+                log.warning('Could not find package associated package')
+
+        return top_packages
+
+    def read(self):
+        '''
+        Lists the most popular datasets across all publishers
+        '''
+        return self.read_publisher(None)
+
+    def read_publisher(self, id):
+        '''
+        Lists the most popular datasets for a publisher (or across all publishers)
+        '''
+        count = 20
+
+        c.publishers = _get_publishers()
+
+        id = request.params.get('publisher', id)
+        if id and id != 'all':
+            c.publisher = model.Group.get(id)
+            if not c.publisher:
+                abort(404, 'A publisher with that name could not be found')
+            c.publisher_name = c.publisher.name
         c.top_packages = [] # package, dataset_views in c.top_packages
 
         # Get the month details by fetching distinct values and determining the
@@ -140,22 +281,65 @@
         c.months = _month_details(GA_Url)
 
         # Work out which month to show, based on query params of the first item
-        c.month = request.params.get('month', c.months[0][0] if c.months else '')
-        c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
-        entry = model.Session.query(GA_Url).\
-            filter(GA_Url.url=='/publisher/%s' % c.publisher.name).\
-            filter(GA_Url.period_name==c.month).first()
+        c.month = request.params.get('month', '')
+        if not c.month:
+            c.month_desc = 'all months'
+        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)
+        entry = q.filter(GA_Url.period_name==c.month).first()
         c.publisher_page_views = entry.pageviews if entry else 0
 
-        entries = model.Session.query(GA_Url).\
-            filter(GA_Url.department_id==c.publisher.name).\
-            filter(GA_Url.period_name==c.month).\
-            order_by('ga_url.pageviews::int desc')[:20]
-        for entry in entries:
-            if entry.url.startswith('/dataset/'):
-                p = model.Package.get(entry.url[len('/dataset/'):])
-                c.top_packages.append((p,entry.pageviews,entry.visitors))
+        c.top_packages = self._get_packages(c.publisher, 20)
 
         return render('ga_report/publisher/read.html')
 
+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
+        from ga_url
+        where department_id <> ''
+          and package_id <> ''
+          and url like '/dataset/%%'
+          and period_name=%s
+        group by department_id order by visits desc
+        """
+    if limit:
+        q = q + " limit %s;" % (limit)
+
+    top_publishers = []
+    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_publishers():
+    '''
+    Returns a list of all publishers. Each item is a tuple:
+      (names, title)
+    '''
+    publishers = []
+    for pub in model.Session.query(model.Group).\
+               filter(model.Group.type=='publisher').\
+               filter(model.Group.state=='active').\
+               order_by(model.Group.name):
+        publishers.append((pub.name, pub.title))
+    return publishers
+
+def _percent(num, total):
+    p = 100 * float(num)/float(total)
+    return "%.2f%%" % round(p, 2)
+

--- a/ckanext/ga_report/download_analytics.py
+++ b/ckanext/ga_report/download_analytics.py
@@ -1,9 +1,9 @@
 import os
 import logging
 import datetime
-
+import collections
 from pylons import config
-
+from ga_model import _normalize_url
 import ga_model
 
 #from ga_client import GA
@@ -11,15 +11,17 @@
 log = logging.getLogger('ckanext.ga-report')
 
 FORMAT_MONTH = '%Y-%m'
+MIN_VIEWS = 50
+MIN_VISITS = 20
 
 class DownloadAnalytics(object):
     '''Downloads and stores analytics info'''
 
-    def __init__(self, service=None, profile_id=None):
+    def __init__(self, service=None, profile_id=None, delete_first=False):
         self.period = config['ga-report.period']
         self.service = service
         self.profile_id = profile_id
-
+        self.delete_first = delete_first
 
     def specific_month(self, date):
         import calendar
@@ -90,17 +92,26 @@
 
     def download_and_store(self, periods):
         for period_name, period_complete_day, start_date, end_date in periods:
+            if self.delete_first:
+                log.info('Deleting existing Analytics for 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-_]+')
+            # Clean up the entries before we run this
+            ga_model.pre_update_url_stats(period_name)
+
+            accountName = config.get('googleanalytics.account')
+
+            data = self.download(start_date, end_date, '~/%s/dataset/[a-z0-9-_]+' % accountName)
             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-_]+')
+            data = self.download(start_date, end_date, '~/%s/publisher/[a-z0-9-_]+' % accountName)
             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,)
@@ -108,13 +119,40 @@
             ga_model.update_publisher_stats(period_name) # about 30 seconds.
             self.sitewide_stats( period_name )
 
-
-    def download(self, start_date, end_date, path='~/dataset/[a-z0-9-_]+'):
+            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')
+        end_date = end_date.strftime('%Y-%m-%d')
+        query = 'ga:hasSocialSourceReferral=~Yes$'
+        metrics = 'ga:entrances'
+        sort = '-ga:entrances'
+
+        # Supported query params at
+        # https://developers.google.com/analytics/devguides/reporting/core/v3/reference
+        results = self.service.data().ga().get(
+                                 ids='ga:' + self.profile_id,
+                                 filters=query,
+                                 start_date=start_date,
+                                 metrics=metrics,
+                                 sort=sort,
+                                 dimensions="ga:landingPagePath,ga:socialNetwork",
+                                 max_results=10000,
+                                 end_date=end_date).execute()
+        data = collections.defaultdict(list)
+        rows = results.get('rows',[])
+        for row in rows:
+            data[_normalize_url(row[0])].append( (row[1], int(row[2]),) )
+        ga_model.update_social(period_name, data)
+
+
+    def download(self, start_date, end_date, path=None):
         '''Get data from GA for a given time period'''
         start_date = start_date.strftime('%Y-%m-%d')
         end_date = end_date.strftime('%Y-%m-%d')
         query = 'ga:pagePath=%s$' % path
-        metrics = 'ga:uniquePageviews, ga:visitors'
+        metrics = 'ga:uniquePageviews, ga:visits'
         sort = '-ga:uniquePageviews'
 
         # Supported query params at
@@ -129,15 +167,13 @@
                                  max_results=10000,
                                  end_date=end_date).execute()
 
-        if os.getenv('DEBUG'):
-            import pprint
-            pprint.pprint(results)
-            print 'Total results: %s' % results.get('totalResults')
-
         packages = []
         for entry in results.get('rows'):
             (loc,pageviews,visits) = entry
-            packages.append( ('http:/' + loc, pageviews, visits,) ) # Temporary hack
+            url = _normalize_url('http:/' + loc)
+            if not url.startswith('/dataset/') and not url.startswith('/publisher/'):
+                continue
+            packages.append( (url, pageviews, visits,) ) # Temporary hack
         return dict(url=packages)
 
     def store(self, period_name, period_complete_day, data):
@@ -176,23 +212,43 @@
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
-        ga_model.update_sitewide_stats(period_name, "Totals", {'Total pageviews': result_data[0][0]})
-
-        results = self.service.data().ga().get(
-                                 ids='ga:' + self.profile_id,
-                                 start_date=start_date,
-                                 metrics='ga:pageviewsPerVisit,ga:bounces,ga:avgTimeOnSite,ga:percentNewVisits',
+        ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]})
+
+        results = self.service.data().ga().get(
+                                 ids='ga:' + self.profile_id,
+                                 start_date=start_date,
+                                 metrics='ga:pageviewsPerVisit,ga:avgTimeOnSite,ga:percentNewVisits,ga:visits',
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
         data = {
             'Pages per visit': result_data[0][0],
-            'Bounces': result_data[0][1],
-            'Average time on site': result_data[0][2],
-            'Percent new visits': result_data[0][3],
+            'Average time on site': result_data[0][1],
+            'New visits': result_data[0][2],
+            'Total visits': result_data[0][3],
         }
         ga_model.update_sitewide_stats(period_name, "Totals", data)
 
+        # Bounces from /data. This url is specified in configuration because
+        # for DGU we don't want /.
+        path = config.get('ga-report.bounce_url','/')
+        print path
+        results = self.service.data().ga().get(
+                                 ids='ga:' + self.profile_id,
+                                 filters='ga:pagePath=~%s$' % (path,),
+                                 start_date=start_date,
+                                 metrics='ga:bounces,ga:uniquePageviews',
+                                 dimensions='ga:pagePath',
+                                 max_results=10000,
+                                 end_date=end_date).execute()
+        result_data = results.get('rows')
+        for results in result_data:
+            if results[0] == path:
+                bounce, total = [float(x) for x in results[1:]]
+                pct = 100 * bounce/total
+                print "%d bounces from %d total == %s" % (bounce, total, pct)
+                ga_model.update_sitewide_stats(period_name, "Totals", {'Bounces': pct})
+
 
     def _locale_stats(self, start_date, end_date, period_name):
         """ Fetches stats about language and country """
@@ -208,11 +264,13 @@
         data = {}
         for result in result_data:
             data[result[0]] = data.get(result[0], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Languages", data)
 
         data = {}
         for result in result_data:
             data[result[1]] = data.get(result[1], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Country", data)
 
 
@@ -227,13 +285,11 @@
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
-        twitter_links = []
         data = {}
         for result in result_data:
             if not result[0] == '(not set)':
                 data[result[0]] = data.get(result[0], 0) + int(result[2])
-                if result[0] == 'Twitter':
-                    twitter_links.append(result[1])
+        self._filter_out_long_tail(data, 3)
         ga_model.update_sitewide_stats(period_name, "Social sources", data)
 
 
@@ -251,12 +307,14 @@
         data = {}
         for result in result_data:
             data[result[0]] = data.get(result[0], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Operating Systems", data)
 
         data = {}
         for result in result_data:
-            key = "%s (%s)" % (result[0],result[1])
-            data[key] = result[2]
+            if int(result[2]) >= MIN_VIEWS:
+                key = "%s %s" % (result[0],result[1])
+                data[key] = result[2]
         ga_model.update_sitewide_stats(period_name, "Operating Systems versions", data)
 
 
@@ -271,17 +329,42 @@
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
+        # e.g. [u'Firefox', u'19.0', u'20']
+
         data = {}
         for result in result_data:
             data[result[0]] = data.get(result[0], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Browsers", data)
 
         data = {}
         for result in result_data:
-            key = "%s (%s)" % (result[0], result[1])
-            data[key] = result[2]
+            key = "%s %s" % (result[0], self._filter_browser_version(result[0], result[1]))
+            data[key] = data.get(key, 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Browser versions", data)
 
+    @classmethod
+    def _filter_browser_version(cls, browser, version_str):
+        '''
+        Simplifies a browser version string if it is detailed.
+        i.e. groups together Firefox 3.5.1 and 3.5.2 to be just 3.
+        This is helpful when viewing stats and good to protect privacy.
+        '''
+        ver = version_str
+        parts = ver.split('.')
+        if len(parts) > 1:
+            if parts[1][0] == '0':
+                ver = parts[0]
+            else:
+                ver = "%s" % (parts[0])
+        # Special case complex version nums
+        if browser in ['Safari', 'Android Browser']:
+            ver = parts[0]
+            if len(ver) > 2:
+                num_hidden_digits = len(ver) - 2
+                ver = ver[0] + ver[1] + 'X' * num_hidden_digits
+        return ver
 
     def _mobile_stats(self, start_date, end_date, period_name):
         """ Info about mobile devices """
@@ -299,10 +382,23 @@
         data = {}
         for result in result_data:
             data[result[0]] = data.get(result[0], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Mobile brands", data)
 
         data = {}
         for result in result_data:
             data[result[1]] = data.get(result[1], 0) + int(result[2])
+        self._filter_out_long_tail(data, MIN_VIEWS)
         ga_model.update_sitewide_stats(period_name, "Mobile devices", data)
 
+    @classmethod
+    def _filter_out_long_tail(cls, data, threshold=10):
+        '''
+        Given data which is a frequency distribution, filter out
+        results which are below a threshold count. This is good to protect
+        privacy.
+        '''
+        for key, value in data.items():
+            if value < threshold:
+                del data[key]
+

--- a/ckanext/ga_report/ga_auth.py
+++ b/ckanext/ga_report/ga_auth.py
@@ -53,7 +53,11 @@
         return None
 
     accountName = config.get('googleanalytics.account')
+    if not accountName:
+        raise Exception('googleanalytics.account needs to be configured')
     webPropertyId = config.get('googleanalytics.id')
+    if not webPropertyId:
+        raise Exception('googleanalytics.id needs to be configured')
     for acc in accounts.get('items'):
         if acc.get('name') == accountName:
             accountId = acc.get('id')

--- a/ckanext/ga_report/ga_model.py