Merge branch 'master' of github.com:datagovuk/ckanext-ga-report
Merge branch 'master' of github.com:datagovuk/ckanext-ga-report

file:a/README.rst -> file:b/README.rst
--- a/README.rst
+++ b/README.rst
@@ -31,14 +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
 
    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 /.
-
-   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.
 
 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)::
 
@@ -47,6 +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
@@ -83,7 +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
@@ -9,7 +9,7 @@
 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')
 
@@ -22,8 +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)
@@ -135,29 +136,7 @@
             c.social_referrer_totals.append((shorten_name(entry[0]), fill_out_url(entry[0]),'',
                                             entry[1]))
 
-
-        browser_version_re = re.compile("(.*)\((.*)\)")
         for k, v in keys.iteritems():
-
-            def clean_field(key):
-                if k != 'Browser versions':
-                    return key
-                m = browser_version_re.match(key)
-                browser = m.groups()[0].strip()
-                ver = m.groups()[1]
-                parts = ver.split('.')
-                if len(parts) > 1:
-                    if parts[1][0] == '0':
-                        ver = parts[0]
-                    else:
-                        ver = "%s.%s" % (parts[0],parts[1])
-                if browser in ['Safari','Android Browser']:  # Special case complex version nums
-                    ver = parts[0]
-                    if len(ver) > 2:
-                        ver = "%s%sX" % (ver[0], ver[1])
-
-                return "%s (%s)" % (browser, ver,)
-
             q = model.Session.query(GA_Stat).\
                 filter(GA_Stat.stat_name==k)
             if c.month:
@@ -167,7 +146,7 @@
 
             d = collections.defaultdict(int)
             for e in q.all():
-                d[clean_field(e.key)] += int(e.value)
+                d[e.key] += int(e.value)
             entries = []
             for key, val in d.iteritems():
                 entries.append((key,val,))
@@ -184,52 +163,60 @@
         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 csv(self, month):
-
-        c.month = month if not month =='all' else ''
+    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", "Views", "Visits", "Period Name"])
-
-        for publisher,view,visit in _get_publishers(None):
+        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 publisher_csv(self, id, month):
-
-        c.month = month if not month =='all' else ''
-        c.publisher = model.Group.get(id)
-        if not c.publisher:
-            abort(404, 'A publisher with that name could not be found')
+    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=%s_%s.csv' % (c.publisher.name, month,))
+            str('attachment; filename=datasets_%s_%s.csv' % (c.publisher_name, month,))
 
         writer = csv.writer(response)
-        writer.writerow(["Publisher", "Views", "Visits", "Period Name"])
+        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 index(self):
+    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.
@@ -241,52 +228,52 @@
         if c.month:
             c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
 
-        c.top_publishers = _get_publishers()
-
+        c.top_publishers = _get_top_publishers()
         return render('ga_report/publisher/index.html')
 
-
-    def _get_packages(self, publisher, count=-1):
+    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 = []
-        q =  model.Session.query(GA_Url).\
-            filter(GA_Url.department_id==publisher.name).\
-            filter(GA_Url.url.like('/dataset/%'))
-        if c.month:
-            q = q.filter(GA_Url.period_name==c.month)
-        q = q.order_by('ga_url.pageviews::int desc')
-
-        if c.month:
-            for entry in q[:count]:
-                p = model.Package.get(entry.url[len('/dataset/'):])
-                top_packages.append((p,entry.pageviews,entry.visitors))
-        else:
-            ds = {}
-            for entry in q.all():
-                if len(ds) >= count:
-                    break
-                p = model.Package.get(entry.url[len('/dataset/'):])
-                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)
-
-            results = []
-            for k, v in ds.iteritems():
-                results.append((k,v['views'],v['visits']))
-
-            top_packages = sorted(results, key=operator.itemgetter(1), reverse=True)
+        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, id):
+    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.publisher = model.Group.get(id)
-        if not c.publisher:
-            abort(404, 'A publisher with that name could not be found')
+        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
@@ -300,43 +287,38 @@
         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)
+            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
 
         c.top_packages = self._get_packages(c.publisher, 20)
 
         return render('ga_report/publisher/read.html')
 
-def _get_publishers(limit=20):
+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 <> ''"""
-    if c.month:
-        q = q + """
-                and period_name=%s
-        """
-    q = q + """
-            group by department_id order by views desc
+        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)
 
-    # 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:
@@ -344,6 +326,19 @@
     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
@@ -3,7 +3,7 @@
 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
@@ -90,25 +92,42 @@
 
     def download_and_store(self, periods):
         for period_name, period_complete_day, start_date, end_date in periods:
-            log.info('Downloading Analytics for period "%s" (%s - %s)',
+            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'))
-
-            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))
+                     start_date.strftime('%Y-%m-%d'),
+                     end_date.strftime('%Y-%m-%d'))
+ 
+            if self.delete_first:
+                log.info('Deleting existing Analytics for this period "%s"',
+                         period_name)
+                ga_model.delete(period_name)
+
+            # Clean up the entries before we run this
+            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, )
 
-            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))
+            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,)
 
+            log.info('Aggregating datasets by 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 )
 
+            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')
@@ -131,17 +150,16 @@
         data = collections.defaultdict(list)
         rows = results.get('rows',[])
         for row in rows:
-            from ga_model import _normalize_url
             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='~/dataset/[a-z0-9-_]+'):
+    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
@@ -159,7 +177,10 @@
         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):
@@ -173,12 +194,10 @@
 
         start_date = '%s-01' % period_name
         end_date = '%s-%s' % (period_name, last_day_of_month)
-        print 'Sitewide_stats for %s (%s -> %s)' % (period_name, start_date, end_date)
-
         funcs = ['_totals_stats', '_social_stats', '_os_stats',
                  '_locale_stats', '_browser_stats', '_mobile_stats']
         for f in funcs:
-            print ' + Fetching %s stats' % f.split('_')[1]
+            log.info('Downloading analytics for %s' % f.split('_')[1])
             getattr(self, f)(start_date, end_date, period_name)
 
     def _get_results(result_data, f):
@@ -203,7 +222,7 @@
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
                                  start_date=start_date,
-                                 metrics='ga:pageviewsPerVisit,ga:avgTimeOnSite,ga:percentNewVisits,ga:visitors',
+                                 metrics='ga:pageviewsPerVisit,ga:avgTimeOnSite,ga:percentNewVisits,ga:visits',
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
@@ -250,11 +269,13 @@
         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)
 
 
@@ -269,13 +290,11 @@
                                  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)
 
 
@@ -293,12 +312,14 @@
         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)
 
 
@@ -313,17 +334,42 @@
                                  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 """
@@ -341,10 +387,23 @@
         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
+++ b/ckanext/ga_report/ga_model.py
@@ -1,10 +1,10 @@
 import re
 import uuid
 
-from sqlalchemy import Table, Column, MetaData
+from sqlalchemy import Table, Column, MetaData, ForeignKey
 from sqlalchemy import types
 from sqlalchemy.sql import select
-from sqlalchemy.orm import mapper
+from sqlalchemy.orm import mapper, relation
 from sqlalchemy import func
 
 import ckan.model as model
@@ -14,8 +14,6 @@
     return unicode(uuid.uuid4())
 
 metadata = MetaData()
-
-
 
 class GA_Url(object):
 
@@ -32,6 +30,7 @@
                       Column('visitors', types.UnicodeText),
                       Column('url', types.UnicodeText),
                       Column('department_id', types.UnicodeText),
+                      Column('package_id', types.UnicodeText),
                 )
 mapper(GA_Url, url_table)
 
@@ -117,7 +116,7 @@
     return '/' + '/'.join(url.split('/')[2:])
 
 
-def _get_department_id_of_url(url):
+def _get_package_and_publisher(url):
     # e.g. /dataset/fuel_prices
     # e.g. /dataset/fuel_prices/resource/e63380d4
     dataset_match = re.match('/dataset/([^/]+)(/.*)?', url)
@@ -127,12 +126,13 @@
         if dataset:
             publisher_groups = dataset.get_groups('publisher')
             if publisher_groups:
-                return publisher_groups[0].name
+                return dataset_ref,publisher_groups[0].name
+        return dataset_ref, None
     else:
         publisher_match = re.match('/publisher/([^/]+)(/.*)?', url)
         if publisher_match:
-            return publisher_match.groups()[0]
-
+            return None, publisher_match.groups()[0]
+    return None, None
 
 def update_sitewide_stats(period_name, stat_name, data):
     for k,v in data.iteritems():
@@ -157,34 +157,76 @@
         model.Session.commit()
 
 
+def update_url_stat_totals(period_name):
+
+    """
+        items = model.Session.query(GA_Url).\
+            filter(GA_Url.period_name != "All").\
+            filter(GA_Url.url==url).all()
+        values = {'id': make_uuid(),
+                  'period_name': "All",
+                  'period_complete_day': "0",
+                  'url': url,
+                  'pageviews': sum([int(x.pageviews) for x in items]),
+                  'visitors': sum([int(x.visitors) for x in items]),
+                  'department_id': department_id,
+                  'package_id': package
+                 }
+        model.Session.add(GA_Url(**values))
+        model.Session.commit()
+    """
+
+def pre_update_url_stats(period_name):
+    model.Session.query(GA_Url).\
+            filter(GA_Url.period_name==period_name).delete()
+    model.Session.query(GA_Url).\
+            filter(GA_Url.period_name=='All').delete()
+
 
 def update_url_stats(period_name, period_complete_day, url_data):
+
     for url, views, visitors in url_data:
-        url = _normalize_url(url)
-        department_id = _get_department_id_of_url(url)
-
-        # see if the row for this url & month is in the table already
+        package, publisher = _get_package_and_publisher(url)
+
         item = model.Session.query(GA_Url).\
             filter(GA_Url.period_name==period_name).\
             filter(GA_Url.url==url).first()
         if item:
-            item.period_name = period_name
-            item.pageviews = views
-            item.visitors = visitors
-            item.department_id = department_id
+            item.pageviews = item.pageviews + views
+            item.visitors = item.visitors + visitors
             model.Session.add(item)
         else:
-            # create the row
             values = {'id': make_uuid(),
                       'period_name': period_name,
                       'period_complete_day': period_complete_day,
                       'url': url,
                       'pageviews': views,
                       'visitors': visitors,
-                      'department_id': department_id
+                      'department_id': publisher,
+                      'package_id': package
                      }
             model.Session.add(GA_Url(**values))
         model.Session.commit()
+
+        if package:
+            entries = model.Session.query(GA_Url).\
+                filter(GA_Url.period_name!='All').\
+                filter(GA_Url.url==url).all()
+            values = {'id': make_uuid(),
+                      'period_name': 'All',
+                      'period_complete_day': 0,
+                      'url': url,
+                      'pageviews': sum([int(e.pageviews) for e in entries]),
+                      'visitors': sum([int(e.visitors) for e in entries]),
+                      'department_id': publisher,
+                      'package_id': package
+                     }
+
+            model.Session.add(GA_Url(**values))
+            model.Session.commit()
+
+
+
 
 
 def update_social(period_name, data):
@@ -295,3 +337,15 @@
         for grandchild in go_down_tree(child):
             yield grandchild
 
+def delete(period_name):
+    '''
+    Deletes table data for the specified period, or specify 'all'
+    for all periods.
+    '''
+    for object_type in (GA_Url, GA_Stat, GA_Publisher, GA_ReferralStat):
+        q = model.Session.query(object_type)
+        if period_name != 'all':
+            q = q.filter_by(period_name=period_name)
+        q.delete()
+    model.Session.commit()
+

--- a/ckanext/ga_report/helpers.py
+++ b/ckanext/ga_report/helpers.py
@@ -1,7 +1,9 @@
 import logging
 import operator
+
 import ckan.lib.base as base
 import ckan.model as model
+from ckan.logic import get_action
 
 from ckanext.ga_report.ga_model import GA_Url, GA_Publisher
 from ckanext.ga_report.controller import _get_publishers
@@ -28,28 +30,50 @@
     return base.render_snippet('ga_report/ga_popular_datasets.html', **ctx)
 
 def single_popular_dataset(top=20):
+    '''Returns a random dataset from the most popular ones.
+
+    :param top: the number of top datasets to select from
+    '''
     import random
 
-    datasets = {}
-    rand = random.randrange(0, top)