Update to handle missing downloads data
Update to handle missing downloads data

file:a/.gitignore -> file:b/.gitignore
--- a/.gitignore
+++ b/.gitignore
@@ -1,6 +1,7 @@
 *.py[co]
 *.py~
 .gitignore
+ckan.log
 
 # Packages
 *.egg

file:a/README.rst -> file:b/README.rst
--- a/README.rst
+++ b/README.rst
@@ -32,11 +32,11 @@
 
       googleanalytics.id = UA-1010101-1
       googleanalytics.account = Account name (e.g. data.gov.uk, see top level item at https://www.google.com/analytics)
+      googleanalytics.token.filepath = ~/pyenv/token.dat
       ga-report.period = monthly
-      ga-report.bounce_url = /data
+      ga-report.bounce_url = /
 
-   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 /.
+   The ga-report.bounce_url specifies a particular path to record the bounce rate for. Typically it is / (the home page).
 
 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)::
 
@@ -83,13 +83,17 @@
 
     $ paster getauthtoken --config=../ckan/development.ini
 
+Now ensure you reference the correct path to your token.dat in your CKAN config file (e.g. development.ini)::
+
+    googleanalytics.token.filepath = ~/pyenv/token.dat
+
 
 Tutorial
 --------
 
 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
+    $ paster loadanalytics latest --config=../ckan/development.ini
 
 The value after the token file is how much data you want to retrieve, this can be
 

--- a/ckanext/ga_report/command.py
+++ b/ckanext/ga_report/command.py
@@ -1,5 +1,8 @@
 import logging
 import datetime
+import os
+
+from pylons import config
 
 from ckan.lib.cli import CkanCommand
 # No other CKAN imports allowed until _load_config is run,
@@ -58,20 +61,17 @@
     """Get data from Google Analytics API and save it
     in the ga_model
 
-    Usage: paster loadanalytics <tokenfile> <time-period>
+    Usage: paster loadanalytics <time-period>
 
-    Where <tokenfile> is the name of the auth token file from
-    the getauthtoken step.
-
-    And where <time-period> is:
+    Where <time-period> is:
         all         - data for all time
         latest      - (default) just the 'latest' data
         YYYY-MM     - just data for the specific month
     """
     summary = __doc__.split('\n')[0]
     usage = __doc__
-    max_args = 2
-    min_args = 1
+    max_args = 1
+    min_args = 0
 
     def __init__(self, name):
         super(LoadAnalytics, self).__init__(name)
@@ -80,6 +80,11 @@
                                default=False,
                                dest='delete_first',
                                help='Delete data for the period first')
+        self.parser.add_option('-s', '--skip_url_stats',
+                               action='store_true',
+                               default=False,
+                               dest='skip_url_stats',
+                               help='Skip the download of URL data - just do site-wide stats')
 
     def command(self):
         self._load_config()
@@ -87,18 +92,25 @@
         from download_analytics import DownloadAnalytics
         from ga_auth import (init_service, get_profile_id)
 
+        ga_token_filepath = os.path.expanduser(config.get('googleanalytics.token.filepath', ''))
+        if not ga_token_filepath:
+            print 'ERROR: In the CKAN config you need to specify the filepath of the ' \
+                  'Google Analytics token file under key: googleanalytics.token.filepath'
+            return
+
         try:
-            svc = init_service(self.args[0], None)
+            svc = init_service(ga_token_filepath, None)
         except TypeError:
             print ('Have you correctly run the getauthtoken task and '
-                   'specified the correct token file?')
+                   'specified the correct token file in the CKAN config under '
+                   '"googleanalytics.token.filepath"?')
             return
 
         downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc),
-                                       delete_first=self.options.delete_first)
+                                       delete_first=self.options.delete_first,
+                                       skip_url_stats=self.options.skip_url_stats)
 
-        time_period = self.args[1] if self.args and len(self.args) > 1 \
-            else 'latest'
+        time_period = self.args[0] if self.args else 'latest'
         if time_period == 'all':
             downloader.all_()
         elif time_period == 'latest':

--- a/ckanext/ga_report/controller.py
+++ b/ckanext/ga_report/controller.py
@@ -22,12 +22,29 @@
 
 
 def _month_details(cls):
-    '''Returns a list of all the month names'''
+    '''
+    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).filter(cls.period_name!='All').distinct().all()
+    day = None
+
+    vals = model.Session.query(cls.period_name,cls.period_complete_day)\
+        .filter(cls.period_name!='All').distinct(cls.period_name)\
+        .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):
@@ -35,7 +52,7 @@
     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()
@@ -52,11 +69,31 @@
                              entry.key.encode('utf-8'),
                              entry.value.encode('utf-8')])
 
+    def csv_downloads(self, month):
+        import csv
+
+        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.key').all()
+
+        response.headers['Content-Type'] = "text/csv; charset=utf-8"
+        response.headers['Content-Disposition'] = str('attachment; filename=downloads_%s.csv' % (month,))
+
+        writer = csv.writer(response)
+        writer.writerow(["Period", "Resource URL", "Count"])
+
+        for entry in entries:
+            writer.writerow([entry.period_name.encode('utf-8'),
+                             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'
@@ -71,13 +108,13 @@
         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)
@@ -97,7 +134,7 @@
                 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))
                 key, val = clean_key(k,v)
 
                 c.global_totals.append((key, val))
@@ -162,6 +199,37 @@
 
         return render('ga_report/site/index.html')
 
+    def downloads(self):
+
+        # Get the month details by fetching distinct values and determining the
+        # month names from the values.
+        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'
+        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])
+
+        c.downloads = []
+        q = model.Session.query(GA_Stat).filter(GA_Stat.stat_name=='Downloads')
+        q = q.filter(GA_Stat.period_name==c.month) if c.month else q
+        q = q.order_by("ga_stat.value::int desc")
+
+        data = collections.defaultdict(int)
+        for entry in q.all():
+            r = model.Session.query(model.Resource).filter(model.Resource.url==entry.key).first()
+            if not r:
+                continue
+            data[r] += int(entry.value)
+
+        for k,v in data.iteritems():
+            c.downloads.append((k,v))
+
+        c.downloads = sorted(c.downloads, key=operator.itemgetter(1), reverse=True)
+
+        return render('ga_report/site/downloads.html')
+
 
 class GaDatasetReport(BaseController):
     """
@@ -220,7 +288,7 @@
 
         # 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', '')
@@ -232,7 +300,7 @@
         return render('ga_report/publisher/index.html')
 
     def _get_packages(self, publisher=None, count=-1):
-        '''Returns the datasets in order of visits'''
+        '''Returns the datasets in order of views'''
         if count == -1:
             count = sys.maxint
 
@@ -244,7 +312,7 @@
         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.visits::int desc')
+        q = q.order_by('ga_url.pageviews::int desc')
         top_packages = []
         for entry,package in q.limit(count):
             if package:
@@ -278,7 +346,7 @@
 
         # 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', '')
@@ -312,7 +380,7 @@
           and package_id <> ''
           and url like '/dataset/%%'
           and period_name=%s
-        group by department_id order by visits desc
+        group by department_id order by views desc
         """
     if limit:
         q = q + " limit %s;" % (limit)
@@ -329,7 +397,7 @@
 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).\

--- a/ckanext/ga_report/download_analytics.py
+++ b/ckanext/ga_report/download_analytics.py
@@ -13,15 +13,18 @@
 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, 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
 
     def specific_month(self, date):
         import calendar
@@ -102,28 +105,29 @@
                          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, )
-
-            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.
+            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,)
+
+                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 )
+            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)
@@ -150,7 +154,8 @@
         data = collections.defaultdict(list)
         rows = results.get('rows',[])
         for row in rows:
-            data[_normalize_url(row[0])].append( (row[1], int(row[2]),) )
+            url = _normalize_url('http:/' + row[0])
+            data[url].append( (row[1], int(row[2]),) )
         ga_model.update_social(period_name, data)
 
 
@@ -177,8 +182,12 @@
         packages = []
         for entry in results.get('rows'):
             (loc,pageviews,visits) = entry
-            url = _normalize_url('http:/' + loc)
+            url = _normalize_url('http:/' + loc) # strips off domain e.g. www.data.gov.uk or data.gov.uk
+
             if not url.startswith('/dataset/') and not url.startswith('/publisher/'):
+                # filter out strays like:
+                # /data/user/login?came_from=http://data.gov.uk/dataset/os-code-point-open
+                # /403.html?page=/about&from=http://data.gov.uk/publisher/planning-inspectorate
                 continue
             packages.append( (url, pageviews, visits,) ) # Temporary hack
         return dict(url=packages)
@@ -187,7 +196,7 @@
         if 'url' in data:
             ga_model.update_url_stats(period_name, period_complete_day, data['url'])
 
-    def sitewide_stats(self, period_name):
+    def sitewide_stats(self, period_name, period_complete_day):
         import calendar
         year, month = period_name.split('-')
         _, last_day_of_month = calendar.monthrange(int(year), int(month))
@@ -195,10 +204,10 @@
         start_date = '%s-01' % period_name
         end_date = '%s-%s' % (period_name, last_day_of_month)
         funcs = ['_totals_stats', '_social_stats', '_os_stats',
-                 '_locale_stats', '_browser_stats', '_mobile_stats']
+                 '_locale_stats', '_browser_stats', '_mobile_stats', '_download_stats']
         for f in funcs:
             log.info('Downloading analytics for %s' % f.split('_')[1])
-            getattr(self, f)(start_date, end_date, period_name)
+            getattr(self, f)(start_date, end_date, period_name, period_complete_day)
 
     def _get_results(result_data, f):
         data = {}
@@ -207,7 +216,7 @@
             data[key] = data.get(key,0) + result[1]
         return data
 
-    def _totals_stats(self, start_date, end_date, period_name):
+    def _totals_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Fetches distinct totals, total pageviews etc """
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -217,7 +226,8 @@
                                  max_results=10000,
                                  end_date=end_date).execute()
         result_data = results.get('rows')
-        ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]})
+        ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]},
+            period_complete_day)
 
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -232,30 +242,33 @@
             '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:pageviews',
+        ga_model.update_sitewide_stats(period_name, "Totals", data, period_complete_day)
+
+        # Bounces from / or another configurable page.
+        path = '/%s%s' % (config.get('googleanalytics.account'),
+                          config.get('ga-report.bounce_url', '/'))
+        results = self.service.data().ga().get(
+                                 ids='ga:' + self.profile_id,
+                                 filters='ga:pagePath==%s' % (path,),
+                                 start_date=start_date,
+                                 metrics='ga:visitBounceRate',
                                  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):
+        if not result_data or len(result_data) != 1:
+            log.error('Could not pinpoint the bounces for path: %s. Got results: %r',
+                      path, result_data)
+            return
+        results = result_data[0]
+        bounces = float(results[1])
+        # visitBounceRate is already a %
+        log.info('Google reports visitBounceRate as %s', bounces)
+        ga_model.update_sitewide_stats(period_name, "Totals", {'Bounce rate (home page)': float(bounces)},
+            period_complete_day)
+
+
+    def _locale_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Fetches stats about language and country """
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -270,16 +283,41 @@
         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)
+        ga_model.update_sitewide_stats(period_name, "Languages", data, period_complete_day)
 
         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)
-
-
-    def _social_stats(self, start_date, end_date, period_name):
+        ga_model.update_sitewide_stats(period_name, "Country", data, period_complete_day)
+
+
+    def _download_stats(self, start_date, end_date, period_name, period_complete_day):
+        """ Fetches stats about language and country """
+        results = self.service.data().ga().get(
+                                 ids='ga:' + self.profile_id,
+                                 start_date=start_date,
+                                 filters='ga:eventAction==download',
+                                 metrics='ga:totalEvents',
+                                 sort='-ga:totalEvents',
+                                 dimensions="ga:eventLabel",
+                                 max_results=10000,
+                                 end_date=end_date).execute()
+        result_data = results.get('rows')
+        if not result_data:
+            # We may not have data for this time period, so we need to bail
+            # early.
+            log.info("There is no downloads data for this time period")
+            return
+
+        # [[url, count], [url],count]
+        data = {}
+        for result in result_data:
+            data[result[0]] = data.get(result[0], 0) + int(result[1])
+        self._filter_out_long_tail(data, MIN_DOWNLOADS)
+        ga_model.update_sitewide_stats(period_name, "Downloads", data, period_complete_day)
+
+    def _social_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Finds out which social sites people are referred from """
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -295,10 +333,10 @@
             if not result[0] == '(not set)':
                 data[result[0]] = data.get(result[0], 0) + int(result[2])
         self._filter_out_long_tail(data, 3)
-        ga_model.update_sitewide_stats(period_name, "Social sources", data)
-
-
-    def _os_stats(self, start_date, end_date, period_name):
+        ga_model.update_sitewide_stats(period_name, "Social sources", data, period_complete_day)
+
+
+    def _os_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Operating system stats """
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -313,17 +351,17 @@
         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)
+        ga_model.update_sitewide_stats(period_name, "Operating Systems", data, period_complete_day)
 
         data = {}
         for result in result_data:
             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)
-
-
-    def _browser_stats(self, start_date, end_date, period_name):
+        ga_model.update_sitewide_stats(period_name, "Operating Systems versions", data, period_complete_day)
+
+
+    def _browser_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Information about browsers and browser versions """
         results = self.service.data().ga().get(
                                  ids='ga:' + self.profile_id,
@@ -340,14 +378,14 @@
         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)
+        ga_model.update_sitewide_stats(period_name, "Browsers", data, period_complete_day)
 
         data = {}
         for result in result_data:
             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)
+        ga_model.update_sitewide_stats(period_name, "Browser versions", data, period_complete_day)
 
     @classmethod
     def _filter_browser_version(cls, browser, version_str):
@@ -371,7 +409,7 @@
                 ver = ver[0] + ver[1] + 'X' * num_hidden_digits
         return ver
 
-    def _mobile_stats(self, start_date, end_date, period_name):
+    def _mobile_stats(self, start_date, end_date, period_name, period_complete_day):
         """ Info about mobile devices """
 
         results = self.service.data().ga().get(
@@ -388,13 +426,13 @@
         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)
+        ga_model.update_sitewide_stats(period_name, "Mobile brands", data, period_complete_day)
 
         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)
+        ga_model.update_sitewide_stats(period_name, "Mobile devices", data, period_complete_day)
 
     @classmethod
     def _filter_out_long_tail(cls, data, threshold=10):

--- a/ckanext/ga_report/ga_model.py
+++ b/ckanext/ga_report/ga_model.py
@@ -9,6 +9,8 @@
 
 import ckan.model as model
 from ckan.lib.base import *
+
+log = __import__('logging').getLogger(__name__)
 
 def make_uuid():
     return unicode(uuid.uuid4())
@@ -45,6 +47,7 @@
                   Column('id', types.UnicodeText, primary_key=True,
                          default=make_uuid),
                   Column('period_name', types.UnicodeText),
+                  Column('period_complete_day', types.UnicodeText),
                   Column('stat_name', types.UnicodeText),
                   Column('key', types.UnicodeText),
                   Column('value', types.UnicodeText), )
@@ -111,9 +114,7 @@
     >>> normalize_url('http://data.gov.uk/dataset/weekly_fuel_prices')
     '/dataset/weekly_fuel_prices'
     '''
-    # Deliberately leaving a /
-    url = url.replace('http:/','')
-    return '/' + '/'.join(url.split('/')[2:])
+    return '/' + '/'.join(url.split('/')[3:])
 
 
 def _get_package_and_publisher(url):
@@ -134,7 +135,7 @@
             return None, publisher_match.groups()[0]
     return None, None
 
-def update_sitewide_stats(period_name, stat_name, data):
+def update_sitewide_stats(period_name, stat_name, data, period_complete_day):
     for k,v in data.iteritems():
         item = model.Session.query(GA_Stat).\
             filter(GA_Stat.period_name==period_name).\
@@ -144,11 +145,13 @@
             item.period_name = period_name
             item.key = k
             item.value = v
+            item.period_complete_day = period_complete_day
             model.Session.add(item)
         else:
             # create the row
             values = {'id': make_uuid(),
                      'period_name': period_name,
+                     'period_complete_day': period_complete_day,
                      'key': k,
                      'value': v,
                      'stat_name': stat_name
@@ -158,14 +161,28 @@
 
 
 def pre_update_url_stats(period_name):
+    log.debug("Deleting '%s' records" % 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()
+
+    count = model.Session.query(GA_Url).\
+            filter(GA_Url.period_name == 'All').count()
+    log.debug("Deleting %d 'All' records" % count)
+    count = model.Session.query(GA_Url).\
+            filter(GA_Url.period_name == 'All').delete()
+    log.debug("Deleted %d 'All' records" % count)
+
+    model.Session.flush()
+    model.Session.commit()
+    model.repo.commit_and_remove()
 
 
 def update_url_stats(period_name, period_complete_day, url_data):
-
+    '''
+    Given a list of urls and number of hits for each during a given period,
+    stores them in GA_Url under the period and recalculates the totals for
+    the 'All' period.
+    '''
     for url, views, visits in url_data:
         package, publisher = _get_package_and_publisher(url)
 
@@ -209,8 +226,8 @@
                       'period_name': 'All',
                       'period_complete_day': 0,
                       'url': url,
-                      'pageviews': sum([int(e.pageviews) for e in entries]) + old_pageviews,
-                      'visits': sum([int(e.visits) for e in entries]) + old_visits,
+                      'pageviews': sum([int(e.pageviews) for e in entries]) + int(old_pageviews),
+                      'visits': sum([int(e.visits or 0) for e in entries]) + int(old_visits),
                       'department_id': publisher,
                       'package_id': package
                      }
@@ -336,8 +353,39 @@
     '''
     for object_type in (GA_Url, GA_Stat, GA_Publisher, GA_ReferralStat):
         q = model.Session.query(object_type)
-        if period_name != 'all':
+        if period_name != 'All':
             q = q.filter_by(period_name=period_name)
         q.delete()
-    model.Session.commit()
-
+    model.repo.commit_and_remove()
+
+def get_score_for_dataset(dataset_name):
+    '''
+    Returns a "current popularity" score for a dataset,
+    based on how many views it has had recently.
+    '''
+    import datetime
+    now = datetime.datetime.now()
+    last_month = now - datetime.timedelta(days=30)
+    period_names = ['%s-%02d' % (last_month.year, last_month.month),
+                    '%s-%02d' % (now.year, now.month),
+                    ]
+
+    score = 0
+    for period_name in period_names:
+        score /= 2 # previous periods are discounted by 50%
+        entry = model.Session.query(GA_Url)\
+                .filter(GA_Url.period_name==period_name)\
+                .filter(GA_Url.package_id==dataset_name).first()
+        # score
+        if entry:
+            views = float(entry.pageviews)
+            if entry.period_complete_day:
+                views_per_day = views / entry.period_complete_day
+            else:
+                views_per_day = views / 15 # guess
+            score += views_per_day
+
+    score = int(score * 100)
+    log.debug('Popularity %s: %s', score, dataset_name)
+    return score
+

--- a/ckanext/ga_report/helpers.py
+++ b/ckanext/ga_report/helpers.py
@@ -50,9 +50,12 @@
             dataset = model.Package.get(ga_url.url[len('/dataset/'):])
             if dataset and not dataset.state == 'active':
                 dataset = None
-                count += 1
-                if count > 10:
-                    break
+            # When testing, it is possible that top datasets are not available
+            # so only go round this loop a few times before falling back on
+            # a random dataset.
+            count += 1
+            if count > 10:
+                break
     if not dataset:
         # fallback
         dataset = model.Session.query(model.Package)\
@@ -60,7 +63,8 @@
         if not dataset:
             return None
     dataset_dict = get_action('package_show')({'model': model,
-                                               'session': model.Session},
+                                               'session': model.Session,
+                                               'validate': False},