[304] Don't show No Graph Loaded text on Publishers page.
[304] Don't show No Graph Loaded text on Publishers page.

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
@@ -26,16 +26,17 @@
 1. Activate you CKAN python environment and install this extension's software::
 
     $ pyenv/bin/activate
-    $ pip install -e  git+https://github.com/okfn/ckanext-ga-report.git#egg=ckanext-ga-report
+    $ pip install -e  git+https://github.com/datagovuk/ckanext-ga-report.git#egg=ckanext-ga-report
 
 2. Ensure you development.ini (or similar) contains the info about your Google Analytics account and configuration::
 
       googleanalytics.id = UA-1010101-1
-      googleanalytics.username = googleaccount@gmail.com
-      googleanalytics.password = googlepassword
+      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 = /
 
-   Note that your password 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 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)::
 
@@ -44,6 +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
@@ -72,17 +79,30 @@
 
 Once you have set up your credentials.json file you can generate an oauth token file by using the
 following command, which will store your oauth token in a file called token.dat once you have finished
-giving permission in the browser.
+giving permission in the browser::
 
     $ 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 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
+    $ paster loadanalytics latest --config=../ckan/development.ini
+
+The value after the token file is how much data you want to retrieve, this can be
+
+* **all**         - data for all time (since 2010)
+
+* **latest**      - (default) just the 'latest' data
+
+* **YYYY-MM-DD**  - just data for all time periods going back to (and including) this date
+
 
 
 Software Licence

--- a/ckanext/ga_report/command.py
+++ b/ckanext/ga_report/command.py
@@ -1,7 +1,13 @@
 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, or logging is disabled
+# No other CKAN imports allowed until _load_config is run,
+# or logging is disabled
+
 
 class InitDB(CkanCommand):
     """Initialise the extension's database tables
@@ -17,7 +23,7 @@
         import ckan.model as model
         model.Session.remove()
         model.Session.configure(bind=model.meta.engine)
-        log = logging.getLogger('ckanext.ga-report')
+        log = logging.getLogger('ckanext.ga_report')
 
         import ga_model
         ga_model.init_tables()
@@ -26,6 +32,12 @@
 
 class GetAuthToken(CkanCommand):
     """ Get's the Google auth token
+
+    Usage: paster getauthtoken <credentials_file>
+
+    Where <credentials_file> is the file name containing the details
+    for the service (obtained from https://code.google.com/apis/console).
+    By default this is set to credentials.json
     """
     summary = __doc__.split('\n')[0]
     usage = __doc__
@@ -33,52 +45,108 @@
     min_args = 0
 
     def command(self):
-        from ga_auth import initialize_service
-        initialize_service('token.dat',
-                           self.args[0] if self.args
-                                        else 'credentials.json')
+        """
+        In this case we don't want a valid service, but rather just to
+        force the user through the auth flow. We allow this to complete to
+        act as a form of verification instead of just getting the token and
+        assuming it is correct.
+        """
+        from ga_auth import init_service
+        init_service('token.dat',
+                      self.args[0] if self.args
+                                   else 'credentials.json')
+
+class FixTimePeriods(CkanCommand):
+    """
+    Fixes the 'All' records for GA_Urls
+
+    It is possible that older urls that haven't recently been visited
+    do not have All records.  This command will traverse through those
+    records and generate valid All records for them.
+    """
+    summary = __doc__.split('\n')[0]
+    usage = __doc__
+    max_args = 0
+    min_args = 0
+
+    def __init__(self, name):
+        super(FixTimePeriods, self).__init__(name)
+
+    def command(self):
+        import ckan.model as model
+        from ga_model import post_update_url_stats
+        self._load_config()
+        model.Session.remove()
+        model.Session.configure(bind=model.meta.engine)
+
+        log = logging.getLogger('ckanext.ga_report')
+
+        log.info("Updating 'All' records for old URLs")
+        post_update_url_stats()
+        log.info("Processing complete")
+
+
 
 class LoadAnalytics(CkanCommand):
     """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-DD  - just data for all time periods going
-                      back to (and including) this date
+        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)
+        self.parser.add_option('-d', '--delete-first',
+                               action='store_true',
+                               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()
 
-        from ga_auth import initialize_service
-        try:
-            svc = initialize_service(self.args[0], None)
-        except TypeError:
-            print 'Have you correctly run the getauthtoken task and specified the correct file here'
+        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
 
-        from download_analytics import DownloadAnalytics
-        from ga_auth import get_profile_id
-        downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc))
+        try:
+            svc = init_service(ga_token_filepath, None)
+        except TypeError:
+            print ('Have you correctly run the getauthtoken task and '
+                   'specified the correct token file in the CKAN config under '
+                   '"googleanalytics.token.filepath"?')
+            return
 
-        time_period = self.args[1] if self.args and len(self.args) > 1 else 'latest'
+        downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc),
+                                       delete_first=self.options.delete_first,
+                                       skip_url_stats=self.options.skip_url_stats)
+
+        time_period = self.args[0] if self.args else 'latest'
         if time_period == 'all':
             downloader.all_()
         elif time_period == 'latest':
             downloader.latest()
         else:
-            since_date = datetime.datetime.strptime(time_period, '%Y-%m-%d')
-            downloader.since_date(since_date)
+            # The month to use
+            for_date = datetime.datetime.strptime(time_period, '%Y-%m')
+            downloader.specific_month(for_date)
 
-

--- a/ckanext/ga_report/controller.py
+++ b/ckanext/ga_report/controller.py
@@ -1,10 +1,538 @@
+import re
+import csv
+import sys
+import json
 import logging
-from ckan.lib.base import BaseController, c, render
-import report_model
+import operator
+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, GA_ReferralStat, GA_Publisher
 
 log = logging.getLogger('ckanext.ga-report')
 
+DOWNLOADS_AVAILABLE_FROM = '2012-12'
+
+def _get_month_name(strdate):
+    import calendar
+    from time import strptime
+    d = strptime(strdate, '%Y-%m')
+    return '%s %s' % (calendar.month_name[d.tm_mon], d.tm_year)
+
+def _get_unix_epoch(strdate):
+    from time import strptime,mktime
+    d = strptime(strdate, '%Y-%m')
+    return int(mktime(d))
+
+def _month_details(cls, stat_key=None):
+    '''
+    Returns a list of all the periods for which we have data, unfortunately
+    knows too much about the type of the cls being passed as GA_Url has a
+    more complex query
+
+    This may need extending if we add a period_name to the stats
+    '''
+    months = []
+    day = None
+
+    q = model.Session.query(cls.period_name,cls.period_complete_day)\
+        .filter(cls.period_name!='All').distinct(cls.period_name)
+    if stat_key:
+        q=  q.filter(cls.stat_name==stat_key)
+
+    vals = q.order_by("period_name desc").all()
+
+    if vals and vals[0][1]:
+        day = int(vals[0][1])
+        ordinal = 'th' if 11 <= day <= 13 \
+            else {1:'st',2:'nd',3:'rd'}.get(day % 10, 'th')
+        day = "{day}{ordinal}".format(day=day, ordinal=ordinal)
+
+    for m in vals:
+        months.append( (m[0], _get_month_name(m[0])))
+
+    return months, day
+
+
 class GaReport(BaseController):
+
+    def csv(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.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"])
+
+        for entry in entries:
+            writer.writerow([entry.period_name.encode('utf-8'),
+                             entry.stat_name.encode('utf-8'),
+                             entry.key.encode('utf-8'),
+                             entry.value.encode('utf-8')])
+
+
     def index(self):
-        return render('index.html')
-
+
+        # 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])
+
+        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', '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','Bounce rate (home page)']:
+                    val = "%s%%" % val
+            if key in ['Total page views', 'Total visits']:
+                val = int(val)
+
+            return key, val
+
+        # Query historic values for sparkline rendering
+        sparkline_query = model.Session.query(GA_Stat)\
+                .filter(GA_Stat.stat_name=='Totals')\
+                .order_by(GA_Stat.period_name)
+        sparkline_data = {}
+        for x in sparkline_query:
+            sparkline_data[x.key] = sparkline_data.get(x.key,[])
+            key, val = clean_key(x.key,float(x.value))
+            tooltip = '%s: %s' % (_get_month_name(x.period_name), val)
+            sparkline_data[x.key].append( (tooltip,x.value) )
+        # Trim the latest month, as it looks like a huge dropoff
+        for key in sparkline_data:
+            sparkline_data[key] = sparkline_data[key][:-1]
+
+        c.global_totals = []
+        if c.month:
+            for e in entries:
+                key, val = clean_key(e.key, e.value)
+                sparkline = sparkline_data[e.key]
+                c.global_totals.append((key, val, sparkline))
+        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))/float(len(v))
+                sparkline = sparkline_data[k]
+                key, val = clean_key(k,v)
+
+                c.global_totals.append((key, val, sparkline))
+        # Sort the global totals into a more pleasant order
+        def sort_func(x):
+            key = x[0]
+            total_order = ['Total page views','Total visits','Pages per visit']
+            if key in total_order:
+                return total_order.index(key)
+            return 999
+        c.global_totals = sorted(c.global_totals, key=sort_func)
+
+        keys = {
+            'Browser versions': 'browser_versions',
+            '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():
+            q = model.Session.query(GA_Stat).\
+                filter(GA_Stat.stat_name==k).\
+                order_by(GA_Stat.period_name)
+            # Buffer the tabular data
+            if c.month:
+                entries = []
+                q = q.filter(GA_Stat.period_name==c.month).\
+                          order_by('ga_stat.value::int desc')
+            d = collections.defaultdict(int)
+            for e in q.all():
+                d[e.key] += int(e.value)
+            entries = []
+            for key, val in d.iteritems():
+                entries.append((key,val,))
+            entries = sorted(entries, key=operator.itemgetter(1), reverse=True)
+
+            # Run a query on all months to gather graph data
+            graph_query = model.Session.query(GA_Stat).\
+                filter(GA_Stat.stat_name==k).\
+                order_by(GA_Stat.period_name)
+            graph_dict = {}
+            for stat in graph_query:
+                graph_dict[ stat.key ] = graph_dict.get(stat.key,{
+                    'name':stat.key, 
+                    'data': []
+                    })
+                graph_dict[ stat.key ]['data'].append({
+                    'x':_get_unix_epoch(stat.period_name),
+                    'y':float(stat.value)
+                    })
+            graph = [ graph_dict[x[0]] for x in entries ]
+            setattr(c, v+'_graph', json.dumps( _to_rickshaw(graph,percentageMode=True) ))
+
+            # Get the total for each set of values and then set the value as
+            # a percentage of the total
+            if k == 'Social sources':
+                total = sum([x for n,x,graph in c.global_totals if n == 'Total visits'])
+            else:
+                total = sum([num for _,num in entries])
+            setattr(c, v, [(k,_percent(v,total)) for k,v in entries ])
+
+        return render('ga_report/site/index.html')
+
+
+class GaDatasetReport(BaseController):
+    """
+    Displays the pageview and visit count for datasets
+    with options to filter by publisher and time period.
+    """
+    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"])
+
+        top_publishers, top_publishers_graph = _get_top_publishers(None)
+
+        for publisher,view,visit in top_publishers:
+            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", "Resource downloads", "Period Name"])
+
+        for package,view,visit,downloads in packages:
+            writer.writerow([package.title.encode('utf-8'),
+                             package.name.encode('utf-8'),
+                             view,
+                             visit,
+                             downloads,
+                             month])
+
+    def publishers(self):
+        '''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, 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', '')
+        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, graph_data = _get_top_publishers()
+        c.top_publishers_graph = json.dumps( _to_rickshaw(graph_data) )
+
+        return render('ga_report/publisher/index.html')
+
+    def _get_packages(self, publisher=None, count=-1):
+        '''Returns the datasets in order of views'''
+        have_download_data = True
+        month = c.month or 'All'
+        if month != 'All':
+            have_download_data = month >= DOWNLOADS_AVAILABLE_FROM
+
+        q = model.Session.query(GA_Url,model.Package)\
+            .filter(model.Package.name==GA_Url.package_id)\
+            .filter(GA_Url.url.like('/dataset/%'))
+        if publisher:
+            q = q.filter(GA_Url.department_id==publisher.name)
+        q = q.filter(GA_Url.period_name==month)
+        q = q.order_by('ga_url.pageviews::int desc')
+        top_packages = []
+        if count == -1:
+            entries = q.all()
+        else:
+            entries = q.limit(count)
+
+        for entry,package in entries:
+            if package:
+                # Downloads ....
+                if have_download_data:
+                    dls = model.Session.query(GA_Stat).\
+                        filter(GA_Stat.stat_name=='Downloads').\
+                        filter(GA_Stat.key==package.name)
+                    if month != 'All':  # Fetch everything unless the month is specific
+                        dls = dls.filter(GA_Stat.period_name==month)
+                    downloads = 0
+                    for x in dls:
+                        downloads += int(x.value)
+                else:
+                    downloads = 'No data'
+                top_packages.append((package, entry.pageviews, entry.visits, downloads))
+            else:
+                log.warning('Could not find package associated package')
+
+        return top_packages
+
+    def read(self):
+        '''
+        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
+        # month names from the values.
+        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', '')
+        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
+
+        c.top_packages = self._get_packages(c.publisher, 20)
+
+        # Graph query
+        top_package_names = [ x[0].name for x in c.top_packages ]
+        graph_query = model.Session.query(GA_Url,model.Package)\
+            .filter(model.Package.name==GA_Url.package_id)\
+            .filter(GA_Url.url.like('/dataset/%'))\
+            .filter(GA_Url.package_id.in_(top_package_names))
+        graph_dict = {}
+        for entry,package in graph_query:
+            if not package: continue
+            if entry.period_name=='All': continue
+            graph_dict[package.name] = graph_dict.get(package.name,{
+                'name':package.title,
+                'data':[]
+                })
+            graph_dict[package.name]['data'].append({
+                'x':_get_unix_epoch(entry.period_name),
+                'y':int(entry.pageviews),
+                })
+        graph = [ graph_dict[x] for x in top_package_names ]
+
+        c.graph_data = json.dumps( _to_rickshaw(graph) )
+
+        return render('ga_report/publisher/read.html')
+
+def _to_rickshaw(data, percentageMode=False):
+    if data==[]:
+        return data
+    # Create a consistent x-axis between all series
+    num_points = [ len(series['data']) for series in data ]
+    ideal_index = num_points.index( max(num_points) )
+    x_axis = []
+    for series in data:
+        for point in series['data']:
+            x_axis.append(point['x'])
+    x_axis = sorted( list( set(x_axis) ) )
+    # Zero pad any missing values
+    for series in data:
+        xs = [ point['x'] for point in series['data'] ]
+        for x in set(x_axis).difference(set(xs)):
+            series['data'].append( {'x':x, 'y':0} )
+    if percentageMode:
+        def get_totals(series_list):
+            totals = {}
+            for series in series_list:
+                for point in series['data']:
+                    totals[point['x']] = totals.get(point['x'],0) + point['y']
+            return totals
+        # Transform data into percentage stacks
+        totals = get_totals(data)
+        # Roll insignificant series into a catch-all
+        THRESHOLD = 0.01
+        raw_data = data
+        data = []
+        for series in raw_data:
+            for point in series['data']:
+                fraction = float(point['y']) / totals[point['x']]
+                if not (series in data) and fraction>THRESHOLD:
+                    data.append(series)
+        # Overwrite data with a set of intereting series
+        others = [ x for x in raw_data if not (x in data) ]
+        data.append({ 
+            'name':'Other',
+            'data': [ {'x':x,'y':y} for x,y in get_totals(others).items() ] 
+            })
+        # Turn each point into a percentage
+        for series in data:
+            for point in series['data']:
+                point['y'] = (point['y']*100) / totals[point['x']]
+    # Sort the points
+    for series in data:
+        series['data'] = sorted( series['data'], key=lambda x:x['x'] )
+        # Strip the latest month's incomplete analytics
+        series['data'] = series['data'][:-1]
+    return data
+
+
+def _get_top_publishers(limit=20):
+    '''
+    Returns a list of the top 20 publishers by dataset visits.
+    (The number to show can be varied with 'limit')
+    '''
+    month = c.month or 'All'
+    connection = model.Session.connection()
+    q = """
+        select department_id, sum(pageviews::int) views, sum(visits::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 views desc
+        """
+    if limit:
+        q = q + " limit %s;" % (limit)
+
+    top_publishers = []
+    res = connection.execute(q, month)
+    department_ids = []
+    for row in res:
+        g = model.Group.get(row[0])
+        if g:
+            department_ids.append(row[0])
+            top_publishers.append((g, row[1], row[2]))
+
+    graph = []
+    if limit is not None:
+        # Query for a history graph of these publishers
+        q = model.Session.query(
+                GA_Url.department_id, 
+                GA_Url.period_name, 
+                func.sum(cast(GA_Url.pageviews,sqlalchemy.types.INT)))\
+            .filter( GA_Url.department_id.in_(department_ids) )\
+            .filter( GA_Url.period_name!='All' )\
+            .filter( GA_Url.url.like('/dataset/%') )\
+            .filter( GA_Url.package_id!='' )\
+            .group_by( GA_Url.department_id, GA_Url.period_name )
+        graph_dict = {}
+        for dept_id,period_name,views in q:
+            graph_dict[dept_id] = graph_dict.get( dept_id, {
+                'name' : model.Group.get(dept_id).title,
+                'data' : []
+                })
+            graph_dict[dept_id]['data'].append({
+                'x': _get_unix_epoch(period_name),
+                'y': views
+                })
+        # Sort dict into ordered list
+        for id in department_ids:
+            graph.append( graph_dict[id] )
+    return top_publishers, graph
+
+
+def _get_publishers():
+    '''
+    Returns a list of all publishers. Each item is a tuple:
+      (name, 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)
+