--- a/ckanext/ga_report/controller.py +++ b/ckanext/ga_report/controller.py @@ -1,6 +1,7 @@ import re import csv import sys +import json import logging import operator import collections @@ -21,6 +22,10 @@ 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): ''' @@ -108,24 +113,24 @@ return key, val # Query historic values for sparkline rendering - graph_query = model.Session.query(GA_Stat)\ + sparkline_query = model.Session.query(GA_Stat)\ .filter(GA_Stat.stat_name=='Totals')\ .order_by(GA_Stat.period_name) - graph_data = {} - for x in graph_query: - graph_data[x.key] = graph_data.get(x.key,[]) + 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) - graph_data[x.key].append( (tooltip,x.value) ) + sparkline_data[x.key].append( (tooltip,x.value) ) # Trim the latest month, as it looks like a huge dropoff - for key in graph_data: - graph_data[key] = graph_data[key][:-1] + 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 = graph_data[e.key] + sparkline = sparkline_data[e.key] c.global_totals.append((key, val, sparkline)) else: d = collections.defaultdict(list) @@ -136,11 +141,18 @@ v = sum(v) else: v = float(sum(v))/float(len(v)) - sparkline = graph_data[k] + sparkline = sparkline_data[k] key, val = clean_key(k,v) c.global_totals.append((key, val, sparkline)) - c.global_totals = sorted(c.global_totals, key=operator.itemgetter(0)) + # 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', @@ -177,12 +189,13 @@ for k, v in keys.iteritems(): q = model.Session.query(GA_Stat).\ - filter(GA_Stat.stat_name==k) + filter(GA_Stat.stat_name==k).\ + order_by(GA_Stat.period_name) + # Buffer the tabular data if c.month: entries = [] q = q.filter(GA_Stat.period_name==c.month).\ order_by('ga_stat.value::int desc') - d = collections.defaultdict(int) for e in q.all(): d[e.key] += int(e.value) @@ -191,6 +204,23 @@ entries.append((key,val,)) entries = sorted(entries, key=operator.itemgetter(1), reverse=True) + # Run a query on all months to gather graph data + graph_query = model.Session.query(GA_Stat).\ + filter(GA_Stat.stat_name==k).\ + order_by(GA_Stat.period_name) + graph_dict = {} + for stat in graph_query: + graph_dict[ stat.key ] = graph_dict.get(stat.key,{ + 'name':stat.key, + 'raw': {} + }) + graph_dict[ stat.key ]['raw'][stat.period_name] = float(stat.value) + stats_in_table = [x[0] for x in entries] + stats_not_in_table = set(graph_dict.keys()) - set(stats_in_table) + stats = stats_in_table + sorted(list(stats_not_in_table)) + graph = [graph_dict[x] for x in stats] + setattr(c, v+'_graph', json.dumps( _to_rickshaw(graph,percentageMode=True) )) + # Get the total for each set of values and then set the value as # a percentage of the total if k == 'Social sources': @@ -219,7 +249,9 @@ writer = csv.writer(response) writer.writerow(["Publisher Title", "Publisher Name", "Views", "Visits", "Period Name"]) - for publisher,view,visit in _get_top_publishers(None): + top_publishers = _get_top_publishers(limit=None) + + for publisher,view,visit in top_publishers: writer.writerow([publisher.title.encode('utf-8'), publisher.name.encode('utf-8'), view, @@ -239,7 +271,7 @@ if not c.publisher: abort(404, 'A publisher with that name could not be found') - packages = self._get_packages(c.publisher) + packages = self._get_packages(publisher=c.publisher, month=c.month) response.headers['Content-Type'] = "text/csv; charset=utf-8" response.headers['Content-Disposition'] = \ str('attachment; filename=datasets_%s_%s.csv' % (c.publisher_name, month,)) @@ -269,12 +301,17 @@ c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month]) c.top_publishers = _get_top_publishers() - return render('ga_report/publisher/index.html') - - def _get_packages(self, publisher=None, count=-1): + graph_data = _get_top_publishers_graph() + c.top_publishers_graph = json.dumps( _to_rickshaw(graph_data) ) + + x = render('ga_report/publisher/index.html') + + return x + + def _get_packages(self, publisher=None, month='', count=-1): '''Returns the datasets in order of views''' have_download_data = True - month = c.month or 'All' + month = month or 'All' if month != 'All': have_download_data = month >= DOWNLOADS_AVAILABLE_FROM @@ -300,11 +337,13 @@ 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 = sum(int(d.value) for d in dls.all()) + downloads = 0 + for x in dls: + downloads += int(x.value) else: downloads = 'No data' - top_packages.append((package, entry.pageviews, entry.visits, downloads)) + if package.private == False: + top_packages.append((package, entry.pageviews, entry.visits, downloads)) else: log.warning('Could not find package associated package') @@ -350,9 +389,73 @@ 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) + c.top_packages = self._get_packages(publisher=c.publisher, count=20, month=c.month) + + # Graph query + top_packages_all_time = self._get_packages(publisher=c.publisher, count=20, month='All') + top_package_names = [ x[0].name for x in top_packages_all_time ] + graph_query = model.Session.query(GA_Url,model.Package)\ + .filter(model.Package.name==GA_Url.package_id)\ + .filter(GA_Url.url.like('/dataset/%'))\ + .filter(GA_Url.package_id.in_(top_package_names)) + all_series = {} + for entry,package in graph_query: + if not package: continue + if entry.period_name=='All': continue + all_series[package.name] = all_series.get(package.name,{ + 'name':package.title, + 'raw': {} + }) + all_series[package.name]['raw'][entry.period_name] = int(entry.pageviews) + graph = [ all_series[series_name] for series_name in top_package_names ] + c.graph_data = json.dumps( _to_rickshaw(graph) ) return render('ga_report/publisher/read.html') + +def _to_rickshaw(data, percentageMode=False): + if data==[]: + return data + # x-axis is every month in c.months. Note that data might not exist + # for entire history, eg. for recently-added datasets + x_axis = [x[0] for x in c.months] + x_axis.reverse() # Ascending order + x_axis = x_axis[:-1] # Remove latest month + totals = {} + for series in data: + series['data'] = [] + for x_string in x_axis: + x = _get_unix_epoch( x_string ) + y = series['raw'].get(x_string,0) + series['data'].append({'x':x,'y':y}) + totals[x] = totals.get(x,0)+y + if not percentageMode: + return data + # Turn all data into percentages + # Roll insignificant series into a catch-all + THRESHOLD = 1 + raw_data = data + data = [] + for series in raw_data: + for point in series['data']: + percentage = (100*float(point['y'])) / totals[point['x']] + if not (series in data) and percentage>THRESHOLD: + data.append(series) + point['y'] = percentage + others = [ x for x in raw_data if not (x in data) ] + if len(others): + data_other = [] + for i in range(len(x_axis)): + x = _get_unix_epoch(x_axis[i]) + y = 0 + for series in others: + y += series['data'][i]['y'] + data_other.append({'x':x,'y':y}) + data.append({ + 'name':'Other', + 'data': data_other + }) + return data + def _get_top_publishers(limit=20): ''' @@ -382,6 +485,46 @@ return top_publishers +def _get_top_publishers_graph(limit=20): + ''' + Returns a list of the top 20 publishers by dataset visits. + (The number to show can be varied with 'limit') + ''' + connection = model.Session.connection() + q = """ + select department_id, sum(pageviews::int) views + from ga_url + where department_id <> '' + and package_id <> '' + and url like '/dataset/%%' + and period_name='All' + group by department_id order by views desc + """ + if limit: + q = q + " limit %s;" % (limit) + + res = connection.execute(q) + department_ids = [ row[0] for row in res ] + + # Query for a history graph of these department ids + q = model.Session.query( + GA_Url.department_id, + GA_Url.period_name, + func.sum(cast(GA_Url.pageviews,sqlalchemy.types.INT)))\ + .filter( GA_Url.department_id.in_(department_ids) )\ + .filter( GA_Url.url.like('/dataset/%') )\ + .filter( GA_Url.package_id!='' )\ + .group_by( GA_Url.department_id, GA_Url.period_name ) + graph_dict = {} + for dept_id,period_name,views in q: + graph_dict[dept_id] = graph_dict.get( dept_id, { + 'name' : model.Group.get(dept_id).title, + 'raw' : {} + }) + graph_dict[dept_id]['raw'][period_name] = views + return [ graph_dict[id] for id in department_ids ] + + def _get_publishers(): ''' Returns a list of all publishers. Each item is a tuple: @@ -389,7 +532,7 @@ ''' publishers = [] for pub in model.Session.query(model.Group).\ - filter(model.Group.type=='publisher').\ + filter(model.Group.type=='organization').\ filter(model.Group.state=='active').\ order_by(model.Group.name): publishers.append((pub.name, pub.title))