Refactor graph code. Querying to build a Top Packages graph.
--- a/ckanext/ga_report/controller.py
+++ b/ckanext/ga_report/controller.py
@@ -1,14 +1,20 @@
+import re
+import csv
+import sys
+import json
import logging
import operator
-from ckan.lib.base import BaseController, c, render, request, response
+import collections
+from ckan.lib.base import (BaseController, c, g, render, request, response, abort)
import sqlalchemy
from sqlalchemy import func, cast, Integer
import ckan.model as model
-from ga_model import GA_Url, GA_Stat
+from ga_model import GA_Url, GA_Stat, GA_ReferralStat, GA_Publisher
log = logging.getLogger('ckanext.ga-report')
+DOWNLOADS_AVAILABLE_FROM = '2012-12'
def _get_month_name(strdate):
import calendar
@@ -16,13 +22,39 @@
d = strptime(strdate, '%Y-%m')
return '%s %s' % (calendar.month_name[d.tm_mon], d.tm_year)
-
-def _month_details(cls):
+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 = []
- vals = model.Session.query(cls.period_name).distinct().all()
+ 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 sorted(months, key=operator.itemgetter(0), reverse=True)
+
+ return months, day
class GaReport(BaseController):
@@ -30,12 +62,13 @@
def csv(self, month):
import csv
- entries = model.Session.query(GA_Stat).\
- filter(GA_Stat.period_name==month).\
- order_by('GA_Stat.stat_name, GA_Stat.key').all()
-
- response.headers['Content-disposition'] = 'attachment; filename=dgu_analytics_%s.csv' % (month)
+ 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"])
@@ -46,98 +79,378 @@
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 = request.params.get('month', c.months[0][0] if c.months else '')
- c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
- entries = model.Session.query(GA_Stat).\
- filter(GA_Stat.stat_name=='Totals').\
- filter(GA_Stat.period_name==c.month).\
- order_by('ga_stat.key').all()
- c.global_totals = [(s.key, s.value) for s in entries ]
+ 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': 'browsers',
- 'Operating Systems versions': 'os',
+ 'Browser versions': 'browser_versions',
+ 'Browsers': 'browsers',
+ 'Operating Systems versions': 'os_versions',
+ 'Operating Systems': 'os',
'Social sources': 'social_networks',
'Languages': 'languages',
'Country': 'country'
}
+ def shorten_name(name, length=60):
+ return (name[:length] + '..') if len(name) > 60 else name
+
+ def fill_out_url(url):
+ import urlparse
+ return urlparse.urljoin(g.site_url, url)
+
+ c.social_referrer_totals, c.social_referrers = [], []
+ q = model.Session.query(GA_ReferralStat)
+ q = q.filter(GA_ReferralStat.period_name==c.month) if c.month else q
+ q = q.order_by('ga_referrer.count::int desc')
+ for entry in q.all():
+ c.social_referrers.append((shorten_name(entry.url), fill_out_url(entry.url),
+ entry.source,entry.count))
+
+ q = model.Session.query(GA_ReferralStat.url,
+ func.sum(GA_ReferralStat.count).label('count'))
+ q = q.filter(GA_ReferralStat.period_name==c.month) if c.month else q
+ q = q.order_by('count desc').group_by(GA_ReferralStat.url)
+ for entry in q.all():
+ c.social_referrer_totals.append((shorten_name(entry[0]), fill_out_url(entry[0]),'',
+ entry[1]))
+
for k, v in keys.iteritems():
- entries = model.Session.query(GA_Stat).\
+ q = model.Session.query(GA_Stat).\
filter(GA_Stat.stat_name==k).\
- filter(GA_Stat.period_name==c.month).\
- order_by('ga_stat.value::int desc').all()
- setattr(c, v, [(s.key, s.value) for s in entries ])
-
+ order_by(GA_Stat.period_name)
+ # Run the query on all months to gather graph data
+ series = {}
+ x_axis = set()
+ for stat in q:
+ x_val = _get_unix_epoch(stat.period_name)
+ series[ stat.key ] = series.get(stat.key,{})
+ series[ stat.key ][x_val] = float(stat.value)
+ x_axis.add(x_val)
+ # Common x-axis for all series. Exclude this month (incomplete data)
+ x_axis = sorted(list(x_axis))[:-1]
+ # Buffer a rickshaw dataset from the series
+ def create_graph(series_name, series_data):
+ return {
+ 'name':series_name,
+ 'data':[ {'x':x,'y':series_data.get(x,0)} for x in x_axis ]
+ }
+ rickshaw = [ create_graph(name,data) for name, data in series.items() ]
+ rickshaw = sorted(rickshaw,key=lambda x:x['data'][-1]['y'])
+ setattr(c, v+'_graph', json.dumps(rickshaw))
+
+ # 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)
+
+ # 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 GaPublisherReport(BaseController):
+class GaDatasetReport(BaseController):
"""
- Displays the pageview and visit count for specific publishers based on
- the datasets associated with the publisher.
+ Displays the pageview and visit count for datasets
+ with options to filter by publisher and time period.
"""
-
- def index(self):
+ def publisher_csv(self, month):
+ '''
+ Returns a CSV of each publisher with the total number of dataset
+ views & visits.
+ '''
+ c.month = month if not month == 'all' else ''
+ response.headers['Content-Type'] = "text/csv; charset=utf-8"
+ response.headers['Content-Disposition'] = str('attachment; filename=publishers_%s.csv' % (month,))
+
+ writer = csv.writer(response)
+ writer.writerow(["Publisher Title", "Publisher Name", "Views", "Visits", "Period Name"])
+
+ for publisher,view,visit in _get_top_publishers(None):
+ writer.writerow([publisher.title.encode('utf-8'),
+ publisher.name.encode('utf-8'),
+ view,
+ visit,
+ month])
+
+ def dataset_csv(self, id='all', month='all'):
+ '''
+ Returns a CSV with the number of views & visits for each dataset.
+
+ :param id: A Publisher ID or None if you want for all
+ :param month: The time period, or 'all'
+ '''
+ c.month = month if not month == 'all' else ''
+ if id != 'all':
+ c.publisher = model.Group.get(id)
+ if not c.publisher:
+ abort(404, 'A publisher with that name could not be found')
+
+ packages = self._get_packages(c.publisher)
+ response.headers['Content-Type'] = "text/csv; charset=utf-8"
+ response.headers['Content-Disposition'] = \
+ str('attachment; filename=datasets_%s_%s.csv' % (c.publisher_name, month,))
+
+ writer = csv.writer(response)
+ writer.writerow(["Dataset Title", "Dataset Name", "Views", "Visits", "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 = _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', c.months[0][0] if c.months else '')
- c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
- connection = model.Session.connection()
- q = """
- select department_id, sum(pageviews::int) views, sum(visitors::int) visits
- from ga_url
- where department_id <> ''
- and not url like '/publisher/%%'
- and period_name=%s
- group by department_id order by views desc limit 20;
- """
- c.top_publishers = []
- res = connection.execute(q, c.month)
- for row in res:
- c.top_publishers.append((model.Group.get(row[0]), row[1], row[2]))
-
+ c.month = request.params.get('month', '')
+ c.month_desc = 'all months'
+ if c.month:
+ c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
+
+ c.top_publishers = _get_top_publishers()
return render('ga_report/publisher/index.html')
-
- def read(self, id):
- c.publisher = model.Group.get(id)
+ 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 = _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', c.months[0][0] if c.months else '')
- c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
-
- entry = model.Session.query(GA_Url).\
- filter(GA_Url.url=='/publisher/%s' % c.publisher.name).\
- filter(GA_Url.period_name==c.month).first()
+ c.month = request.params.get('month', '')
+ if not c.month:
+ c.month_desc = 'all months'
+ else:
+ c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month])
+
+ month = c.month or 'All'
+ c.publisher_page_views = 0
+ q = model.Session.query(GA_Url).\
+ filter(GA_Url.url=='/publisher/%s' % c.publisher_name)
+ entry = q.filter(GA_Url.period_name==c.month).first()
c.publisher_page_views = entry.pageviews if entry else 0
- entries = model.Session.query(GA_Url).\
- filter(GA_Url.department_id==c.publisher.name).\
- filter(GA_Url.period_name==c.month).\
- order_by('ga_url.pageviews::int desc')[:20]
- for entry in entries:
- if entry.url.startswith('/dataset/'):
- p = model.Package.get(entry.url[len('/dataset/'):])
- c.top_packages.append((p,entry.pageviews,entry.visitors))
+ c.top_packages = self._get_packages(c.publisher, 20)
+
+ # 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_data = {}
+ for entry,package in graph_query:
+ if not package: continue
+ if entry.period_name=='All': continue
+ graph_data[package.id] = graph_data.get(package.id,{
+ 'name':package.title,
+ 'data':[]
+ })
+ graph_data[package.id]['data'].append({
+ 'x':_get_unix_epoch(entry.period_name),
+ 'y':int(entry.pageviews),
+ })
+ c.graph_data = json.dumps(graph_data.values())
return render('ga_report/publisher/read.html')
+def _get_top_publishers(limit=20):
+ '''
+ Returns a list of the top 20 publishers by dataset visits.
+ (The number to show can be varied with 'limit')
+ '''
+ month = c.month or 'All'
+ connection = model.Session.connection()
+ q = """
+ select department_id, sum(pageviews::int) views, sum(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)
+ for row in res:
+ g = model.Group.get(row[0])
+ if g:
+ top_publishers.append((g, row[1], row[2]))
+ return top_publishers
+
+
+def _get_publishers():
+ '''
+ Returns a list of all publishers. Each item is a tuple:
+ (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)
+
--- a/ckanext/ga_report/download_analytics.py
+++ b/ckanext/ga_report/download_analytics.py
@@ -1,9 +1,9 @@
import os
import logging
import datetime
-
+import collections
from pylons import config
-
+from ga_model import _normalize_url
import ga_model
#from ga_client import GA
@@ -11,15 +11,20 @@
log = logging.getLogger('ckanext.ga-report')
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):
+ 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
@@ -90,32 +95,81 @@
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))
- 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))
- self.store(period_name, period_complete_day, data,)
-
- ga_model.update_publisher_stats(period_name) # about 30 seconds.
- self.sitewide_stats( period_name )
-
-
- def download(self, start_date, end_date, path='~/dataset/[a-z0-9-_]+'):
+ 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)
+
+ 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,)
+
+ # Make sure the All records are correct.
+ ga_model.post_update_url_stats()
+
+ 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, period_complete_day )
+
+ log.info('Downloading and storing analytics for social networks')
+ self.update_social_info(period_na