From: Ross Jones Date: Wed, 07 Nov 2012 14:11:35 +0000 Subject: Handling duplicate urls in ga_url processing X-Git-Url: https://maxious.lambdacomplex.org/git/?p=ckanext-ga-report.git&a=commitdiff&h=6a953b2c1630aabdd8d7ea629b3f06c8c2f6eb47 --- Handling duplicate urls in ga_url processing --- --- a/README.rst +++ b/README.rst @@ -31,10 +31,12 @@ 2. Ensure you development.ini (or similar) contains the info about your Google Analytics account and configuration:: googleanalytics.id = UA-1010101-1 - googleanalytics.account = Account name (i.e. data.gov.uk, see top level item at https://www.google.com/analytics) + googleanalytics.account = Account name (e.g. data.gov.uk, see top level item at https://www.google.com/analytics) ga-report.period = monthly + ga-report.bounce_url = /data - Note that your credentials will be readable by system administrators on your server. Rather than use sensitive account details, it is suggested you give access to the GA account to a new Google account that you create just for this purpose. + 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 /. 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):: @@ -43,6 +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 @@ -79,7 +87,7 @@ Tutorial -------- -Download some GA data and store it in CKAN's db. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file) and specifying the name of your auth file (token.dat by default) from the previous step:: +Download some GA data and store it in CKAN's database. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file) and specifying the name of your auth file (token.dat by default) from the previous step:: $ paster loadanalytics token.dat latest --config=../ckan/development.ini --- a/ckanext/ga_report/command.py +++ b/ckanext/ga_report/command.py @@ -73,6 +73,14 @@ max_args = 2 min_args = 1 + def __init__(self, name): + super(LoadAnalytics, self).__init__(name) + self.parser.add_option('-d', '--delete-first', + action='store_true', + default=False, + dest='delete_first', + help='Delete data for the period first') + def command(self): self._load_config() @@ -83,10 +91,11 @@ svc = init_service(self.args[0], None) except TypeError: print ('Have you correctly run the getauthtoken task and ' - 'specified the correct file here') + 'specified the correct token file?') return - downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc)) + downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc), + delete_first=self.options.delete_first) time_period = self.args[1] if self.args and len(self.args) > 1 \ else 'latest' --- a/ckanext/ga_report/controller.py +++ b/ckanext/ga_report/controller.py @@ -1,11 +1,15 @@ +import re +import csv +import sys import logging import operator -from ckan.lib.base import BaseController, c, render, request, response, abort +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') @@ -18,8 +22,9 @@ def _month_details(cls): + '''Returns a list of all the month names''' months = [] - vals = model.Session.query(cls.period_name).distinct().all() + vals = model.Session.query(cls.period_name).filter(cls.period_name!='All').distinct().all() for m in vals: months.append( (m[0], _get_month_name(m[0]))) return sorted(months, key=operator.itemgetter(0), reverse=True) @@ -30,11 +35,13 @@ 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() + q = model.Session.query(GA_Stat) + 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"]) @@ -52,87 +59,221 @@ c.months = _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 = [] - for e in entries: - val = e.value - if e.key in ['Average time on site', 'Pages per visit', 'Percent new visits']: - val = "%.2f" % round(float(e.value), 2) - if e.key == 'Average time on site': + 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', 'Bounces']: + 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) - e.key = '%s *' % e.key - c.global_totals.append((e.key, val)) - + if key in ['New visits','Bounces']: + val = "%s%%" % val + if key in ['Total page views', 'Total visits']: + val = int(val) + + return key, val + + c.global_totals = [] + if c.month: + for e in entries: + key, val = clean_key(e.key, e.value) + c.global_totals.append((key, val)) + 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))/len(v) + key, val = clean_key(k,v) + + c.global_totals.append((key, val)) + c.global_totals = sorted(c.global_totals, key=operator.itemgetter(0)) 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).\ - 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 ]) - + q = model.Session.query(GA_Stat).\ + filter(GA_Stat.stat_name==k) + 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 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", "Period Name"]) + + for package,view,visit in packages: + writer.writerow([package.title.encode('utf-8'), + package.name.encode('utf-8'), + view, + visit, + 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) # 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 period_name=%s - group by department_id order by views desc limit 20; - """ - # Add this back (before and period_name =%s) if you want to ignore publisher - # homepage views - # and not url like '/publisher/%%' - - 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) - if not c.publisher: - abort(404, 'A publisher with that name could not be found') + def _get_packages(self, publisher=None, count=-1): + '''Returns the datasets in order of visits''' + if count == -1: + count = sys.maxint + + month = c.month or 'All' + + q = model.Session.query(GA_Url,model.Package)\ + .filter(model.Package.name==GA_Url.package_id)\ + .filter(GA_Url.url.like('/dataset/%')) + if publisher: + q = q.filter(GA_Url.department_id==publisher.name) + q = q.filter(GA_Url.period_name==month) + q = q.order_by('ga_url.visitors::int desc') + top_packages = [] + for entry,package in q.limit(count): + if package: + top_packages.append((package, entry.pageviews, entry.visitors)) + else: + log.warning('Could not find package associated package') + + return top_packages + + def read(self): + ''' + 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 @@ -140,22 +281,65 @@ c.months = _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) 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(visitors::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 visits 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: + (names, title) + ''' + publishers = [] + for pub in model.Session.query(model.Group).\ + filter(model.Group.type=='publisher').\ + filter(model.Group.state=='active').\ + order_by(model.Group.name): + publishers.append((pub.name, pub.title)) + return publishers + +def _percent(num, total): + p = 100 * float(num)/float(total) + return "%.2f%%" % round(p, 2) + --- a/ckanext/ga_report/download_analytics.py +++ b/ckanext/ga_report/download_analytics.py @@ -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 +11,17 @@ log = logging.getLogger('ckanext.ga-report') FORMAT_MONTH = '%Y-%m' +MIN_VIEWS = 50 +MIN_VISITS = 20 class DownloadAnalytics(object): '''Downloads and stores analytics info''' - def __init__(self, service=None, profile_id=None): + def __init__(self, service=None, profile_id=None, delete_first=False): self.period = config['ga-report.period'] self.service = service self.profile_id = profile_id - + self.delete_first = delete_first def specific_month(self, date): import calendar @@ -90,17 +92,26 @@ def download_and_store(self, periods): for period_name, period_complete_day, start_date, end_date in periods: + if self.delete_first: + log.info('Deleting existing Analytics for period "%s"', + period_name) + ga_model.delete(period_name) log.info('Downloading Analytics for 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-_]+') + # Clean up the entries before we run this + ga_model.pre_update_url_stats(period_name) + + accountName = config.get('googleanalytics.account') + + data = self.download(start_date, end_date, '~/%s/dataset/[a-z0-9-_]+' % accountName) 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-_]+') + data = self.download(start_date, end_date, '~/%s/publisher/[a-z0-9-_]+' % accountName) 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,) @@ -108,13 +119,40 @@ 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-_]+'): + self.update_social_info(period_name, start_date, end_date) + + + def update_social_info(self, period_name, start_date, end_date): + start_date = start_date.strftime('%Y-%m-%d') + end_date = end_date.strftime('%Y-%m-%d') + query = 'ga:hasSocialSourceReferral=~Yes$' + metrics = 'ga:entrances' + sort = '-ga:entrances' + + # Supported query params at + # https://developers.google.com/analytics/devguides/reporting/core/v3/reference + results = self.service.data().ga().get( + ids='ga:' + self.profile_id, + filters=query, + start_date=start_date, + metrics=metrics, + sort=sort, + dimensions="ga:landingPagePath,ga:socialNetwork", + max_results=10000, + end_date=end_date).execute() + data = collections.defaultdict(list) + rows = results.get('rows',[]) + for row in rows: + data[_normalize_url(row[0])].append( (row[1], int(row[2]),) ) + ga_model.update_social(period_name, data) + + + def download(self, start_date, end_date, path=None): '''Get data from GA for a given time period''' start_date = start_date.strftime('%Y-%m-%d') end_date = end_date.strftime('%Y-%m-%d') query = 'ga:pagePath=%s$' % path - metrics = 'ga:uniquePageviews, ga:visitors' + metrics = 'ga:uniquePageviews, ga:visits' sort = '-ga:uniquePageviews' # Supported query params at @@ -129,15 +167,13 @@ max_results=10000, end_date=end_date).execute() - if os.getenv('DEBUG'): - import pprint - pprint.pprint(results) - print 'Total results: %s' % results.get('totalResults') - packages = [] for entry in results.get('rows'): (loc,pageviews,visits) = entry - packages.append( ('http:/' + loc, pageviews, visits,) ) # Temporary hack + url = _normalize_url('http:/' + loc) + if not url.startswith('/dataset/') and not url.startswith('/publisher/'): + continue + packages.append( (url, pageviews, visits,) ) # Temporary hack return dict(url=packages) def store(self, period_name, period_complete_day, data): @@ -176,23 +212,43 @@ max_results=10000, end_date=end_date).execute() result_data = results.get('rows') - ga_model.update_sitewide_stats(period_name, "Totals", {'Total pageviews': result_data[0][0]}) - - results = self.service.data().ga().get( - ids='ga:' + self.profile_id, - start_date=start_date, - metrics='ga:pageviewsPerVisit,ga:bounces,ga:avgTimeOnSite,ga:percentNewVisits', + ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]}) + + results = self.service.data().ga().get( + ids='ga:' + self.profile_id, + start_date=start_date, + metrics='ga:pageviewsPerVisit,ga:avgTimeOnSite,ga:percentNewVisits,ga:visits', max_results=10000, end_date=end_date).execute() result_data = results.get('rows') data = { 'Pages per visit': result_data[0][0], - 'Bounces': result_data[0][1], - 'Average time on site': result_data[0][2], - 'Percent new visits': result_data[0][3], + 'Average time on site': result_data[0][1], + '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:uniquePageviews', + 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): """ Fetches stats about language and country """ @@ -208,11 +264,13 @@ data = {} for result in result_data: data[result[0]] = data.get(result[0], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Languages", data) data = {} for result in result_data: data[result[1]] = data.get(result[1], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Country", data) @@ -227,13 +285,11 @@ max_results=10000, end_date=end_date).execute() result_data = results.get('rows') - twitter_links = [] data = {} for result in result_data: if not result[0] == '(not set)': data[result[0]] = data.get(result[0], 0) + int(result[2]) - if result[0] == 'Twitter': - twitter_links.append(result[1]) + self._filter_out_long_tail(data, 3) ga_model.update_sitewide_stats(period_name, "Social sources", data) @@ -251,12 +307,14 @@ data = {} for result in result_data: data[result[0]] = data.get(result[0], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Operating Systems", data) data = {} for result in result_data: - key = "%s (%s)" % (result[0],result[1]) - data[key] = result[2] + if int(result[2]) >= MIN_VIEWS: + key = "%s %s" % (result[0],result[1]) + data[key] = result[2] ga_model.update_sitewide_stats(period_name, "Operating Systems versions", data) @@ -271,17 +329,42 @@ max_results=10000, end_date=end_date).execute() result_data = results.get('rows') + # e.g. [u'Firefox', u'19.0', u'20'] + data = {} for result in result_data: data[result[0]] = data.get(result[0], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Browsers", data) data = {} for result in result_data: - key = "%s (%s)" % (result[0], result[1]) - data[key] = result[2] + key = "%s %s" % (result[0], self._filter_browser_version(result[0], result[1])) + data[key] = data.get(key, 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Browser versions", data) + @classmethod + def _filter_browser_version(cls, browser, version_str): + ''' + Simplifies a browser version string if it is detailed. + i.e. groups together Firefox 3.5.1 and 3.5.2 to be just 3. + This is helpful when viewing stats and good to protect privacy. + ''' + ver = version_str + parts = ver.split('.') + if len(parts) > 1: + if parts[1][0] == '0': + ver = parts[0] + else: + ver = "%s" % (parts[0]) + # Special case complex version nums + if browser in ['Safari', 'Android Browser']: + ver = parts[0] + if len(ver) > 2: + num_hidden_digits = len(ver) - 2 + ver = ver[0] + ver[1] + 'X' * num_hidden_digits + return ver def _mobile_stats(self, start_date, end_date, period_name): """ Info about mobile devices """ @@ -299,10 +382,23 @@ data = {} for result in result_data: data[result[0]] = data.get(result[0], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Mobile brands", data) data = {} for result in result_data: data[result[1]] = data.get(result[1], 0) + int(result[2]) + self._filter_out_long_tail(data, MIN_VIEWS) ga_model.update_sitewide_stats(period_name, "Mobile devices", data) + @classmethod + def _filter_out_long_tail(cls, data, threshold=10): + ''' + Given data which is a frequency distribution, filter out + results which are below a threshold count. This is good to protect + privacy. + ''' + for key, value in data.items(): + if value < threshold: + del data[key] + --- a/ckanext/ga_report/ga_auth.py +++ b/ckanext/ga_report/ga_auth.py @@ -53,7 +53,11 @@ return None accountName = config.get('googleanalytics.account') + if not accountName: + raise Exception('googleanalytics.account needs to be configured') webPropertyId = config.get('googleanalytics.id') + if not webPropertyId: + raise Exception('googleanalytics.id needs to be configured') for acc in accounts.get('items'): if acc.get('name') == accountName: accountId = acc.get('id') --- a/ckanext/ga_report/ga_model.py +++ b/ckanext/ga_report/ga_model.py @@ -1,10 +1,10 @@ import re import uuid -from sqlalchemy import Table, Column, MetaData +from sqlalchemy import Table, Column, MetaData, ForeignKey from sqlalchemy import types from sqlalchemy.sql import select -from sqlalchemy.orm import mapper +from sqlalchemy.orm import mapper, relation from sqlalchemy import func import ckan.model as model @@ -13,7 +13,7 @@ def make_uuid(): return unicode(uuid.uuid4()) - +metadata = MetaData() class GA_Url(object): @@ -21,20 +21,6 @@ for k,v in kwargs.items(): setattr(self, k, v) -class GA_Stat(object): - - def __init__(self, **kwargs): - for k,v in kwargs.items(): - setattr(self, k, v) - -class GA_Publisher(object): - - def __init__(self, **kwargs): - for k,v in kwargs.items(): - setattr(self, k, v) - - -metadata = MetaData() url_table = Table('ga_url', metadata, Column('id', types.UnicodeText, primary_key=True, default=make_uuid), @@ -44,8 +30,16 @@ Column('visitors', types.UnicodeText), Column('url', types.UnicodeText), Column('department_id', types.UnicodeText), + Column('package_id', types.UnicodeText), ) mapper(GA_Url, url_table) + + +class GA_Stat(object): + + def __init__(self, **kwargs): + for k,v in kwargs.items(): + setattr(self, k, v) stat_table = Table('ga_stat', metadata, Column('id', types.UnicodeText, primary_key=True, @@ -56,6 +50,12 @@ Column('value', types.UnicodeText), ) mapper(GA_Stat, stat_table) + +class GA_Publisher(object): + + def __init__(self, **kwargs): + for k,v in kwargs.items(): + setattr(self, k, v) pub_table = Table('ga_publisher', metadata, Column('id', types.UnicodeText, primary_key=True, @@ -71,6 +71,24 @@ mapper(GA_Publisher, pub_table) +class GA_ReferralStat(object): + + def __init__(self, **kwargs): + for k,v in kwargs.items(): + setattr(self, k, v) + +referrer_table = Table('ga_referrer', metadata, + Column('id', types.UnicodeText, primary_key=True, + default=make_uuid), + Column('period_name', types.UnicodeText), + Column('source', types.UnicodeText), + Column('url', types.UnicodeText), + Column('count', types.Integer), + ) +mapper(GA_ReferralStat, referrer_table) + + + def init_tables(): metadata.create_all(model.meta.engine) @@ -93,11 +111,12 @@ >>> normalize_url('http://data.gov.uk/dataset/weekly_fuel_prices') '/dataset/weekly_fuel_prices' ''' - url = re.sub('https?://(www\.)?data.gov.uk', '', url) - return url - - -def _get_department_id_of_url(url): + # Deliberately leaving a / + url = url.replace('http:/','') + return '/' + '/'.join(url.split('/')[2:]) + + +def _get_package_and_publisher(url): # e.g. /dataset/fuel_prices # e.g. /dataset/fuel_prices/resource/e63380d4 dataset_match = re.match('/dataset/([^/]+)(/.*)?', url) @@ -107,12 +126,13 @@ if dataset: publisher_groups = dataset.get_groups('publisher') if publisher_groups: - return publisher_groups[0].name + return dataset_ref,publisher_groups[0].name + return dataset_ref, None else: publisher_match = re.match('/publisher/([^/]+)(/.*)?', url) if publisher_match: - return publisher_match.groups()[0] - + return None, publisher_match.groups()[0] + return None, None def update_sitewide_stats(period_name, stat_name, data): for k,v in data.iteritems(): @@ -137,36 +157,105 @@ model.Session.commit() +def update_url_stat_totals(period_name): + + """ + items = model.Session.query(GA_Url).\ + filter(GA_Url.period_name != "All").\ + filter(GA_Url.url==url).all() + values = {'id': make_uuid(), + 'period_name': "All", + 'period_complete_day': "0", + 'url': url, + 'pageviews': sum([int(x.pageviews) for x in items]), + 'visitors': sum([int(x.visitors) for x in items]), + 'department_id': department_id, + 'package_id': package + } + model.Session.add(GA_Url(**values)) + model.Session.commit() + """ + +def pre_update_url_stats(period_name): + model.Session.query(GA_Url).\ + filter(GA_Url.period_name==period_name).delete() + model.Session.query(GA_Url).\ + filter(GA_Url.period_name=='All').delete() + def update_url_stats(period_name, period_complete_day, url_data): + for url, views, visitors in url_data: - url = _normalize_url(url) - department_id = _get_department_id_of_url(url) - - # see if the row for this url & month is in the table already + package, publisher = _get_package_and_publisher(url) + item = model.Session.query(GA_Url).\ filter(GA_Url.period_name==period_name).\ filter(GA_Url.url==url).first() if item: - item.period_name = period_name - item.pageviews = views - item.visitors = visitors - item.department_id = department_id + item.pageviews = item.pageviews + views + item.visitors = item.visitors + visitors model.Session.add(item) else: - # create the row values = {'id': make_uuid(), 'period_name': period_name, 'period_complete_day': period_complete_day, 'url': url, 'pageviews': views, 'visitors': visitors, - 'department_id': department_id + 'department_id': publisher, + 'package_id': package } model.Session.add(GA_Url(**values)) model.Session.commit() - + if package: + entries = model.Session.query(GA_Url).\ + filter(GA_Url.period_name!='All').\ + filter(GA_Url.url==url).all() + values = {'id': make_uuid(), + 'period_name': 'All', + 'period_complete_day': 0, + 'url': url, + 'pageviews': sum([int(e.pageviews) for e in entries]), + 'visitors': sum([int(e.visitors) for e in entries]), + 'department_id': publisher, + 'package_id': package + } + + model.Session.add(GA_Url(**values)) + model.Session.commit() + + + + + +def update_social(period_name, data): + # Clean up first. + model.Session.query(GA_ReferralStat).\ + filter(GA_ReferralStat.period_name==period_name).delete() + + for url,data in data.iteritems(): + for entry in data: + source = entry[0] + count = entry[1] + + item = model.Session.query(GA_ReferralStat).\ + filter(GA_ReferralStat.period_name==period_name).\ + filter(GA_ReferralStat.source==source).\ + filter(GA_ReferralStat.url==url).first() + if item: + item.count = item.count + count + model.Session.add(item) + else: + # create the row + values = {'id': make_uuid(), + 'period_name': period_name, + 'source': source, + 'url': url, + 'count': count, + } + model.Session.add(GA_ReferralStat(**values)) + model.Session.commit() def update_publisher_stats(period_name): """ @@ -248,3 +337,15 @@ for grandchild in go_down_tree(child): yield grandchild +def delete(period_name): + ''' + Deletes table data for the specified period, or specify 'all' + for all periods. + ''' + for object_type in (GA_Url, GA_Stat, GA_Publisher, GA_ReferralStat): + q = model.Session.query(object_type) + if period_name != 'all': + q = q.filter_by(period_name=period_name) + q.delete() + model.Session.commit() + --- a/ckanext/ga_report/helpers.py +++ b/ckanext/ga_report/helpers.py @@ -1,37 +1,91 @@ import logging import operator + import ckan.lib.base as base import ckan.model as model +from ckan.logic import get_action +from ckanext.ga_report.ga_model import GA_Url, GA_Publisher +from ckanext.ga_report.controller import _get_publishers _log = logging.getLogger(__name__) +def popular_datasets(count=10): + import random + + publisher = None + publishers = _get_publishers(30) + total = len(publishers) + while not publisher or not datasets: + rand = random.randrange(0, total) + publisher = publishers[rand][0] + if not publisher.state == 'active': + publisher = None + continue + datasets = _datasets_for_publisher(publisher, 10)[:count] + + ctx = { + 'datasets': datasets, + 'publisher': publisher + } + return base.render_snippet('ga_report/ga_popular_datasets.html', **ctx) + +def single_popular_dataset(top=20): + '''Returns a random dataset from the most popular ones. + + :param top: the number of top datasets to select from + ''' + import random + + top_datasets = model.Session.query(GA_Url).\ + filter(GA_Url.url.like('/dataset/%')).\ + order_by('ga_url.pageviews::int desc') + num_top_datasets = top_datasets.count() + + dataset = None + if num_top_datasets: + count = 0 + while not dataset: + rand = random.randrange(0, min(top, num_top_datasets)) + ga_url = top_datasets[rand] + dataset = model.Package.get(ga_url.url[len('/dataset/'):]) + if dataset and not dataset.state == 'active': + dataset = None + count += 1 + if count > 10: + break + if not dataset: + # fallback + dataset = model.Session.query(model.Package)\ + .filter_by(state='active').first() + if not dataset: + return None + dataset_dict = get_action('package_show')({'model': model, + 'session': model.Session}, + {'id':dataset.id}) + return dataset_dict + +def single_popular_dataset_html(top=20): + dataset_dict = single_popular_dataset(top) + groups = package.get('groups', []) + publishers = [ g for g in groups if g.get('type') == 'publisher' ] + publisher = publishers[0] if publishers else {'name':'', 'title': ''} + context = { + 'dataset': dataset_dict, + 'publisher': publisher_dict + } + return base.render_snippet('ga_report/ga_popular_single.html', **context) + + def most_popular_datasets(publisher, count=20): - from ckanext.ga_report.ga_model import GA_Url if not publisher: _log.error("No valid publisher passed to 'most_popular_datasets'") return "" - datasets = {} - entries = model.Session.query(GA_Url).\ - filter(GA_Url.department_id==publisher.name).\ - filter(GA_Url.url.like('/dataset/%')).\ - order_by('ga_url.pageviews::int desc')[:count] - for entry in entries: - p = model.Package.get(entry.url[len('/dataset/'):]) - if not p in datasets: - datasets[p] = {'views':0, 'visits': 0} - datasets[p]['views'] = datasets[p]['views'] + int(entry.pageviews) - datasets[p]['visits'] = datasets[p]['visits'] + int(entry.visitors) - - results = [] - for k, v in datasets.iteritems(): - results.append((k,v['views'],v['visits'])) - - results = sorted(results, key=operator.itemgetter(1), reverse=True) + results = _datasets_for_publisher(publisher, count) ctx = { - 'dataset_count': len(datasets), + 'dataset_count': len(results), 'datasets': results, 'publisher': publisher @@ -39,3 +93,23 @@ return base.render_snippet('ga_report/publisher/popular.html', **ctx) +def _datasets_for_publisher(publisher, count): + datasets = {} + entries = model.Session.query(GA_Url).\ + filter(GA_Url.department_id==publisher.name).\ + filter(GA_Url.url.like('/dataset/%')).\ + order_by('ga_url.pageviews::int desc').all() + for entry in entries: + if len(datasets) < count: + p = model.Package.get(entry.url[len('/dataset/'):]) + if not p in datasets: + datasets[p] = {'views':0, 'visits': 0} + datasets[p]['views'] = datasets[p]['views'] + int(entry.pageviews) + datasets[p]['visits'] = datasets[p]['visits'] + int(entry.visitors) + + results = [] + for k, v in datasets.iteritems(): + results.append((k,v['views'],v['visits'])) + + return sorted(results, key=operator.itemgetter(1), reverse=True) + --- a/ckanext/ga_report/plugin.py +++ b/ckanext/ga_report/plugin.py @@ -2,6 +2,10 @@ import ckan.lib.helpers as h import ckan.plugins as p from ckan.plugins import implements, toolkit + +from ckanext.ga_report.helpers import (most_popular_datasets, + popular_datasets, + single_popular_dataset) log = logging.getLogger('ckanext.ga-report') @@ -19,32 +23,51 @@ A dictionary of extra helpers that will be available to provide ga report info to templates. """ - from ckanext.ga_report.helpers import most_popular_datasets return { 'ga_report_installed': lambda: True, + 'popular_datasets': popular_datasets, 'most_popular_datasets': most_popular_datasets, + 'single_popular_dataset': single_popular_dataset } def after_map(self, map): + # GaReport map.connect( - '/data/analytics/publisher', - controller='ckanext.ga_report.controller:GaPublisherReport', - action='index' - ) - map.connect( - '/data/analytics/publisher/{id}', - controller='ckanext.ga_report.controller:GaPublisherReport', - action='read' - ) - map.connect( - '/data/analytics', + '/data/site-usage', controller='ckanext.ga_report.controller:GaReport', action='index' ) map.connect( - '/data/analytics/data_{month}.csv', + '/data/site-usage/data_{month}.csv', controller='ckanext.ga_report.controller:GaReport', action='csv' + ) + + # GaDatasetReport + map.connect( + '/data/site-usage/publisher', + controller='ckanext.ga_report.controller:GaDatasetReport', + action='publishers' + ) + map.connect( + '/data/site-usage/publishers_{month}.csv', + controller='ckanext.ga_report.controller:GaDatasetReport', + action='publisher_csv' + ) + map.connect( + '/data/site-usage/dataset/datasets_{id}_{month}.csv', + controller='ckanext.ga_report.controller:GaDatasetReport', + action='dataset_csv' + ) + map.connect( + '/data/site-usage/dataset', + controller='ckanext.ga_report.controller:GaDatasetReport', + action='read' + ) + map.connect( + '/data/site-usage/dataset/{id}', + controller='ckanext.ga_report.controller:GaDatasetReport', + action='read_publisher' ) return map --- a/ckanext/ga_report/report_model.py +++ /dev/null --- /dev/null +++ b/ckanext/ga_report/templates/ga_report/ga_popular_datasets.html @@ -1,1 +1,27 @@ + +
+
+ More popular datasets ยป +
+

Popular datasets

+

${publisher.title}

+ + +
+ + + + + + --- /dev/null +++ b/ckanext/ga_report/templates/ga_report/ga_popular_single.html @@ -1,1 +1,31 @@ + + +
+

Featured dataset

+ +
+ +

${dataset['title']}

+ +

+ Publisher : + ${publisher['title']} +

+
${h.truncate(dataset['notes_rendered'], length=200, whole_word=True)}
+
+
+ Other popular datasets + All usage data +
+
+ + + + + + + --- /dev/null +++ b/ckanext/ga_report/templates/ga_report/ga_util.html @@ -1,1 +1,57 @@ + + + + + + + + + + + + + + +
NameSourceVisits
${name}${source}${count}
+ + + + + + + + + + + + + +
Name% ${title}
${name}${value}
+ + +
+
+
+ +
+
+
+ + + + --- /dev/null +++ b/ckanext/ga_report/templates/ga_report/notes.html @@ -1,1 +1,17 @@ + +
  • +

    Notes

    + +
    • + + --- a/ckanext/ga_report/templates/ga_report/publisher/index.html +++ b/ckanext/ga_report/templates/ga_report/publisher/index.html @@ -3,42 +3,53 @@ xmlns:xi="http://www.w3.org/2001/XInclude" py:strip=""> - Publisher Analytics for ${g.site_title} + + + Usage by Publisher
  • -

    Publishers

    -

    The table shows the top 20 publishers as recorded by page views of datasets owned by that publisher, and the number of visits to each publisher's home page.

    +

    Download

    +

    + Download as CSV
    +

    • +     +
    -

    Publisher Analytics

    -

    The top 20 publishers

    -
    +

    Site Usage

    + + ${usage_nav('Publishers')} + + +
    - +
    + - - + -
    PublisherDataset Visits Dataset ViewsVisits
    ${h.link_to(publisher.title, h.url_for(controller='ckanext.ga_report.controller:GaPublisherReport', action='read', id=publisher.name))} + ${h.link_to(publisher.title, h.url_for(controller='ckanext.ga_report.controller:GaDatasetReport', action='read_publisher', id=publisher.name))} ${visits} ${views}${visits}
    @@ -60,5 +71,3 @@ - - --- a/ckanext/ga_report/templates/ga_report/publisher/popular.html +++ b/ckanext/ga_report/templates/ga_report/publisher/popular.html @@ -3,7 +3,7 @@ xmlns:xi="http://www.w3.org/2001/XInclude" py:strip=""> -

    We do not currently have analytics data for ${publisher.title}

    +

    We do not currently have usage data for ${publisher.title}

     @@ -15,7 +15,7 @@ -

    ${h.link_to("More analytics for " + publisher.title, h.url_for(controller='ckanext.ga_report.controller:GaPublisherReport',action='read',id=publisher.name))}

    +

    ${h.link_to("More usage data for " + publisher.title, h.url_for(controller='ckanext.ga_report.controller:GaDatasetReport',action='read_publisher',id=publisher.name))}

    --- a/ckanext/ga_report/templates/ga_report/publisher/read.html +++ b/ckanext/ga_report/templates/ga_report/publisher/read.html @@ -3,56 +3,62 @@ xmlns:xi="http://www.w3.org/2001/XInclude" py:strip=""> - Analytics for ${g.site_title} + - + Usage by Dataset +
  • -

    ${c.publisher.title}

    -

    - The table shows the top 20 most viewed datasets belonging to ${c.publisher.title}. +

    Download

    +

    + Download as CSV

    -

    - As well as showing the number of views within ${c.month_desc}, it will also show the - number of visitors that viewed each dataset. -

    -

    The dataset list page for ${c.publisher.title} was viewed ${c.publisher_page_views} times during ${c.month_desc}

    -

    View the publishers leaderboard

    • +     +
    -

    Analytics for ${c.publisher.title}

    +

    Site Usage

    -

    Top 20 most viewed datasets

    -

    Note: this data does not include API calls

    + ${usage_nav('Datasets')} -
    -
    + +
    - -
    - + + +
    + - +

    ${c.publisher.title}

    + +

    No page views in this period

    +
    + - + - -
    DatasetVisits ViewsVisits
    ${h.link_to(package.title or package.name, h.url_for(controller='package', action='read', id=package.name))} ${visits} ${views}${visits}
    +
    --- a/ckanext/ga_report/templates/ga_report/site/index.html +++ b/ckanext/ga_report/templates/ga_report/site/index.html @@ -3,40 +3,67 @@ xmlns:xi="http://www.w3.org/2001/XInclude" py:strip=""> - Site analytics + + + Site usage
  • -

    Statistics

    -

    It is possible to export the analytics data as a CSV file, which contains all of the information for ${c.month_desc}

    +

    Download

    +

    + Download as CSV
    +

    • -
  • -

    Publisher statistics

    -

    You can view statistics about specific publishers at the Publisher Analytics page

    + -
    • -

    Site statistics

    +

    Site Usage

    + ${usage_nav('Site-wide')}
    - +