From: David Read Date: Fri, 09 Nov 2012 16:15:35 +0000 Subject: Adjust popularity score to take account of number of days in the month. X-Git-Url: http://maxious.lambdacomplex.org/git/?p=ckanext-ga-report.git&a=commitdiff&h=20b6eca0a538a77122ce85cf588045784fa9b67e --- Adjust popularity score to take account of number of days in the month. --- --- a/README.rst +++ b/README.rst @@ -33,10 +33,9 @@ googleanalytics.id = UA-1010101-1 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 + ga-report.bounce_url = / - The ga-report.bounce_url specifies the path to use when calculating bounces. For DGU this is /data - but you may want to set this to /. + The ga-report.bounce_url specifies a particular path to record the bounce rate for. Typically it is / (the home page). 3. Set up this extension's database tables using a paster command. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file):: --- a/ckanext/ga_report/command.py +++ b/ckanext/ga_report/command.py @@ -80,6 +80,11 @@ default=False, dest='delete_first', help='Delete data for the period first') + self.parser.add_option('-s', '--skip_url_stats', + action='store_true', + default=False, + dest='skip_url_stats', + help='Skip the download of URL data - just do site-wide stats') def command(self): self._load_config() @@ -95,7 +100,8 @@ return downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc), - delete_first=self.options.delete_first) + delete_first=self.options.delete_first, + skip_url_stats=self.options.skip_url_stats) time_period = self.args[1] if self.args and len(self.args) > 1 \ else 'latest' --- a/ckanext/ga_report/controller.py +++ b/ckanext/ga_report/controller.py @@ -9,7 +9,7 @@ import sqlalchemy from sqlalchemy import func, cast, Integer import ckan.model as model -from ga_model import GA_Url, GA_Stat, GA_ReferralStat +from ga_model import GA_Url, GA_Stat, GA_ReferralStat, GA_Publisher log = logging.getLogger('ckanext.ga-report') @@ -71,13 +71,13 @@ entries = q.order_by('ga_stat.key').all() def clean_key(key, val): - if key in ['Average time on site', 'Pages per visit', 'New visits', 'Bounces']: + if key in ['Average time on site', 'Pages per visit', 'New visits', 'Bounce rate (home page)']: val = "%.2f" % round(float(val), 2) if key == 'Average time on site': mins, secs = divmod(float(val), 60) hours, mins = divmod(mins, 60) val = '%02d:%02d:%02d (%s seconds) ' % (hours, mins, secs, val) - if key in ['New visits','Bounces']: + if key in ['New visits','Bounce rate (home page)']: val = "%s%%" % val if key in ['Total page views', 'Total visits']: val = int(val) @@ -229,11 +229,10 @@ 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): - '''Returns the datasets in order of visits''' + '''Returns the datasets in order of views''' if count == -1: count = sys.maxint @@ -245,12 +244,11 @@ 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') + q = q.order_by('ga_url.pageviews::int desc') top_packages = [] - for entry,package in q.limit(count): if package: - top_packages.append((package, entry.pageviews, entry.visitors)) + top_packages.append((package, entry.pageviews, entry.visits)) else: log.warning('Could not find package associated package') @@ -289,7 +287,7 @@ else: c.month_desc = ''.join([m[1] for m in c.months if m[0]==c.month]) - month = c.mnth or 'All' + 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) @@ -305,38 +303,33 @@ 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 + select department_id, sum(pageviews::int) views, sum(visits::int) visits from ga_url - where department_id <> ''""" - if c.month: - q = q + """ - and period_name=%s - """ - q = q + """ - group by department_id order by visits desc + 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) - # Add this back (before and period_name =%s) if you want to ignore publisher - # homepage views - # and not url like '/publisher/%%' - top_publishers = [] - res = connection.execute(q, c.month) - + 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) + (name, title) ''' publishers = [] for pub in model.Session.query(model.Group).\ --- a/ckanext/ga_report/download_analytics.py +++ b/ckanext/ga_report/download_analytics.py @@ -3,7 +3,7 @@ import datetime import collections from pylons import config - +from ga_model import _normalize_url import ga_model #from ga_client import GA @@ -17,11 +17,13 @@ class DownloadAnalytics(object): '''Downloads and stores analytics info''' - def __init__(self, service=None, profile_id=None, delete_first=False): + def __init__(self, service=None, profile_id=None, delete_first=False, + skip_url_stats=False): self.period = config['ga-report.period'] self.service = service self.profile_id = profile_id self.delete_first = delete_first + self.skip_url_stats = skip_url_stats def specific_month(self, date): import calendar @@ -92,29 +94,43 @@ def download_and_store(self, periods): for period_name, period_complete_day, start_date, end_date in periods: + 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')) + if self.delete_first: - log.info('Deleting existing Analytics for period "%s"', + log.info('Deleting existing Analytics for this 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-_]+') - 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. + + if not self.skip_url_stats: + # Clean out old url data before storing the new + ga_model.pre_update_url_stats(period_name) + + accountName = config.get('googleanalytics.account') + + log.info('Downloading analytics for dataset views') + data = self.download(start_date, end_date, '~/%s/dataset/[a-z0-9-_]+' % accountName) + + log.info('Storing dataset views (%i rows)', len(data.get('url'))) + self.store(period_name, period_complete_day, data, ) + + log.info('Downloading analytics for publisher views') + data = self.download(start_date, end_date, '~/%s/publisher/[a-z0-9-_]+' % accountName) + + log.info('Storing publisher views (%i rows)', len(data.get('url'))) + self.store(period_name, period_complete_day, data,) + + log.info('Aggregating datasets by publisher') + ga_model.update_publisher_stats(period_name) # about 30 seconds. + + log.info('Downloading and storing analytics for site-wide stats') self.sitewide_stats( period_name ) + log.info('Downloading and storing analytics for social networks') 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') @@ -137,18 +153,17 @@ data = collections.defaultdict(list) rows = results.get('rows',[]) for row in rows: - from ga_model import _normalize_url 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='~/dataset/[a-z0-9-_]+'): + 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:visits' - sort = '-ga:uniquePageviews' + metrics = 'ga:pageviews, ga:visits' + sort = '-ga:pageviews' # Supported query params at # https://developers.google.com/analytics/devguides/reporting/core/v3/reference @@ -165,7 +180,14 @@ packages = [] for entry in results.get('rows'): (loc,pageviews,visits) = entry - packages.append( ('http:/' + loc, pageviews, visits,) ) # Temporary hack + url = _normalize_url('http:/' + loc) # strips off domain e.g. www.data.gov.uk or data.gov.uk + + if not url.startswith('/dataset/') and not url.startswith('/publisher/'): + # filter out strays like: + # /data/user/login?came_from=http://data.gov.uk/dataset/os-code-point-open + # /403.html?page=/about&from=http://data.gov.uk/publisher/planning-inspectorate + continue + packages.append( (url, pageviews, visits,) ) # Temporary hack return dict(url=packages) def store(self, period_name, period_complete_day, data): @@ -179,12 +201,10 @@ start_date = '%s-01' % period_name end_date = '%s-%s' % (period_name, last_day_of_month) - print 'Sitewide_stats for %s (%s -> %s)' % (period_name, start_date, end_date) - funcs = ['_totals_stats', '_social_stats', '_os_stats', '_locale_stats', '_browser_stats', '_mobile_stats'] for f in funcs: - print ' + Fetching %s stats' % f.split('_')[1] + log.info('Downloading analytics for %s' % f.split('_')[1]) getattr(self, f)(start_date, end_date, period_name) def _get_results(result_data, f): @@ -199,8 +219,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', max_results=10000, end_date=end_date).execute() result_data = results.get('rows') @@ -221,25 +241,27 @@ } 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', + # Bounces from / or another configurable page. + path = '/%s%s' % (config.get('googleanalytics.account'), + config.get('ga-report.bounce_url', '/')) + results = self.service.data().ga().get( + ids='ga:' + self.profile_id, + filters='ga:pagePath==%s' % (path,), + start_date=start_date, + metrics='ga:bounces,ga:pageviews', 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}) + if not result_data or len(result_data) != 1: + log.error('Could not pinpoint the bounces for path: %s. Got results: %r', + path, result_data) + return + results = result_data[0] + bounces, total = [float(x) for x in result_data[0][1:]] + pct = 100 * bounces/total + log.info('%d bounces from %d total == %s', bounces, total, pct) + ga_model.update_sitewide_stats(period_name, "Totals", {'Bounce rate (home page)': pct}) def _locale_stats(self, start_date, end_date, period_name): @@ -247,8 +269,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', dimensions="ga:language,ga:country", max_results=10000, end_date=end_date).execute() @@ -271,8 +293,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', dimensions="ga:socialNetwork,ga:referralPath", max_results=10000, end_date=end_date).execute() @@ -290,8 +312,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', dimensions="ga:operatingSystem,ga:operatingSystemVersion", max_results=10000, end_date=end_date).execute() @@ -315,8 +337,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', dimensions="ga:browser,ga:browserVersion", max_results=10000, end_date=end_date).execute() @@ -364,8 +386,8 @@ results = self.service.data().ga().get( ids='ga:' + self.profile_id, start_date=start_date, - metrics='ga:uniquePageviews', - sort='-ga:uniquePageviews', + metrics='ga:pageviews', + sort='-ga:pageviews', dimensions="ga:mobileDeviceBranding, ga:mobileDeviceInfo", max_results=10000, end_date=end_date).execute() --- a/ckanext/ga_report/ga_model.py +++ b/ckanext/ga_report/ga_model.py @@ -9,6 +9,8 @@ import ckan.model as model from ckan.lib.base import * + +log = __import__('logging').getLogger(__name__) def make_uuid(): return unicode(uuid.uuid4()) @@ -27,7 +29,7 @@ Column('period_name', types.UnicodeText), Column('period_complete_day', types.Integer), Column('pageviews', types.UnicodeText), - Column('visitors', types.UnicodeText), + Column('visits', types.UnicodeText), Column('url', types.UnicodeText), Column('department_id', types.UnicodeText), Column('package_id', types.UnicodeText), @@ -63,7 +65,7 @@ Column('period_name', types.UnicodeText), Column('publisher_name', types.UnicodeText), Column('views', types.UnicodeText), - Column('visitors', types.UnicodeText), + Column('visits', types.UnicodeText), Column('toplevel', types.Boolean, default=False), Column('subpublishercount', types.Integer, default=0), Column('parent', types.UnicodeText), @@ -111,12 +113,10 @@ >>> normalize_url('http://data.gov.uk/dataset/weekly_fuel_prices') '/dataset/weekly_fuel_prices' ''' - # Deliberately leaving a / - url = url.replace('http:/','') - return '/' + '/'.join(url.split('/')[2:]) - - -def _get_department_id_of_url(url): + return '/' + '/'.join(url.split('/')[3:]) + + +def _get_package_and_publisher(url): # e.g. /dataset/fuel_prices # e.g. /dataset/fuel_prices/resource/e63380d4 dataset_match = re.match('/dataset/([^/]+)(/.*)?', url) @@ -126,12 +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(): @@ -156,61 +157,72 @@ 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) - - package = None - if url.startswith('/dataset/'): - package = url[len('/dataset/'):] - - # see if the row for this url & month is in the table already + ''' + Given a list of urls and number of hits for each during a given period, + stores them in GA_Url under the period and recalculates the totals for + the 'All' period. + ''' + for url, views, visits in url_data: + package, publisher = _get_package_and_publisher(url) + + 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.package_id = package + item.pageviews = item.pageviews + views + item.visits = item.visits + visits + if not item.package_id: + item.package_id = package + if not item.department_id: + item.department_id = publisher 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, + 'visits': visits, + 'department_id': publisher, 'package_id': package } model.Session.add(GA_Url(**values)) - - # We now need to recaculate the ALL time_period from the data we have - # Delete the old 'All' - old = model.Session.query(GA_Url).\ - filter(GA_Url.period_name == "All").\ - filter(GA_Url.url==url).delete() - - 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() + + if package: + old_pageviews, old_visits = 0, 0 + old = model.Session.query(GA_Url).\ + filter(GA_Url.period_name=='All').\ + filter(GA_Url.url==url).all() + old_pageviews = sum([int(o.pageviews) for o in old]) + old_visits = sum([int(o.visits) for o in old]) + + 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]) + old_pageviews, + 'visits': sum([int(e.visits or 0) for e in entries]) + old_visits, + 'department_id': publisher, + 'package_id': package + } + + model.Session.add(GA_Url(**values)) + model.Session.commit() + + def update_social(period_name, data): @@ -252,7 +264,7 @@ filter(model.Group.type=='publisher').\ filter(model.Group.state=='active').all() for publisher in publishers: - views, visitors, subpub = update_publisher(period_name, publisher, publisher.name) + views, visits, subpub = update_publisher(period_name, publisher, publisher.name) parent, parents = '', publisher.get_groups('publisher') if parents: parent = parents[0].name @@ -261,7 +273,7 @@ filter(GA_Publisher.publisher_name==publisher.name).first() if item: item.views = views - item.visitors = visitors + item.visits = visits item.publisher_name = publisher.name item.toplevel = publisher in toplevel item.subpublishercount = subpub @@ -273,7 +285,7 @@ 'period_name': period_name, 'publisher_name': publisher.name, 'views': views, - 'visitors': visitors, + 'visits': visits, 'toplevel': publisher in toplevel, 'subpublishercount': subpub, 'parent': parent @@ -283,7 +295,7 @@ def update_publisher(period_name, pub, part=''): - views,visitors,subpub = 0, 0, 0 + views,visits,subpub = 0, 0, 0 for publisher in go_down_tree(pub): subpub = subpub + 1 items = model.Session.query(GA_Url).\ @@ -291,9 +303,9 @@ filter(GA_Url.department_id==publisher.name).all() for item in items: views = views + int(item.pageviews) - visitors = visitors + int(item.visitors) - - return views, visitors, (subpub-1) + visits = visits + int(item.visits) + + return views, visits, (subpub-1) def get_top_level(): @@ -333,3 +345,34 @@ q.delete() model.Session.commit() +def get_score_for_dataset(dataset_name): + ''' + Returns a "current popularity" score for a dataset, + based on how many views it has had recently. + ''' + import datetime + now = datetime.datetime.now() + last_month = now - datetime.timedelta(days=30) + period_names = ['%s-%02d' % (last_month.year, last_month.month), + '%s-%02d' % (now.year, now.month), + ] + + score = 0 + for period_name in period_names: + score /= 2 # previous periods are discounted by 50% + entry = model.Session.query(GA_Url)\ + .filter(GA_Url.period_name==period_name)\ + .filter(GA_Url.package_id==dataset_name).first() + # score + if entry: + views = float(entry.pageviews) + if entry.period_complete_day: + views_per_day = views / entry.period_complete_day + else: + views_per_day = views / 15 # guess + score += views_per_day + + score = int(score * 100) + log.debug('Popularity %s: %s', score, dataset_name) + return score + --- a/ckanext/ga_report/helpers.py +++ b/ckanext/ga_report/helpers.py @@ -60,7 +60,8 @@ if not dataset: return None dataset_dict = get_action('package_show')({'model': model, - 'session': model.Session}, + 'session': model.Session, + 'validate': False}, {'id':dataset.id}) return dataset_dict @@ -105,7 +106,7 @@ 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) + datasets[p]['visits'] = datasets[p]['visits'] + int(entry.visits) results = [] for k, v in datasets.iteritems(): --- a/ckanext/ga_report/templates/ga_report/notes.html +++ b/ckanext/ga_report/templates/ga_report/notes.html @@ -6,11 +6,10 @@

  • Notes

    • --- a/ckanext/ga_report/templates/ga_report/publisher/index.html +++ b/ckanext/ga_report/templates/ga_report/publisher/index.html @@ -41,15 +41,15 @@ + - + -
    Publisher Dataset ViewsDataset Visits
    ${h.link_to(publisher.title, h.url_for(controller='ckanext.ga_report.controller:GaDatasetReport', action='read_publisher', id=publisher.name))} ${views}${visits}
    --- a/ckanext/ga_report/templates/ga_report/publisher/read.html +++ b/ckanext/ga_report/templates/ga_report/publisher/read.html @@ -47,15 +47,15 @@ + - + -
    Dataset ViewsVisits
    ${h.link_to(package.title or package.name, h.url_for(controller='package', action='read', id=package.name))} ${views}${visits}
    --- /dev/null +++ b/ckanext/ga_report/tests/test_model.py @@ -1,1 +1,18 @@ +from nose.tools import assert_equal +from ckanext.ga_report.ga_model import _normalize_url + +class TestNormalizeUrl: + def test_normal(self): + assert_equal(_normalize_url('http://data.gov.uk/dataset/weekly_fuel_prices'), + '/dataset/weekly_fuel_prices') + + def test_www_dot(self): + assert_equal(_normalize_url('http://www.data.gov.uk/dataset/weekly_fuel_prices'), + '/dataset/weekly_fuel_prices') + + def test_https(self): + assert_equal(_normalize_url('https://data.gov.uk/dataset/weekly_fuel_prices'), + '/dataset/weekly_fuel_prices') + +