From: Tom Rees Date: Tue, 02 Apr 2013 16:20:29 +0000 Subject: [403] Rickshaw graph always renders all series, no matter which month is selected. X-Git-Url: http://maxious.lambdacomplex.org/git/?p=ckanext-ga-report.git&a=commitdiff&h=2f22129335cc8bba78b345c9094bfcca2fec5b1e --- [403] Rickshaw graph always renders all series, no matter which month is selected. --- --- a/.gitignore +++ b/.gitignore @@ -1,6 +1,7 @@ *.py[co] *.py~ .gitignore +ckan.log # Packages *.egg --- a/README.rst +++ b/README.rst @@ -32,6 +32,7 @@ googleanalytics.id = UA-1010101-1 googleanalytics.account = Account name (e.g. data.gov.uk, see top level item at https://www.google.com/analytics) + googleanalytics.token.filepath = ~/pyenv/token.dat ga-report.period = monthly ga-report.bounce_url = / @@ -82,13 +83,17 @@ $ paster getauthtoken --config=../ckan/development.ini +Now ensure you reference the correct path to your token.dat in your CKAN config file (e.g. development.ini):: + + googleanalytics.token.filepath = ~/pyenv/token.dat + Tutorial -------- Download some GA data and store it in CKAN's database. (Ensure your CKAN pyenv is still activated, run the command from ``src/ckanext-ga-report``, alter the ``--config`` option to point to your site config file) and specifying the name of your auth file (token.dat by default) from the previous step:: - $ paster loadanalytics token.dat latest --config=../ckan/development.ini + $ paster loadanalytics latest --config=../ckan/development.ini The value after the token file is how much data you want to retrieve, this can be --- a/ckanext/ga_report/command.py +++ b/ckanext/ga_report/command.py @@ -1,5 +1,8 @@ import logging import datetime +import os + +from pylons import config from ckan.lib.cli import CkanCommand # No other CKAN imports allowed until _load_config is run, @@ -20,7 +23,7 @@ import ckan.model as model model.Session.remove() model.Session.configure(bind=model.meta.engine) - log = logging.getLogger('ckanext.ga-report') + log = logging.getLogger('ckanext.ga_report') import ga_model ga_model.init_tables() @@ -53,25 +56,52 @@ self.args[0] if self.args else 'credentials.json') +class FixTimePeriods(CkanCommand): + """ + Fixes the 'All' records for GA_Urls + + It is possible that older urls that haven't recently been visited + do not have All records. This command will traverse through those + records and generate valid All records for them. + """ + summary = __doc__.split('\n')[0] + usage = __doc__ + max_args = 0 + min_args = 0 + + def __init__(self, name): + super(FixTimePeriods, self).__init__(name) + + def command(self): + import ckan.model as model + from ga_model import post_update_url_stats + self._load_config() + model.Session.remove() + model.Session.configure(bind=model.meta.engine) + + log = logging.getLogger('ckanext.ga_report') + + log.info("Updating 'All' records for old URLs") + post_update_url_stats() + log.info("Processing complete") + + class LoadAnalytics(CkanCommand): """Get data from Google Analytics API and save it in the ga_model - Usage: paster loadanalytics + Usage: paster loadanalytics - Where is the name of the auth token file from - the getauthtoken step. - - And where is: + Where is: all - data for all time latest - (default) just the 'latest' data YYYY-MM - just data for the specific month """ summary = __doc__.split('\n')[0] usage = __doc__ - max_args = 2 - min_args = 1 + max_args = 1 + min_args = 0 def __init__(self, name): super(LoadAnalytics, self).__init__(name) @@ -92,19 +122,25 @@ from download_analytics import DownloadAnalytics from ga_auth import (init_service, get_profile_id) + ga_token_filepath = os.path.expanduser(config.get('googleanalytics.token.filepath', '')) + if not ga_token_filepath: + print 'ERROR: In the CKAN config you need to specify the filepath of the ' \ + 'Google Analytics token file under key: googleanalytics.token.filepath' + return + try: - svc = init_service(self.args[0], None) + svc = init_service(ga_token_filepath, None) except TypeError: print ('Have you correctly run the getauthtoken task and ' - 'specified the correct token file?') + 'specified the correct token file in the CKAN config under ' + '"googleanalytics.token.filepath"?') return downloader = DownloadAnalytics(svc, profile_id=get_profile_id(svc), delete_first=self.options.delete_first, skip_url_stats=self.options.skip_url_stats) - time_period = self.args[1] if self.args and len(self.args) > 1 \ - else 'latest' + time_period = self.args[0] if self.args else 'latest' if time_period == 'all': downloader.all_() elif time_period == 'latest': --- 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 @@ -13,6 +14,7 @@ log = logging.getLogger('ckanext.ga-report') +DOWNLOADS_AVAILABLE_FROM = '2012-12' def _get_month_name(strdate): import calendar @@ -20,14 +22,39 @@ d = strptime(strdate, '%Y-%m') return '%s %s' % (calendar.month_name[d.tm_mon], d.tm_year) - -def _month_details(cls): - '''Returns a list of all the month names''' +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).filter(cls.period_name!='All').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): @@ -35,7 +62,7 @@ def csv(self, month): import csv - q = model.Session.query(GA_Stat) + 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() @@ -52,11 +79,12 @@ 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_desc = 'all months' @@ -84,11 +112,26 @@ 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) - c.global_totals.append((key, val)) + sparkline = sparkline_data[e.key] + c.global_totals.append((key, val, sparkline)) else: d = collections.defaultdict(list) for e in entries: @@ -97,11 +140,19 @@ if k in ['Total page views', 'Total visits']: v = sum(v) else: - v = float(sum(v))/len(v) + v = float(sum(v))/float(len(v)) + sparkline = sparkline_data[k] key, val = clean_key(k,v) - c.global_totals.append((key, val)) - c.global_totals = sorted(c.global_totals, key=operator.itemgetter(0)) + c.global_totals.append((key, val, sparkline)) + # Sort the global totals into a more pleasant order + def sort_func(x): + key = x[0] + total_order = ['Total page views','Total visits','Pages per visit'] + if key in total_order: + return total_order.index(key) + return 999 + c.global_totals = sorted(c.global_totals, key=sort_func) keys = { 'Browser versions': 'browser_versions', @@ -138,12 +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) @@ -152,10 +204,30 @@ entries.append((key,val,)) entries = sorted(entries, key=operator.itemgetter(1), reverse=True) + # Run a query on all months to gather graph data + graph_query = model.Session.query(GA_Stat).\ + filter(GA_Stat.stat_name==k).\ + order_by(GA_Stat.period_name) + graph_dict = {} + for stat in graph_query: + graph_dict[ stat.key ] = graph_dict.get(stat.key,{ + 'name':stat.key, + 'data': [] + }) + graph_dict[ stat.key ]['data'].append({ + 'x':_get_unix_epoch(stat.period_name), + 'y':float(stat.value) + }) + 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': - total = sum([x for n,x in c.global_totals if n == 'Total visits']) + 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 ]) @@ -180,7 +252,9 @@ 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, top_publishers_graph = _get_top_publishers(None) + + for publisher,view,visit in top_publishers: writer.writerow([publisher.title.encode('utf-8'), publisher.name.encode('utf-8'), view, @@ -206,13 +280,14 @@ 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(["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): @@ -220,7 +295,7 @@ # 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', '') @@ -228,15 +303,17 @@ 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() + c.top_publishers, graph_data = _get_top_publishers() + c.top_publishers_graph = json.dumps( _to_rickshaw(graph_data) ) + return render('ga_report/publisher/index.html') def _get_packages(self, publisher=None, count=-1): '''Returns the datasets in order of views''' - if count == -1: - count = sys.maxint - + 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)\ @@ -246,9 +323,26 @@ q = q.filter(GA_Url.period_name==month) q = q.order_by('ga_url.pageviews::int desc') top_packages = [] - for entry,package in q.limit(count): + if count == -1: + entries = q.all() + else: + entries = q.limit(count) + + for entry,package in entries: if package: - top_packages.append((package, entry.pageviews, entry.visits)) + # 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') @@ -278,7 +372,7 @@ # 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', '') @@ -296,7 +390,81 @@ c.top_packages = self._get_packages(c.publisher, 20) + # Graph query + top_package_names = [ x[0].name for x in c.top_packages ] + graph_query = model.Session.query(GA_Url,model.Package)\ + .filter(model.Package.name==GA_Url.package_id)\ + .filter(GA_Url.url.like('/dataset/%'))\ + .filter(GA_Url.package_id.in_(top_package_names)) + graph_dict = {} + for entry,package in graph_query: + if not package: continue + if entry.period_name=='All': continue + graph_dict[package.name] = graph_dict.get(package.name,{ + 'name':package.title, + 'data':[] + }) + graph_dict[package.name]['data'].append({ + 'x':_get_unix_epoch(entry.period_name), + 'y':int(entry.pageviews), + }) + graph = [ graph_dict[x] for x in top_package_names ] + + c.graph_data = json.dumps( _to_rickshaw(graph) ) + return render('ga_report/publisher/read.html') + +def _to_rickshaw(data, percentageMode=False): + if data==[]: + return data + # Create a consistent x-axis between all series + num_points = [ len(series['data']) for series in data ] + ideal_index = num_points.index( max(num_points) ) + x_axis = [] + for series in data: + for point in series['data']: + x_axis.append(point['x']) + x_axis = sorted( list( set(x_axis) ) ) + # Zero pad any missing values + for series in data: + xs = [ point['x'] for point in series['data'] ] + for x in set(x_axis).difference(set(xs)): + series['data'].append( {'x':x, 'y':0} ) + if percentageMode: + def get_totals(series_list): + totals = {} + for series in series_list: + for point in series['data']: + totals[point['x']] = totals.get(point['x'],0) + point['y'] + return totals + # Transform data into percentage stacks + totals = get_totals(data) + # Roll insignificant series into a catch-all + THRESHOLD = 0.01 + raw_data = data + data = [] + for series in raw_data: + for point in series['data']: + fraction = float(point['y']) / totals[point['x']] + if not (series in data) and fraction>THRESHOLD: + data.append(series) + # Overwrite data with a set of interesting series + others = [ x for x in raw_data if not (x in data) ] + data.append({ + 'name':'Other', + 'data': [ {'x':x,'y':y} for x,y in get_totals(others).items() ] + }) + # Turn each point into a percentage + for series in data: + for point in series['data']: + point['y'] = (point['y']*100) / totals[point['x']] + # Sort the points + for series in data: + series['data'] = sorted( series['data'], key=lambda x:x['x'] ) + # Strip the latest month's incomplete analytics + series['data'] = series['data'][:-1] + return data + def _get_top_publishers(limit=20): ''' @@ -319,11 +487,39 @@ top_publishers = [] res = connection.execute(q, month) + department_ids = [] for row in res: g = model.Group.get(row[0]) if g: + department_ids.append(row[0]) top_publishers.append((g, row[1], row[2])) - return top_publishers + + graph = [] + if limit is not None: + # Query for a history graph of these publishers + q = model.Session.query( + GA_Url.department_id, + GA_Url.period_name, + func.sum(cast(GA_Url.pageviews,sqlalchemy.types.INT)))\ + .filter( GA_Url.department_id.in_(department_ids) )\ + .filter( GA_Url.period_name!='All' )\ + .filter( GA_Url.url.like('/dataset/%') )\ + .filter( GA_Url.package_id!='' )\ + .group_by( GA_Url.department_id, GA_Url.period_name ) + graph_dict = {} + for dept_id,period_name,views in q: + graph_dict[dept_id] = graph_dict.get( dept_id, { + 'name' : model.Group.get(dept_id).title, + 'data' : [] + }) + graph_dict[dept_id]['data'].append({ + 'x': _get_unix_epoch(period_name), + 'y': views + }) + # Sort dict into ordered list + for id in department_ids: + graph.append( graph_dict[id] ) + return top_publishers, graph def _get_publishers(): --- a/ckanext/ga_report/download_analytics.py +++ b/ckanext/ga_report/download_analytics.py @@ -13,6 +13,7 @@ FORMAT_MONTH = '%Y-%m' MIN_VIEWS = 50 MIN_VISITS = 20 +MIN_DOWNLOADS = 10 class DownloadAnalytics(object): '''Downloads and stores analytics info''' @@ -31,6 +32,11 @@ first_of_this_month = datetime.datetime(date.year, date.month, 1) _, last_day_of_month = calendar.monthrange(int(date.year), int(date.month)) last_of_this_month = datetime.datetime(date.year, date.month, last_day_of_month) + # if this is the latest month, note that it is only up until today + now = datetime.datetime.now() + if now.year == date.year and now.month == date.month: + last_day_of_month = now.day + last_of_this_month = now periods = ((date.strftime(FORMAT_MONTH), last_day_of_month, first_of_this_month, last_of_this_month),) @@ -122,11 +128,15 @@ 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') + # Make sure the All records are correct. + ga_model.post_update_url_stats() + + log.info('Associating datasets with their 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 ) + self.sitewide_stats( period_name, period_complete_day ) log.info('Downloading and storing analytics for social networks') self.update_social_info(period_name, start_date, end_date) @@ -153,7 +163,8 @@ data = collections.defaultdict(list) rows = results.get('rows',[]) for row in rows: - data[_normalize_url(row[0])].append( (row[1], int(row[2]),) ) + url = _normalize_url('http:/' + row[0]) + data[url].append( (row[1], int(row[2]),) ) ga_model.update_social(period_name, data) @@ -178,6 +189,7 @@ end_date=end_date).execute() packages = [] + log.info("There are %d results" % results['totalResults']) for entry in results.get('rows'): (loc,pageviews,visits) = entry url = _normalize_url('http:/' + loc) # strips off domain e.g. www.data.gov.uk or data.gov.uk @@ -194,7 +206,7 @@ if 'url' in data: ga_model.update_url_stats(period_name, period_complete_day, data['url']) - def sitewide_stats(self, period_name): + def sitewide_stats(self, period_name, period_complete_day): import calendar year, month = period_name.split('-') _, last_day_of_month = calendar.monthrange(int(year), int(month)) @@ -202,10 +214,10 @@ start_date = '%s-01' % period_name end_date = '%s-%s' % (period_name, last_day_of_month) funcs = ['_totals_stats', '_social_stats', '_os_stats', - '_locale_stats', '_browser_stats', '_mobile_stats'] + '_locale_stats', '_browser_stats', '_mobile_stats', '_download_stats'] for f in funcs: log.info('Downloading analytics for %s' % f.split('_')[1]) - getattr(self, f)(start_date, end_date, period_name) + getattr(self, f)(start_date, end_date, period_name, period_complete_day) def _get_results(result_data, f): data = {} @@ -214,7 +226,7 @@ data[key] = data.get(key,0) + result[1] return data - def _totals_stats(self, start_date, end_date, period_name): + def _totals_stats(self, start_date, end_date, period_name, period_complete_day): """ Fetches distinct totals, total pageviews etc """ results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -224,7 +236,8 @@ max_results=10000, end_date=end_date).execute() result_data = results.get('rows') - ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]}) + ga_model.update_sitewide_stats(period_name, "Totals", {'Total page views': result_data[0][0]}, + period_complete_day) results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -239,7 +252,7 @@ 'New visits': result_data[0][2], 'Total visits': result_data[0][3], } - ga_model.update_sitewide_stats(period_name, "Totals", data) + ga_model.update_sitewide_stats(period_name, "Totals", data, period_complete_day) # Bounces from / or another configurable page. path = '/%s%s' % (config.get('googleanalytics.account'), @@ -248,7 +261,7 @@ ids='ga:' + self.profile_id, filters='ga:pagePath==%s' % (path,), start_date=start_date, - metrics='ga:bounces,ga:pageviews', + metrics='ga:visitBounceRate', dimensions='ga:pagePath', max_results=10000, end_date=end_date).execute() @@ -258,13 +271,14 @@ 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): + bounces = float(results[1]) + # visitBounceRate is already a % + log.info('Google reports visitBounceRate as %s', bounces) + ga_model.update_sitewide_stats(period_name, "Totals", {'Bounce rate (home page)': float(bounces)}, + period_complete_day) + + + def _locale_stats(self, start_date, end_date, period_name, period_complete_day): """ Fetches stats about language and country """ results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -279,16 +293,84 @@ 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) + ga_model.update_sitewide_stats(period_name, "Languages", data, period_complete_day) 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) - - - def _social_stats(self, start_date, end_date, period_name): + ga_model.update_sitewide_stats(period_name, "Country", data, period_complete_day) + + + def _download_stats(self, start_date, end_date, period_name, period_complete_day): + """ Fetches stats about data downloads """ + import ckan.model as model + + data = {} + + results = self.service.data().ga().get( + ids='ga:' + self.profile_id, + start_date=start_date, + filters='ga:eventAction==download', + metrics='ga:totalEvents', + sort='-ga:totalEvents', + dimensions="ga:eventLabel", + max_results=10000, + end_date=end_date).execute() + result_data = results.get('rows') + if not result_data: + # We may not have data for this time period, so we need to bail + # early. + log.info("There is no download data for this time period") + return + + def process_result_data(result_data, cached=False): + progress_total = len(result_data) + progress_count = 0 + resources_not_matched = [] + for result in result_data: + progress_count += 1 + if progress_count % 100 == 0: + log.debug('.. %d/%d done so far', progress_count, progress_total) + + url = result[0].strip() + + # Get package id associated with the resource that has this URL. + q = model.Session.query(model.Resource) + if cached: + r = q.filter(model.Resource.cache_url.like("%s%%" % url)).first() + else: + r = q.filter(model.Resource.url.like("%s%%" % url)).first() + + package_name = r.resource_group.package.name if r else "" + if package_name: + data[package_name] = data.get(package_name, 0) + int(result[1]) + else: + resources_not_matched.append(url) + continue + if resources_not_matched: + log.debug('Could not match %i or %i resource URLs to datasets. e.g. %r', + len(resources_not_matched), progress_total, resources_not_matched[:3]) + + log.info('Associating downloads of resource URLs with their respective datasets') + process_result_data(results.get('rows')) + + results = self.service.data().ga().get( + ids='ga:' + self.profile_id, + start_date=start_date, + filters='ga:eventAction==download-cache', + metrics='ga:totalEvents', + sort='-ga:totalEvents', + dimensions="ga:eventLabel", + max_results=10000, + end_date=end_date).execute() + log.info('Associating downloads of cache resource URLs with their respective datasets') + process_result_data(results.get('rows'), cached=False) + + self._filter_out_long_tail(data, MIN_DOWNLOADS) + ga_model.update_sitewide_stats(period_name, "Downloads", data, period_complete_day) + + def _social_stats(self, start_date, end_date, period_name, period_complete_day): """ Finds out which social sites people are referred from """ results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -304,10 +386,10 @@ if not result[0] == '(not set)': data[result[0]] = data.get(result[0], 0) + int(result[2]) self._filter_out_long_tail(data, 3) - ga_model.update_sitewide_stats(period_name, "Social sources", data) - - - def _os_stats(self, start_date, end_date, period_name): + ga_model.update_sitewide_stats(period_name, "Social sources", data, period_complete_day) + + + def _os_stats(self, start_date, end_date, period_name, period_complete_day): """ Operating system stats """ results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -322,17 +404,17 @@ 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) + ga_model.update_sitewide_stats(period_name, "Operating Systems", data, period_complete_day) data = {} for result in result_data: 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) - - - def _browser_stats(self, start_date, end_date, period_name): + ga_model.update_sitewide_stats(period_name, "Operating Systems versions", data, period_complete_day) + + + def _browser_stats(self, start_date, end_date, period_name, period_complete_day): """ Information about browsers and browser versions """ results = self.service.data().ga().get( ids='ga:' + self.profile_id, @@ -349,14 +431,14 @@ 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) + ga_model.update_sitewide_stats(period_name, "Browsers", data, period_complete_day) data = {} for result in result_data: 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) + ga_model.update_sitewide_stats(period_name, "Browser versions", data, period_complete_day) @classmethod def _filter_browser_version(cls, browser, version_str): @@ -380,7 +462,7 @@ ver = ver[0] + ver[1] + 'X' * num_hidden_digits return ver - def _mobile_stats(self, start_date, end_date, period_name): + def _mobile_stats(self, start_date, end_date, period_name, period_complete_day): """ Info about mobile devices """ results = self.service.data().ga().get( @@ -397,13 +479,13 @@ 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) + ga_model.update_sitewide_stats(period_name, "Mobile brands", data, period_complete_day) 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) + ga_model.update_sitewide_stats(period_name, "Mobile devices", data, period_complete_day) @classmethod def _filter_out_long_tail(cls, data, threshold=10): --- 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()) @@ -45,6 +47,7 @@ Column('id', types.UnicodeText, primary_key=True, default=make_uuid), Column('period_name', types.UnicodeText), + Column('period_complete_day', types.UnicodeText), Column('stat_name', types.UnicodeText), Column('key', types.UnicodeText), Column('value', types.UnicodeText), ) @@ -132,7 +135,7 @@ return None, publisher_match.groups()[0] return None, None -def update_sitewide_stats(period_name, stat_name, data): +def update_sitewide_stats(period_name, stat_name, data, period_complete_day): for k,v in data.iteritems(): item = model.Session.query(GA_Stat).\ filter(GA_Stat.period_name==period_name).\ @@ -142,11 +145,13 @@ item.period_name = period_name item.key = k item.value = v + item.period_complete_day = period_complete_day model.Session.add(item) else: # create the row values = {'id': make_uuid(), 'period_name': period_name, + 'period_complete_day': period_complete_day, 'key': k, 'value': v, 'stat_name': stat_name @@ -156,10 +161,64 @@ 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() + q = model.Session.query(GA_Url).\ + filter(GA_Url.period_name==period_name) + log.debug("Deleting %d '%s' records" % (q.count(), period_name)) + q.delete() + + q = model.Session.query(GA_Url).\ + filter(GA_Url.period_name == 'All') + log.debug("Deleting %d 'All' records..." % q.count()) + q.delete() + + model.Session.flush() + model.Session.commit() + model.repo.commit_and_remove() + log.debug('...done') + +def post_update_url_stats(): + + """ Check the distinct url field in ga_url and make sure + it has an All record. If not then create one. + + After running this then every URL should have an All + record regardless of whether the URL has an entry for + the month being currently processed. + """ + log.debug('Post-processing "All" records...') + query = """select url, pageviews::int, visits::int + from ga_url + where url not in (select url from ga_url where period_name ='All')""" + connection = model.Session.connection() + res = connection.execute(query) + + views, visits = {}, {} + # url, views, visits + for row in res: + views[row[0]] = views.get(row[0], 0) + row[1] + visits[row[0]] = visits.get(row[0], 0) + row[2] + + progress_total = len(views.keys()) + progress_count = 0 + for key in views.keys(): + progress_count += 1 + if progress_count % 100 == 0: + log.debug('.. %d/%d done so far', progress_count, progress_total) + + package, publisher = _get_package_and_publisher(key) + + values = {'id': make_uuid(), + 'period_name': "All", + 'period_complete_day': 0, + 'url': key, + 'pageviews': views[key], + 'visits': visits[key], + 'department_id': publisher, + 'package_id': package + } + model.Session.add(GA_Url(**values)) + model.Session.commit() + log.debug('..done') def update_url_stats(period_name, period_complete_day, url_data): @@ -168,9 +227,14 @@ stores them in GA_Url under the period and recalculates the totals for the 'All' period. ''' + progress_total = len(url_data) + progress_count = 0 for url, views, visits in url_data: + progress_count += 1 + if progress_count % 100 == 0: + log.debug('.. %d/%d done so far', progress_count, progress_total) + package, publisher = _get_package_and_publisher(url) - item = model.Session.query(GA_Url).\ filter(GA_Url.period_name==period_name).\ @@ -211,8 +275,8 @@ '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) for e in entries]) + old_visits, + 'pageviews': sum([int(e.pageviews) for e in entries]) + int(old_pageviews), + 'visits': sum([int(e.visits or 0) for e in entries]) + int(old_visits), 'department_id': publisher, 'package_id': package } @@ -338,8 +402,39 @@ ''' for object_type in (GA_Url, GA_Stat, GA_Publisher, GA_ReferralStat): q = model.Session.query(object_type) - if period_name != 'all': + if period_name != 'All': q = q.filter_by(period_name=period_name) q.delete() - model.Session.commit() - + model.repo.commit_and_remove() + +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 @@ -50,9 +50,12 @@ dataset = model.Package.get(ga_url.url[len('/dataset/'):]) if dataset and not dataset.state == 'active': dataset = None - count += 1 - if count > 10: - break + # When testing, it is possible that top datasets are not available + # so only go round this loop a few times before falling back on + # a random dataset. + count += 1 + if count > 10: + break if not dataset: # fallback dataset = model.Session.query(model.Package)\ @@ -77,7 +80,7 @@ return base.render_snippet('ga_report/ga_popular_single.html', **context) -def most_popular_datasets(publisher, count=20): +def most_popular_datasets(publisher, count=20, preview_image=None): if not publisher: _log.error("No valid publisher passed to 'most_popular_datasets'") @@ -89,7 +92,8 @@ 'dataset_count': len(results), 'datasets': results, - 'publisher': publisher + 'publisher': publisher, + 'preview_image': preview_image } return base.render_snippet('ga_report/publisher/popular.html', **ctx) @@ -103,6 +107,10 @@ for entry in entries: if len(datasets) < count: p = model.Package.get(entry.url[len('/dataset/'):]) + if not p: + _log.warning("Could not find Package for {url}".format(url=entry.url)) + continue + if not p in datasets: datasets[p] = {'views':0, 'visits': 0} datasets[p]['views'] = datasets[p]['views'] + int(entry.pageviews) --- a/ckanext/ga_report/plugin.py +++ b/ckanext/ga_report/plugin.py @@ -42,6 +42,16 @@ controller='ckanext.ga_report.controller:GaReport', action='csv' ) + map.connect( + '/data/site-usage/downloads', + controller='ckanext.ga_report.controller:GaReport', + action='downloads' + ) + map.connect( + '/data/site-usage/downloads_{month}.csv', + controller='ckanext.ga_report.controller:GaReport', + action='csv_downloads' + ) # GaDatasetReport map.connect( --- /dev/null +++ b/ckanext/ga_report/public/css/ga_report.css @@ -1,1 +1,64 @@ +.table-condensed td.sparkline-cell { + padding: 1px 0 0 0; + width: 108px; + text-align: center; + /* Hack to hide the momentary flash of text + * before sparklines are fully rendered */ + font-size: 1px; + color: transparent; + overflow: hidden; +} +.rickshaw_chart_container { + position: relative; + height: 350px; + margin: 0 auto 20px auto; +} +.rickshaw_chart { + position: absolute; + left: 40px; + width: 500px; + top: 0; + bottom: 0; +} +.rickshaw_legend { + background: transparent; + width: 100%; + padding-top: 4px; +} +.rickshaw_y_axis { + position: absolute; + top: 0; + bottom: 0; + width: 40px; +} +.rickshaw_legend .label { + background: transparent !important; + color: #000000 !important; + font-weight: normal !important; +} +.rickshaw_legend .instructions { + color: #000; + margin-bottom: 6px; +} +.rickshaw_legend .line .action { + display: none; +} +.rickshaw_legend .line .swatch { + display: block; + float: left; +} +.rickshaw_legend .line .label { + display: block; + white-space: normal; + float: left; + width: 200px; +} +.rickshaw_legend .line .label:hover { + text-decoration: underline; +} + +.ga-reports-table .td-numeric { + text-align: center; +} + --- /dev/null +++ b/ckanext/ga_report/public/scripts/ckanext_ga_reports.js @@ -1,1 +1,122 @@ +var CKAN = CKAN || {}; +CKAN.GA_Reports = {}; +CKAN.GA_Reports.render_rickshaw = function( css_name, data, mode, colorscheme ) { + var graphLegends = $('#graph-legend-container'); + + if (!Modernizr.svg) { + $("#chart_"+css_name) + .html( '
Your browser does not support vector graphics. No graphs can be rendered.
') + .closest('.rickshaw_chart_container').css('height',50); + var myLegend = $('
') + .html('(Graph cannot be rendered)') + .appendTo(graphLegends); + return; + } + var myLegend = $('
').appendTo(graphLegends); + + var palette = new Rickshaw.Color.Palette( { scheme: colorscheme } ); + $.each(data, function(i, object) { + object['color'] = palette.color(); + }); + // Rickshaw renders the legend in reverse order... + data.reverse(); + + var graphElement = document.querySelector("#chart_"+css_name); + + var graph = new Rickshaw.Graph( { + element: document.querySelector("#chart_"+css_name), + renderer: mode, + series: data , + height: 328 + }); + var x_axis = new Rickshaw.Graph.Axis.Time( { graph: graph } ); + var y_axis = new Rickshaw.Graph.Axis.Y( { + graph: graph, + orientation: 'left', + tickFormat: Rickshaw.Fixtures.Number.formatKMBT, + element: document.getElementById('y_axis_'+css_name) + } ); + var legend = new Rickshaw.Graph.Legend( { + element: document.querySelector('#legend_'+css_name), + graph: graph + } ); + var shelving = new Rickshaw.Graph.Behavior.Series.Toggle( { + graph: graph, + legend: legend + } ); + myLegend.prepend('
Click on a series below to isolate its graph:
'); + graph.render(); +}; + +CKAN.GA_Reports.bind_sparklines = function() { + /* + * Bind to the 'totals' tab being on screen, when the + * Sparkline graphs should be drawn. + * Note that they cannot be drawn sooner. + */ + var created = false; + $('a[href="#totals"]').on( + 'shown', + function() { + if (!created) { + var sparkOptions = { + enableTagOptions: true, + type: 'line', + width: 100, + height: 26, + chartRangeMin: 0, + spotColor: '', + maxSpotColor: '', + minSpotColor: '', + highlightSpotColor: '#000000', + lineColor: '#3F8E6D', + fillColor: '#B7E66B' + }; + $('.sparkline').sparkline('html',sparkOptions); + created = true; + } + $.sparkline_display_visible(); + } + ); +}; + +CKAN.GA_Reports.bind_sidebar = function() { + /* + * Bind to changes in the tab behaviour: + * Show the correct rickshaw graph in the sidebar. + * Not to be called before all graphs load. + */ + $('a[data-toggle="hashtab"]').on( + 'shown', + function(e) { + var href = $(e.target).attr('href'); + var pane = $(href); + if (!pane.length) { console.err('bad href',href); return; } + var legend_name = "none"; + var graph = pane.find('.rickshaw_chart'); + if (graph.length) { + legend_name = graph.attr('id').replace('chart_',''); + } + legend_name = '#legend_'+legend_name; + $('#graph-legend-container > *').hide(); + $('#graph-legend-container .instructions').show(); + $(legend_name).show(); + } + ); + /* The first tab might already have been shown */ + $('li.active > a[data-toggle="hashtab"]').trigger('shown'); +}; + +CKAN.GA_Reports.bind_month_selector = function() { + var handler = function(e) { + var target = $(e.delegateTarget); + var form = target.closest('form'); + var url = form.attr('action')+'?month='+target.val()+window.location.hash; + window.location = url; + }; + var selectors = $('select[name="month"]'); + assert(selectors.length>0); + selectors.bind('change', handler); +}; + --- /dev/null +++ b/ckanext/ga_report/public/scripts/rickshaw_ie7_shim.js @@ -1,1 +1,109 @@ +/* + * Collection of shims to allow d3 and Rickshaw to load, error-free + * (but ultimately unusable) on Internet Explorer 7. The browser's + * API lacks several crucial functions which these libraries depend + * upon to load; we try to hide these errors from the user. + * + * With thanks to Array functions from: + * http://stackoverflow.com/questions/2790001/fixing-javascript-array-functions-in-internet-explorer-indexof-foreach-etc + * + * Use (Modernizr.svg==true) to detect whether it's okay to draw a graph. + */ +'use strict'; +window.Element = window.Element || {'prototype': {}}; +window.CSSStyleDeclaration = window.CSSStyleDeclaration || {'prototype':{}}; + +// Add ECMA262-5 method binding if not supported natively +// +if (!('bind' in Function.prototype)) { + Function.prototype.bind= function(owner) { + var that= this; + if (arguments.length<=1) { + return function() { + return that.apply(owner, arguments); + }; + } else { + var args= Array.prototype.slice.call(arguments, 1); + return function() { + return that.apply(owner, arguments.length===0? args : args.concat(Array.prototype.slice.call(arguments))); + }; + } + }; +} + +// Add ECMA262-5 string trim if not supported natively +// +if (!('trim' in String.prototype)) { + String.prototype.trim= function() { + return this.replace(/^\s+/, '').replace(/\s+$/, ''); + }; +} + +// Add ECMA262-5 Array methods if not supported natively +// +if (!('indexOf' in Array.prototype)) { + Array.prototype.indexOf= function(find, i /*opt*/) { + if (i===undefined) i= 0; + if (i<0) i+= this.length; + if (i<0) i= 0; + for (var n= this.length; ithis.length-1) i= this.length-1; + for (i++; i-->0;) /* i++ because from-argument is sadly inclusive */ + if (i in this && this[i]===find) + return i; + return -1; + }; +} +if (!('forEach' in Array.prototype)) { + Array.prototype.forEach= function(action, that /*opt*/) { + for (var i= 0, n= this.length; id&&(c=b,d=e);return c}function u(a){return a.reduce(v,0)}function v(a,b){return a+b[1]}function w(a,b){return x(a,Math.ceil(Math.log(b.length)/Math.LN2+1))}function x(a,b){var c=-1,d=+a[0],e=(a[1]-d)/b,f=[];while(++c<=b)f[c]=e*c+d;return f}function y(a){return[d3.min(a),d3.max(a)]}function z(a,b){return a.sort=d3.rebind(a,b.sort),a.children=d3.rebind(a,b.children),a.links=D,a.value=d3.rebind(a,b.value),a.nodes=function(b){return E=!0,(a.nodes=a)(b)},a}function A(a){return a.children}function B(a){return a.value}function C(a,b){return b.value-a.value}function D(a){return d3.merge(a.map(function(a){return(a.children||[]).map(function(b){return{source:a,target:b}})}))}function F(a,b){return a.value-b.value}function G(a,b){var c=a._pack_next;a._pack_next=b,b._pack_prev=a,b._pack_next=c,c._pack_prev=b}function H(a,b){a._pack_next=b,b._pack_prev=a}function I(a,b){var c=b.x-a.x,d=b.y-a.y,e=a.r+b.r;return e*e-c*c-d*d>.001}function J(a){function l(a){b=Math.min(a.x-a.r,b),c=Math.max(a.x+a.r,c),d=Math.min(a.y-a.r,d),e=Math.max(a.y+a.r,e)}var b=Infinity,c=-Infinity,d=Infinity,e=-Infinity,f=a.length,g,h,i,j,k;a.forEach(K),g=a[0],g.x=-g.r,g.y=0,l(g);if(f>1){h=a[1],h.x=h.r,h.y=0,l(h);if(f>2){i=a[2],O(g,h,i),l(i),G(g,i),g._pack_prev=i,G(i,h),h=g._pack_next;for(var m=3;m0?(H(g,j),h=j,m--):(H(j,h),g=j,m--)}}}var q=(b+c)/2,r=(d+e)/2,s=0;for(var m=0;m0&&(a=d)}return a}function X(a,b){return a.x-b.x}function Y(a,b){return b.x-a.x}function Z(a,b){return a.depth-b.depth}function $(a,b){function c(a,d){var e=a.children;if(e&&(i=e.length)){var f,g=null,h=-1,i;while(++h=0)f=d[e]._tree,f.prelim+=b,f.mod+=b,b+=f.shift+(c+=f.change)}function ba(a,b,c){a=a._tree,b=b._tree;var d=c/(b.number-a.number);a.change+=d,b.change-=d,b.shift+=c,b.prelim+=c,b.mod+=c}function bb(a,b,c){return a._tree.ancestor.parent==b.parent?a._tree.ancestor:c}function bc(a){return{x:a.x,y:a.y,dx:a.dx,dy:a.dy}}function bd(a,b){var c=a.x+b[3],d=a.y+b[0],e=a.dx-b[1]-b[3],f=a.dy-b[0]-b[2];return e<0&&(c+=e/2,e=0),f<0&&(d+=f/2,f=0),{x:c,y:d,dx:e,dy:f}}d3.layout={},d3.layout.bundle=function(){return function(b){var c=[],d=-1,e=b.length;while(++de&&(e=h),d.push(h)}for(g=0;g