def __init__(self):
self.mp = Mixpanel(api_key=ENV["mp"]["api_key"], api_secret=ENV["mp"]["api_secret"])
class MPService():
def __init__(self):
self.mp = Mixpanel(api_key=ENV["mp"]["api_key"], api_secret=ENV["mp"]["api_secret"])
#
# get list of most common events over the past 31 days
#
def get_event_list(self):
return self.mp.request(['events', 'names'], {'type': 'general', 'limit':5})
#
# get Events data
#
def get_events_data(self, events, start_date, end_date, unit, event_type, viz="line"):
interval = (end_date - start_date).days
today_date = datetime.utcnow()
# mix panel always returns events data using end_data as today's date
# diff between the end date requested by app and today's date
diff = (today_date - end_date).days
if diff > 0:
interval = (today_date - start_date).days
interval = self.get_interval(unit, interval)
# call api
response_data = self.mp.request(['events'], {
'event' : events, 'unit' : unit,
'interval' : interval, 'type': event_type
})
# bit of massaging is needed here to remove extra data
dates = []
if diff > 0: # chop values of
diff = self.get_interval(unit, diff)
#dates = response_data["data"]["series"][:-(diff-1)] # remove last n elements
dates = response_data["data"]["series"][-(diff):] # keep firt elements
events_data = []
for event_name in events:
event_data = response_data["data"]["values"][event_name]
for d in dates:
del event_data[d]
# convert date string to time in epoch
if viz == "line":
new_event_data = {}
for event_date, count in event_data.iteritems():
event_date = (calendar.timegm((datetime.strptime(event_date, '%Y-%m-%d')).timetuple())) * 1000
new_event_data[event_date] = count
events_data.append({"name": event_name, "data": sorted(new_event_data.iteritems(), key=lambda x: x[0])})
else:
events_data.append({"name": event_name, "data": sorted(event_data.iteritems(), key=lambda x: x[0])})
return events_data
#
# Given unit and interval in days, get new value of interval
#
def get_interval(self, unit, interval):
if unit == "hour":
interval = interval * 24
elif unit == "week":
interval = interval // 7
elif unit == "month":
interval = interval // 30
return interval
#
# get list of funnels
#
def get_funnel_list(self):
return self.mp.request(['funnels/list'], {})
#
# make api call to get funnel data
#
def get_funnel(self, funnel_id, start_date, end_date, group_by="", where_cond=""):
return self.mp.request(['funnels'], {
'funnel_id': funnel_id, 'from_date': start_date,
'to_date':end_date, 'on': group_by, 'where':where_cond})
#
# get funnel data
#
def get_funnel_data(self, funnel_id, start_date, end_date, viz="bar"):
# get funnel data
funnel_data = self.get_funnel(funnel_id, start_date, end_date)
# construct an empty funnel object
funnel = {'funnel_id': funnel_id, 'from_date': start_date, 'to_date':end_date,
'data': []}
if viz == "bar":
funnel["data"] = self.transform_funnel_data(funnel_data)
else:
funnel["data"] = self.transform_funnel_data(funnel_data, True)
return funnel
#
# get funnel breakdown data
#
def get_funnel_breakdown_data(self, funnel_id, start_date, end_date, viz="bar"):
# get funnel data
funnel_data = self.get_funnel(funnel_id, start_date, end_date)
# construct an empty funnel object
funnel = {'funnel_id': funnel_id, 'from_date': start_date, 'to_date':end_date,
'data': []}
if viz == "bar":
funnel["data"] = self.transform_funnel_breakdown_data(funnel_data)
else:
funnel["data"] = self.transform_funnel_breakdown_data(funnel_data, True)
return funnel
#
# get funnel data
#
def get_filtered_funnel_data(self, funnel_id, start_date, end_date, group_by="", where_cond="", viz="bar"):
# get funnel data
funnel_data = self.get_funnel(funnel_id, start_date, end_date, group_by, where_cond)
# construct an empty funnel object
funnel = {'funnel_id': funnel_id, 'from_date': start_date, 'to_date':end_date,
'data': []}
funnel["data"] = self.transform_filtered_funnel_data(funnel_data)
return funnel
#
# get funnel conversion rate
#
def get_funnel_conversion_rate_data(self, funnel_id, start_date, end_date, viz="bar"):
# construct an empty funnel object
funnel = {'funnel_id': funnel_id, 'from_date': start_date, 'to_date':end_date,
'data': []}
# get funnel data
funnel_data = self.get_funnel(funnel_id, start_date, end_date)
conversion_data = {}
for event_date, values in sorted(funnel_data["data"].iteritems(), key=lambda x: x[0]):
event_date = (calendar.timegm((datetime.strptime(event_date, '%Y-%m-%d')).timetuple())) * 1000
ana = values["analysis"]
if int(ana["starting_amount"]) != 0:
conversion_data[event_date] = round(int(ana["completion"])/int(ana["starting_amount"]), 3)
else:
conversion_data[event_date] = 0
funnel["data"] = [{"name": "Conversion Rate", "data": sorted(conversion_data.iteritems(), key=lambda x: x[0])}]
return funnel
#
# transform funnel data
#
def transform_funnel_data(self, funnel_data, table_view=False):
# reduce all steps related to funnels grouped by date in one list
funnel_step_count_by_day = [step for info in funnel_data["data"].values() for step in info["steps"]]
# now aggregate the list of dictionaries
funnel_steps = defaultdict(int)
for steps in funnel_step_count_by_day:
funnel_steps[steps['event']] += steps['count']
if not table_view:
data = defaultdict(list)
for category,value in sorted(funnel_steps.iteritems(), key=lambda x: x[1], reverse=True):
data["categories"].append(category)
data["series"].append(value)
return data
else: # transform data to table
data = []
row = OrderedDict()
row["Funnel"] = "Overall"
prev_step_val = None
for category,value in sorted(funnel_steps.iteritems(), key=lambda x: x[1], reverse=True):
if prev_step_val is None:
prev_step_val = value
row[category] = "%d (%.2f%%)" % (value, (100 * float(value)/float(prev_step_val)))
prev_step_val = value
data.append(row)
return data
#
# transform funnel breakdown data by step by date
#
def transform_funnel_breakdown_data(self, funnel_data, table_view=False):
#print funnel_data
#by_step_by_date = defaultdict(list)
by_step_by_date = OrderedDict()
categories = []
for event_date, values in sorted(funnel_data["data"].iteritems(), key=lambda x: x[0]):
event_date = datetime.strptime(event_date, '%Y-%m-%d').strftime('%b %d')
#(calendar.timegm((datetime.strptime(event_date, '%Y-%m-%d')).timetuple())) * 1000
categories.append(event_date)
for step in values["steps"]:
series_name = step["event"]
count = step["count"]
if series_name in by_step_by_date:
#by_step_by_date[series_name].append([event_date, count])
by_step_by_date[series_name].append(count)
else:
#by_step_by_date[series_name] = [[event_date, count]]
by_step_by_date[series_name] = [count]
data = {"categories": categories, "series": by_step_by_date}
if not table_view:
return data
else:
grid_data = []
for key, value in data["series"].iteritems():
l = zip(data["categories"], value) # zip date and event count together
l.insert(0, ("Event", key))
grid_row = OrderedDict(l)
grid_data.append(grid_row)
return grid_data
#events_data.append({"name": event_name, "data": sorted(event_data.iteritems(), key=lambda x: x[0])})
#
# transform fitler data
#
def transform_filtered_funnel_data(self, funnel_data):
# first bring all events for a property value under one list
by_prop = defaultdict(list)
for info in funnel_data["data"].values():
for prop, items in info.iteritems():
by_prop[prop].extend(items)
# now aggregate by property
for prop, items in by_prop.iteritems():
by_event = defaultdict(Counter)
for item in items:
key = item["event"]
vals = {"count": item["count"]}
by_event[key].update(vals)
by_prop[prop] = OrderedDict(sorted(by_event.items(), key=lambda x: x[1], reverse=True))
return by_prop
#
# get segment data
#
def get_segment_data(self, event_name, start_date, end_date, unit, event_type, group_by="", where_cond="", viz="bar"):
# get funnel data segment_data =
request_object = {'event' : event_name,
'from_date': start_date,
'to_date': end_date,
'unit': unit,
'type': event_type }
#print request_object
if group_by:
request_object["on"] = group_by
if where_cond:
request_object["where"] = where_cond
#print request_object
segment_data = self.mp.request(['segmentation'], request_object)
return self.transform_segment_data(event_name, segment_data)
#
# transform segment data
#
#
# transform segment data
#
def transform_segment_data(self, event_name, segment_data, table_view=False, viz="bar"):
# first bring all events for a property value under one list
#by_prop = defaultdict(list)
data = defaultdict(list)
for prop, items in segment_data["data"]["values"].iteritems():
data["categories"].append(prop)
data["series"].append(sum(items.values()))
return data
#
# get retention data
#
def get_retention_data(self, born_event, start_date, end_date, unit, interval_count, data_as='number', return_event=None, group_by="", where_cond="", viz="bar"):
# get funnel data segment_data =
request_object = {'born_event' : born_event,
'from_date': start_date,
'to_date': end_date,
'unit': unit,
'interval_count': interval_count}
#print request_object
if return_event:
request_object["event"] = return_event
if where_cond:
request_object["where"] = where_cond
#print request_object
retention_data = self.mp.request(['retention'], request_object)
return self.transform_retention_data(retention_data, unit, data_as)
#
# transform segment data
#
def transform_retention_data(self, retention_data, unit, data_as='number', viz="bar"):
# first bring all events for a property value under one list
table_data = []
for prop, items in retention_data.iteritems():
data = OrderedDict()
data[unit] = prop
data['users'] = items["first"]
for index, ele in enumerate(items["counts"]):
if data_as == 'number':
data["".join([unit, str(index)])] = ele
else:
data["".join([unit, str(index)])] = round(ele * 100.0 / items["first"], 1)
table_data.append(data)
return table_data
