Pull and sort of bunch of session events from Keen IO

gistfile1.rb

require 'rubygems'
require 'net/http' # This is probably not needed, did not validate which things are truly required...
require 'net/https'
require 'keen'
require 'json'
require 'date'
require 'time'
require 'set'

NUM_LATEST_SESSIONS = 1000

# First, we enter the Keen IO project info for the data we'll be reading.
keen_project = Keen::Client.new(
    :project_id => <project_id>,
    :read_key => <read_key>
)

# Extract from Keen IO a bunch of sessions IDs that we would like study
completed_sessions = keen_project.extraction('session_end',
    :target_property => 'session.id',
    :latest => NUM_LATEST_SESSIONS,
)

puts completed_sessions.length  # How many sessions are we looking at?

# Each flow starts with a 'session_start', but beyond that, the following events
# can occur in varying orders (aka flows).
completed_sessions.each do |data|

    puts session_id = data['session']['id']

    # Create a bucket to store all the session events for each session
    # Start by filling it with screen_view events
    session_events = keen_project.extraction('screen_views',
        :filters => [{
            :property_name => 'session.id',
            :operator => 'eq',
            :property_value => session_id,
        }],
        # This screenviews collection happens to have a label for each screen called "session.current.type"
        :property_names => (['session.current.type', 'keen.timestamp']).to_json 
    )

    # Now go get other types of events, such as payment events
    payment = keen_project.extraction('payment',
        :filters => [{
            :property_name => 'session.id',
            :operator => 'eq',
            :property_value => session_id,
        }],
        :property_names => (['keen.timestamp']).to_json
    )

    # Add the payment events to the array of events. 
        payment.each do |m|
            # To match the existing events we have in session_events, we're adding new "session.current.type" events with the label "money"
            # We should probably re-write this script with something more simple and generic like "label" for each session event
            m['session'] = {'current' => {'type' => 'money'}}
        session_events << m
        end
    end

    # Add plays type events to the set of session events
    plays = keen_project.extraction('play',
        :filters => [{
            :property_name => 'session.id',
            :operator => 'eq',
            :property_value => session_id,
        }],
        :property_names => (['event.play_type', 'keen.timestamp']).to_json
    )

    unless plays.empty?
        plays.each do |p|
            # Add the play events to our array using a label based on the "play_type" 
            p['session'] = {'current' => {'type' => p['event']['play_type']}}
        session_events << p
        end
    end

    # Now let's convert the timestamps we get from Keen IO into Epoch integers.
    session_events.each do |e|
        e['keen']['timestamp_epoch'] = DateTime.parse(e['keen']['timestamp']).to_time.to_i
    end

    # Sort the events by their timestamps so that we can see them in the order they happened
    sorted_events = session_events.sort_by { |hsh| hsh['keen']['timestamp_epoch']}
    
    # Do something with the sorted events!
    puts sorted_events
    
end