Bug 1333806 - Investigate pings with missing activePlugins sections

Bug 1333806 - Investigate pings with missing activePlugins sections.ipynb

Bug 1333806 - Investigate pings with missing activePlugins sections.py

# coding: utf-8

# Bug 1333806 - Investigate pings with missing activePlugins sections

# In[1]:

import ujson as json
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import plotly.plotly as py

from plotly.graph_objs import *
from moztelemetry import get_pings_properties, get_one_ping_per_client
from moztelemetry.dataset import Dataset

get_ipython().magic(u'matplotlib inline')


# In[5]:

pings = Dataset.from_source("telemetry")     .where(appName='Firefox')     .where(docType='main')     .where(submissionDate= lambda x: "20170101" <= x <= "20170131")     .where(appUpdateChannel="nightly")     .where(sourceVersion="4")     .records(sc, sample=0.1)


# ... and extract only the attributes we need from the Telemetry submissions:

# In[6]:

subset = get_pings_properties(pings, ["clientId",
                                      "environment/system/os/name",
                                      "environment/addons"])


# In[24]:

ping_count = subset.count()


# Let's try to understand what's going on with the activePlugins section coming from the client pings.

# In[13]:

def to_os_plugins(x):
    os_name = x.get("environment/system/os/name", "Unknown")
    env_addons = x.get("environment/addons", None)

    if (env_addons is None):
        return ((os_name, "No env_addons section"), 1)

    if (not isinstance(env_addons, dict)):
        return ((os_name, "Env_addons is not a dict"), 1)
    
    active_plugins = env_addons["activePlugins"]
    if (active_plugins is None):
        return ((os_name, "No activePlugins"), 1)
    
    if (isinstance(active_plugins, dict)):
        return ((os_name, "activePlugins is a dict, not a list"), 1)
    
    if (not isinstance(active_plugins, list)):
        return ((os_name, "activePlugins is not a list either!"), 1)
    
    if (len(active_plugins) < 1):
        return ((os_name, "empty activePlugins list"), 1)

    return ((os_name, "activePlugins should be fine"), 1)

plugins_errors = subset.map(to_os_plugins)


# In[25]:

error_counts = plugins_errors.countByKey()
error_counts


# It looks like we're receiving *dicts* instead of *lists* for activePlugins in some pings. Here's the breakdown, per platform:
# 
# * Darwin - 4769 over 14499 (ratio 0.32)
# * Linux - 13910 over 27714 (ratio 0.50)
# * Windows - 90009 over 340591 (ratio 0.26)
# 
# Let's also check if the dicts contain any key.

# In[21]:

def filter_valid_dicts(p):
    env_addons = p.get("environment/addons", None)
    if (env_addons is None):
        return False
    
    active_plugins = env_addons["activePlugins"]
    if (active_plugins is None):
        return False
    
    if (isinstance(active_plugins, list)):
        return False
    
    return True

plugins_dicts = subset.filter(filter_valid_dicts)


# In[22]:

dict_keys_counts = plugins_dicts.map(lambda p: len(p.get("environment/addons", {}).get("activePlugins").keys()))


# In[23]:

dict_keys_counts.countByValue()


# Since we're here, also check if we're receiving pings with empty *lists* as activePlugins (which is what we expect).

# In[29]:

def filter_empty_lists(p):
    env_addons = p.get("environment/addons", None)
    if (env_addons is None):
        return False
    
    active_plugins = env_addons["activePlugins"]
    if (active_plugins is None):
        return False
    
    if (isinstance(active_plugins, dict)):
        return False
    
    if (len(active_plugins) > 0):
        return False
    
    return True

plugin_lists = subset.filter(filter_empty_lists)


# In[30]:

plugin_lists.count()


# Oh, that's weird. It looks like every empty *activePlugin* section contains {} rather than [].

# In[ ]: