zulip/analytics/management/commands/analyze_mit.py

88 lines
3.7 KiB
Python

from __future__ import absolute_import
from __future__ import print_function
from typing import Any
from optparse import make_option
from django.core.management.base import BaseCommand, CommandParser
from zerver.models import Recipient, Message
from zerver.lib.timestamp import timestamp_to_datetime
import datetime
import time
import logging
def compute_stats(log_level):
# type: (int) -> None
logger = logging.getLogger()
logger.setLevel(log_level)
one_week_ago = timestamp_to_datetime(time.time()) - datetime.timedelta(weeks=1)
mit_query = Message.objects.filter(sender__realm__string_id="mit",
recipient__type=Recipient.STREAM,
pub_date__gt=one_week_ago)
for bot_sender_start in ["imap.", "rcmd.", "sys."]:
mit_query = mit_query.exclude(sender__email__startswith=(bot_sender_start))
# Filtering for "/" covers tabbott/extra@ and all the daemon/foo bots.
mit_query = mit_query.exclude(sender__email__contains=("/"))
mit_query = mit_query.exclude(sender__email__contains=("aim.com"))
mit_query = mit_query.exclude(
sender__email__in=["rss@mit.edu", "bash@mit.edu", "apache@mit.edu",
"bitcoin@mit.edu", "lp@mit.edu", "clocks@mit.edu",
"root@mit.edu", "nagios@mit.edu",
"www-data|local-realm@mit.edu"])
user_counts = {} # type: Dict[str, Dict[str, int]]
for m in mit_query.select_related("sending_client", "sender"):
email = m.sender.email
user_counts.setdefault(email, {})
user_counts[email].setdefault(m.sending_client.name, 0)
user_counts[email][m.sending_client.name] += 1
total_counts = {} # type: Dict[str, int]
total_user_counts = {} # type: Dict[str, int]
for email, counts in user_counts.items():
total_user_counts.setdefault(email, 0)
for client_name, count in counts.items():
total_counts.setdefault(client_name, 0)
total_counts[client_name] += count
total_user_counts[email] += count
logging.debug("%40s | %10s | %s" % ("User", "Messages", "Percentage Zulip"))
top_percents = {} # type: Dict[int, float]
for size in [10, 25, 50, 100, 200, len(total_user_counts.keys())]:
top_percents[size] = 0.0
for i, email in enumerate(sorted(total_user_counts.keys(),
key=lambda x: -total_user_counts[x])):
percent_zulip = round(100 - (user_counts[email].get("zephyr_mirror", 0)) * 100. /
total_user_counts[email], 1)
for size in top_percents.keys():
top_percents.setdefault(size, 0)
if i < size:
top_percents[size] += (percent_zulip * 1.0 / size)
logging.debug("%40s | %10s | %s%%" % (email, total_user_counts[email],
percent_zulip))
logging.info("")
for size in sorted(top_percents.keys()):
logging.info("Top %6s | %s%%" % (size, round(top_percents[size], 1)))
grand_total = sum(total_counts.values())
print(grand_total)
logging.info("%15s | %s" % ("Client", "Percentage"))
for client in total_counts.keys():
logging.info("%15s | %s%%" % (client, round(100. * total_counts[client] / grand_total, 1)))
class Command(BaseCommand):
help = "Compute statistics on MIT Zephyr usage."
def add_arguments(self, parser):
# type: (CommandParser) -> None
parser.add_argument('--verbose', default=False, action='store_true')
def handle(self, *args, **options):
# type: (*Any, **Any) -> None
level = logging.INFO
if options["verbose"]:
level = logging.DEBUG
compute_stats(level)