mpri-webdam/histories/models.py

130 lines
4.0 KiB
Python

""" Models for the history. This history should be able to generate history
entries, which looks like human-based browsing, according to a dedicated user
interests, keywords...
"""
import random
from math import floor
from queue import Queue
from django.db import models
import profiles.models as profiles
from tor_runner import TorInstance
from crawl import crawl
from pinocchio.settings import HISTORY_MIN
class HistoryEntry(models.Model):
""" A history entry, aka a url, and a timestamp.
"""
search = models.URLField(help_text="The url to be searched")
timestamp = models.DateTimeField()
history = models.ForeignKey(
'History',
on_delete=models.CASCADE
)
def __str__(self):
""" Returns the string representation of a history entry.
"""
return "{} : {}".format(self.timestamp, self.search)
class History(models.Model):
""" A history for a user, containing some web connections (http, https).
Each history is timed, in a human-behaviour manner. """
start_ts = models.DateTimeField(
help_text='The starting timestamp of the history. Useful for cron-like '
'structure.'
)
played = models.BooleanField(default=False)
user = models.ForeignKey(
profiles.Profile,
on_delete=models.CASCADE
)
def return_history(self):
""" Returns the history, sorted by increasing timestamps
"""
history_set = self.history_set.order_by('timestamp')
history_set = [(item.search, item.timestamp.date()) for item in history_set]
return history_set
def __str__(self):
""" Returns the string representation of a history.
"""
history_set = self.history_set.order_by('timestamp')
header = "[History]:\n"
return header + "\n".join(history_set)
def play_histories(self):
""" Actually plays the history.
"""
self.played = True
runner = TorInstance(self.history)
self.save()
def generate_partial_history(user, t_start):
""" Generate the part of the history resulting from the crawl starting at
the given url.
"""
timestamp = t_start
result = []
basis = generate_first_url(user)
result.append((basis, timestamp))
timestamp += 5* random.weibullvariate(1, 1.5)
queue = Queue()
search_engine_query = profiles.SearchEngine.objects.all()
search_engine_list = [item.url for item in search_engine_query]
crawler = crawl.CrawlingThread(user, basis, search_engine_list, queue)
crawler.start()
crawler.join()
urls = queue.get()
for url in urls:
timestamp += 5* random.weibullvariate(1, 1.5)
result.append((url, timestamp))
return result
def generate_first_url(user):
""" Generate the first url of a partial history, based on the user
information. """
interest = random.choice(
[user.interests.keywords.all(), user.interests.places.all(),
user.interests.websites.all(), user.interests.events.all()
]
)
search_term = random.choice(interest)
url = search_term.generate_url(user)
return url
def generate_history(user, ts_start):
""" Generate a new history for the user `user`, starting from timestamp
`ts_start`.
A few heuristics are used in order to give the impression that the history
is actually played by a user.
"""
# let's define a new history object.
history = History(start_ts=ts_start, user=user)
length = HISTORY_MIN + floor(10 * random.weibullvariate(1, 1.5))
history_line = 0
while history_line < length:
ts_start += 5 * random.weibullvariate(1, 2.8)
history_list = generate_partial_history(user, ts_start)
ts_start = history_list[-1].timestamp + 5 * random.weibullvariate(1, 5)
for (url, timestamp) in history_list:
new_line = HistoryEntry(
search=url,
timestamp=timestamp,
history=history
)
new_line.save()