mpri-webdam/histories/models.py

""" 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 django.db import models
import profiles.models as profiles
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')
        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_history(self):
        """ Actually plays the history.
        """
        self.played = True
        self.save()


def generate_partial_history(user, t_start, url, history):
    """ Generate the part of the history resulting from the crawl starting at
    the given url.
    """
    #crawler = crawl.CrawlingThread()
    return []

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(interset)
    if isinstance(search_term, profiles.Website):
        if user.uses_url:
            url = search_term.url
        elif random.random() <= 0.1:
            url = random.choice(search_term.notable_pages).url
        elif random.random() >= 0.3:
            search_term_text = search_term.name +  " " + random.choice(search_term.keywords)
            url = user.search_engine.search_url(search_term_text)
    #if
    #elif isinstance(search_term, profiles.Website):
    #    url = user.search_engine.search_url(search_term)



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 += random.uniform(1, 10)
        history = generate_partial_history(user, ts_start, url)
        ts_start = history[-1].timestamp + 5 * weilbullvariate(1, 5)
        for (url, ts) in history:
            new_line = HistoryEntry(
                search=url,
                timestamp=ts,
                history=history
            )