#! /usr/local/bin/python
#-*- coding: utf-8 -*-

__author__ = "Cedric Bonhomme"
__version__ = "$Revision: 0.1 $"
__date__ = "$Date: 2010/02/24 $"
__copyright__ = "Copyright (c) 2010 Cedric Bonhomme"
__license__ = "GPLv3"

import re
import pylab
import sqlite3
import hashlib

from datetime import datetime
from string import punctuation
from collections import defaultdict

from StringIO import StringIO

from oice.langdet import langdet
from oice.langdet import streams
from oice.langdet import languages

def detect_language(text):
    """
    Detect the language of a text.
    English, French or other (not detected).
    """
    text = text.strip()
    try:
        text_stream = streams.Stream(StringIO(text))
        lang = langdet.LanguageDetector.detect(text_stream)
    except:
        return 'other'
    if lang == languages.french:
        return 'french'.encode('utf-8')
    elif lang == languages.english:
        return 'english'.encode('utf-8')
    else:
        return 'other'

def remove_html_tags(data):
    """
    Remove HTML tags for the search.
    """
    p = re.compile(r'<[^<]*?/?>')
    return p.sub('', data)

def top_words(dic_articles, n=10):
    """
    Return the n most frequent words in a list.
    """
    words = {}
    articles_content = ""
    for rss_feed_id in dic_articles.keys():
        for article in dic_articles[rss_feed_id]:
            articles_content += remove_html_tags(article[4].encode('utf-8'))
    words_gen = (word.strip(punctuation).lower() \
                        for word in articles_content.split() \
                                if len(word) >= 5)
    words = defaultdict(int)
    for word in words_gen:
        words[word] += 1
    top_words = sorted(words.iteritems(),
                key=lambda(word, count): (-count, word))[:n]
    return top_words

def create_histogram(words, file_name="./var/histogram.png"):
    """
    Create a histogram.
    """
    length = 10
    ind = pylab.arange(length) # abscissa
    width = 0.35 # bars width

    w = [elem[0] for elem in words]
    count = [int(elem[1]) for elem in words]

    max_count = max(count)  # maximal weight

    p = pylab.bar(ind, count, width, color='r')

    pylab.ylabel("Count")
    pylab.title("Most frequent words")
    pylab.xticks(ind + (width / 2), range(1, len(w)+1))
    pylab.xlim(-width, len(ind))

    # changing the ordinate scale according to the max.
    if max_count <= 100:
        pylab.ylim(0, max_count + 5)
        pylab.yticks(pylab.arange(0, max_count + 5, 5))
    elif max_count <= 200:
        pylab.ylim(0, max_count + 10)
        pylab.yticks(pylab.arange(0, max_count + 10, 10))
    elif max_count <= 600:
        pylab.ylim(0, max_count + 25)
        pylab.yticks(pylab.arange(0, max_count + 25, 25))
    elif max_count <= 800:
        pylab.ylim(0, max_count + 50)
        pylab.yticks(pylab.arange(0, max_count + 50, 50))

    pylab.savefig(file_name, dpi = 80)
    pylab.close()

def compare(stringtime1, stringtime2):
    """
    Compare two dates in the format 'yyyy-mm-dd hh:mm:ss'.
    """
    date1, time1 = stringtime1.split(' ')
    date2, time2 = stringtime2.split(' ')

    year1, month1, day1 = date1.split('-')
    year2, month2, day2 = date2.split('-')

    hour1, minute1, second1 = time1.split(':')
    hour2, minute2, second2 = time2.split(':')

    datetime1 = datetime(year=int(year1), month=int(month1), day=int(day1), \
                        hour=int(hour1), minute=int(minute1), second=int(second1))

    datetime2 = datetime(year=int(year2), month=int(month2), day=int(day2), \
                        hour=int(hour2), minute=int(minute2), second=int(second2))

    if datetime1 < datetime2:
        return -1
    elif datetime1 > datetime2:
        return 1
    return 0

def load_feed():
    """
    Load feeds and articles in a dictionary.
    """
    list_of_feeds = None
    list_of_articles = None
    try:
        conn = sqlite3.connect("./var/feed.db", isolation_level = None)
        c = conn.cursor()
        list_of_feeds = c.execute("SELECT * FROM feeds").fetchall()
    except:
        pass

    # articles[feed_id] = (article_id, article_date, article_title,
    #               article_link, article_description, article_readed,
    #               article_language)
    # feeds[feed_id] = (nb_article, nb_article_unreaded, feed_image,
    #               feed_title, feed_link, feed_site_link)
    articles, feeds = {}, {}
    if list_of_feeds is not None:
        for feed in list_of_feeds:
            list_of_articles = c.execute(\
                    "SELECT * FROM articles WHERE feed_link='" + \
                    feed[2] + "'").fetchall()

            if list_of_articles is not None:
                for article in list_of_articles:
                    sha1_hash = hashlib.sha1()
                    sha1_hash.update(article[5].encode('utf-8'))
                    feed_id = sha1_hash.hexdigest()
                    sha1_hash.update(article[2].encode('utf-8'))
                    article_id = sha1_hash.hexdigest()

                    if article[3] != "":
                        language = detect_language(remove_html_tags(article[3][:80]).encode('utf-8') + \
                                            remove_html_tags(article[1]).encode('utf-8'))
                    else:
                        language = detect_language(remove_html_tags(article[1]).encode('utf-8'))

                    article_list = [article_id, article[0], article[1], \
                        article[2], article[3], article[4], language]

                    if feed_id not in articles:
                        articles[feed_id] = [article_list]
                    else:
                        articles[feed_id].append(article_list)


                # sort articles by date for each feeds
                for rss_feed_id in articles.keys():
                    articles[rss_feed_id].sort(lambda x,y: compare(y[1], x[1]))

                feeds[feed_id] = (len(articles[feed_id]), \
                                len([article for article in articles[feed_id] \
                                    if article[5]=="0"]), \
                                feed[3], feed[0], feed[2], feed[1] \
                                )
        c.close()

        return (articles, feeds)
    return (articles, feeds)