alexis2/parser.py

#!/usr/bin/python3
import threading
from queue import Queue

from db import db
from source import Source
from linklist import LinkList
from news import News
from paid_parser import PaidParser
from time import strftime
import sys


def parseSource(s):
    global q
    q = Queue()
    num_worker_threads = 10
    sel = None
    rss = False
    if "selector" in s:
        sel = s["selector"]
    if "rss" in s:
        rss = bool(s["rss"])
    l = LinkList(s["link"], sel, rss=rss)

    kw = d.get_keywords(s["category"])
    new_articles = 0
#    print(kw)
    if l is not None:
        for a in l.links:
            if not d.article_exists(a.lower()):
                q.put(a)
        print("[%s][Source] %s: %d/%d new/total articles" %
              (strftime("%H:%M:%S"), s["name"], q.qsize(), len(l.links)))
    new_articles = q.qsize()

    if not q.empty():
        num_worker_threads = min(10, q.qsize())
        threads = []
        for i in range(num_worker_threads):
            t = threading.Thread(target=news_worker, args=[s, kw])
            t.daemon = True
            t.start()
            threads.append(t)

#        print("Waiting for %d queue elements" % q.qsize())
        q.join()  # block until all tasks are done
#        print("Finished waiting for the queue")

        for i in range(num_worker_threads):
            q.put(None)
        for t in threads:
            t.join(10)
#    print("[Source] %s => Finished parsing" %s["name"])
    return new_articles


def news_worker(source, kw):
    source_id = source["_id"]
    source_name = source["name"]
    source_category = source["category"]
    paid = False
    if "paid" in source and source["paid"]:
        paid = True

    tname = threading.current_thread().name
    while True:
        try:
            item = q.get(timeout=5)
        except:
            #print("Failed to get item")
            q.task_done()
            break

        if item is None:
            q.task_done()
            break

        #print("[Thread %s] %s" % (tname,item))
        try:
            html = None
            if paid:
                p = PaidParser(item)
                html = p.html

            n = News(item, source_id, source_name,
                     source_category, kw, html=html)
            #print("[Thread %s] Finished" % tname)
            d.insert_article(n.get())
        except Exception as e:
            print("#### EXCEPTION ########")
            print(e)
            print(item)
            print("whut")
            print("#### END EXCEPTION ####")
        finally:
            q.task_done()

q = Queue()
d = db()
d.purge_error()
n = 0
for s in d.sources():
    n = n + parseSource(s)
    sys.stdout.flush()

print("[%s] %d new articles" % (strftime("%H:%M:%S"), n))
print("###############################################################################")