0ct0pu5/mwmbl
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Merge pull request #96 from mwmbl/unique-urls-in-queue

Browse source 
Unique URLs in queue
This commit is contained in:
Daoud Clarke 2023-02-25 10:35:32 +00:00 • committed by GitHub
commit a9e2b48840
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 13 additions and 14 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

27
mwmbl/url_queue.py
View file

@ -5,6 +5,7 @@ from datetime import datetime, timedelta
from logging import getLogger
from multiprocessing import Queue
from queue import Empty
from random import Random
from typing import KeysView, Union
from mwmbl.crawler.urls import BATCH_SIZE, URLDatabase, URLStatus, FoundURL, REASSIGN_MIN_HOURS
@ -25,10 +26,8 @@ MAX_URLS_PER_TOP_DOMAIN = 100
MAX_URLS_PER_OTHER_DOMAIN = 5
MAX_OTHER_DOMAINS = 10000
@dataclass
class URLScore:
url: str
score: float
random = Random(1)
class URLQueue:
@ -39,8 +38,8 @@ class URLQueue:
"""
self._new_item_queue = new_item_queue
self._queued_batches = queued_batches
self._other_urls = defaultdict(list)
self._top_urls = defaultdict(list)
self._other_urls = defaultdict(dict)
self._top_urls = defaultdict(dict)
self._min_top_domains = min_top_domains
def initialize(self):
@ -88,7 +87,7 @@ class URLQueue:
except ValueError:
continue
url_store = self._top_urls if domain in TOP_DOMAINS else self._other_urls
url_store[domain].append(URLScore(found_url.url, found_url.score))
url_store[domain][found_url.url] = found_url.score
logger.info(f"URL store updated: {len(self._top_urls)} top domains, {len(self._other_urls)} other domains")
@ -110,6 +109,7 @@ class URLQueue:
self._queue_urls(urls)
def _queue_urls(self, valid_urls: list[str]):
random.shuffle(valid_urls)
for url_batch in batch(valid_urls, BATCH_SIZE):
self._queued_batches.put(url_batch, block=False)
@ -122,18 +122,17 @@ class URLQueue:
return len(self._top_urls)
def _sort_and_limit_urls(domain_urls: dict[str, list[str]], max_urls: int):
def _sort_and_limit_urls(domain_urls: dict[str, dict[str, float]], max_urls: int):
for domain, urls in domain_urls.items():
domain_urls[domain] = sorted(urls, key=lambda url_score: url_score.score, reverse=True)[:max_urls]
domain_urls[domain] = dict(sorted(urls.items(), key=lambda url_score: url_score[1], reverse=True)[:max_urls])
def _add_urls(domains: Union[set[str], KeysView], domain_urls: dict[str, list[URLScore]], urls: list[str], max_urls: int):
def _add_urls(domains: Union[set[str], KeysView], domain_urls: dict[str, dict[str, float]], urls: list[str], max_urls: int):
for domain in list(domains & domain_urls.keys()):
new_urls = domain_urls[domain][:max_urls]
urls += [url_score.url for url_score in new_urls]
new_domain_urls = domain_urls[domain][max_urls:]
urls += list(domain_urls[domain].keys())[:max_urls]
new_domain_urls = list(domain_urls[domain].items())[max_urls:]
if len(new_domain_urls) > 0:
domain_urls[domain] = new_domain_urls
domain_urls[domain] = dict(new_domain_urls)
else:
del domain_urls[domain]