Parse HTML

[paperdoorknob] / paperdoorknob.py

diff --git a/paperdoorknob.py b/paperdoorknob.py
index c1117c2680d3f2eb6788542116602606423fb6df..bb8bdd1ae38a660b3a6f90aa3d5aa317505beee3 100644 (file)
--- a/paperdoorknob.py
+++ b/paperdoorknob.py
@@ -6,12 +6,20 @@
 
 
 from argparse import ArgumentParser
+import os.path
+from bs4 import BeautifulSoup
 import requests
 import requests_cache
+from xdg_base_dirs import xdg_cache_home
 
 
 def command_line_parser() -> ArgumentParser:
     parser = ArgumentParser(prog='paperdoorknob', description='Print glowfic')
+    parser.add_argument(
+        '--cache_path',
+        metavar='PATH',
+        help='Where to keep the http cache (instead of %(default)s)',
+        default=os.path.join(xdg_cache_home(), "paperdoorknob"))
     parser.add_argument(
         '--timeout',
         help='How long to wait for HTTP requests, in seconds',
@@ -20,14 +28,15 @@ def command_line_parser() -> ArgumentParser:
     return parser
 
 
-def fetch(url: str, session: requests.Session, timeout: int) -> None:
+def fetch(url: str, session: requests.Session, timeout: int) -> BeautifulSoup:
     with session.get(url, timeout=timeout) as r:
         r.raise_for_status()
+        return BeautifulSoup(r.text, 'html.parser')
 
 
 def main() -> None:
     args = command_line_parser().parse_args()
-    with requests_cache.CachedSession() as session:
+    with requests_cache.CachedSession(args.cache_path, cache_control=True) as session:
         fetch(args.url, session, args.timeout)