mirror of
https://github.com/ArchiveBox/ArchiveBox
synced 2024-11-10 06:34:16 +00:00
54 lines
1.7 KiB
Python
54 lines
1.7 KiB
Python
__package__ = 'archivebox.parsers'
|
|
|
|
|
|
from typing import IO, Iterable
|
|
from datetime import datetime
|
|
|
|
from ..index.schema import Link
|
|
from ..util import (
|
|
htmldecode,
|
|
enforce_types,
|
|
str_between,
|
|
)
|
|
|
|
@enforce_types
|
|
def parse_generic_rss_export(rss_file: IO[str], **_kwargs) -> Iterable[Link]:
|
|
"""Parse RSS XML-format files into links"""
|
|
|
|
rss_file.seek(0)
|
|
items = rss_file.read().split('<item>')
|
|
items = items[1:] if items else []
|
|
for item in items:
|
|
# example item:
|
|
# <item>
|
|
# <title><![CDATA[How JavaScript works: inside the V8 engine]]></title>
|
|
# <category>Unread</category>
|
|
# <link>https://blog.sessionstack.com/how-javascript-works-inside</link>
|
|
# <guid>https://blog.sessionstack.com/how-javascript-works-inside</guid>
|
|
# <pubDate>Mon, 21 Aug 2017 14:21:58 -0500</pubDate>
|
|
# </item>
|
|
|
|
trailing_removed = item.split('</item>', 1)[0]
|
|
leading_removed = trailing_removed.split('<item>', 1)[-1].strip()
|
|
rows = leading_removed.split('\n')
|
|
|
|
def get_row(key):
|
|
return [r for r in rows if r.strip().startswith('<{}>'.format(key))][0]
|
|
|
|
url = str_between(get_row('link'), '<link>', '</link>')
|
|
ts_str = str_between(get_row('pubDate'), '<pubDate>', '</pubDate>')
|
|
time = datetime.strptime(ts_str, "%a, %d %b %Y %H:%M:%S %z")
|
|
title = str_between(get_row('title'), '<![CDATA[', ']]').strip()
|
|
|
|
yield Link(
|
|
url=htmldecode(url),
|
|
timestamp=str(time.timestamp()),
|
|
title=htmldecode(title) or None,
|
|
tags=None,
|
|
sources=[rss_file.name],
|
|
)
|
|
|
|
|
|
KEY = 'rss'
|
|
NAME = 'Generic RSS'
|
|
PARSER = parse_generic_rss_export
|