Gitea to RSS scraper

mkfeed_gitea.py

#!/usr/bin/env python3

import html
from datetime import datetime, timedelta, timezone
import re
import json
import sys
import requests
from bs4 import BeautifulSoup

VERSION = '2022.07.25.gitea'

# date funcs
def rfc822date(date):
	return datetime.strftime(date, '%a, %d %b %Y %H:%M:%S %z').strip()

# utility funcs
def sanitize(tag):
	pass

def escape(string):
	return html.escape(string)

# simple function
def build_tree(tag, content, attrs={}):
	encoded_content = ""
	encoded_attrs = ""
	
	if isinstance(content, str):
		encoded_content = escape(content)
	
	elif isinstance(content, dict):
		for k, e in content.items():
			encoded_content += build_tree(k, e)
	
	elif isinstance(content, list):
		for k in content:
			if isinstance(k, dict):
				encoded_content += build_tree(
					k.get("tag"),
					k.get("content"),
					attrs=k.get("attrs", {})
				)

	for k, e in attrs.items():
		encoded_attrs = " %s=\"%s\"" % (k, e)
	
	return "\n<%s%s>%s</%s>" % (tag, encoded_attrs, encoded_content, tag)

def parse_page(url):
	itemsList = []
	
	with requests.get(url) as rq:
		rr = BeautifulSoup(rq.text, 'html.parser')
		
		html_commits = rr.select('.commit-list > tr')
		for hcommit in html_commits:
			author = hcommit.select('.author a')[0].text
			title = hcommit.select('.commit-summary')[0].text
			link = hcommit.select('.default-link')[0].attrs['href']
			baseUrl = re.findall(r'^.+?[^\/:](?=[?\/]|$)', url)[0]
			date = hcommit.select('.time-since')[0].attrs['title']
			sha = hcommit.select('.shortsha')[0].text
			
			itemsList.append({
				"tag": "item",
				"content": {
					"title": title,
					"link": baseUrl + link,
					"author": author,
					"description": sha,
					"pubDate": date
				}
			})
	return itemsList

def make_rss(url):
	# init channel
	channelData = [
				{"tag":"title",       "content":"Gitea"},
				{"tag":"link",        "content":url},
				{"tag":"description", "content":"Latest commits from a gitea repo"},
				{"tag":"language",    "content":"en-US"},
				{"tag":"generator", "content":"mkfeed.py %s" % VERSION},
				{"tag":"lastBuildDate", "content":rfc822date(datetime.now(timezone.utc))},
	]
	# doctype
	print("<?xml version=\"1.0\" encoding=\"utf-8\"?>", end="")

	# make listing
	itemList = []
	itemList += parse_page(url)
	
	channelData += itemList
	
	# generate RSS feed
	print(
		build_tree(
		"rss", {
			"channel": channelData
		},
		attrs={"version": "2.0"})
	)

if __name__ == '__main__':
	try:
		feed = make_rss(sys.argv[1])
	except:
		print('./gitea "https://<Gitea instance>/<Author>/<Repo>/commits/branch/<Branch>"')