PRI_2020-FE/backend/xmlParser.py

#!/usr/bin/env python3

import argparse
from bs4 import BeautifulSoup
from postmarkup import render_bbcode
import html
import json


# arguments
parser = argparse.ArgumentParser(description='Process some xml files.')
parser.add_argument('filename', help='xml forum file')
args = parser.parse_args()

# make a soup
with open(args.filename) as forum:
  soup = BeautifulSoup(forum, "xml")

# put json together
out = {}
out['id'] = soup.forum.get('id')
out['name'] = soup.forum.find('name').text
out['discussions'] = []
for d in soup.forum.find_all('discussion'):
  posts = []
  for p in d.find_all('post'):
    post_soup = BeautifulSoup(html.unescape(str(p.message)), "lxml")
    paragraphs = [render_bbcode(x.text) for x in post_soup.find_all('p')]
    posts.append({
      'id': p.get('id'),
      'parent': p.find('parent').text,
      'author': p.userid.text,
      'message': [x for x in paragraphs if x]
    })
  out['discussions'].append({
    'id':         d.get('id'),
    'title':      d.find('name').text,
    'first_post': d.firstpost.text,
    'posts': posts
  })

with open('parsed.json', 'w') as outfile:
  json.dump(out, outfile, ensure_ascii=False, indent=2)