libpurple-to-markdown/libpurple_to_markdown/libpurple.py

"""Backend for Pidgin/LibPurple.

[Pidgin](https://pidgin.im/) is a multi-protocol instant messaging app. It
stores logs as either plain text files, or as HTML files (default).

This backend parses the HTML files, focusing on the IRC protocol-style logs.

**This backend is not actively maintained.**
"""

import datetime
import logging
from pathlib import Path

import bs4

from .data import Message

logger = logging.getLogger(__name__)


def parse_timestamp(c) -> datetime.time:
    timestamp_obj = c
    if c.font is not None:
        c = c.font
    m = re.match(r'\((\d+):(\d+):(\d+)\)', timestamp_obj.get_text())
    return datetime.time(int(m.group(1)), int(m.group(2)), int(m.group(3)))


def parse_messages_in_chat_file(path: Path, chat_id: str) -> list[Message]:
    logger.info('Parsing %s', path)
    chat_start = datetime.datetime.fromisoformat(
        path.stem.removesuffix('CEST').removesuffix('CET'),
    )

    with open(path) as f:
        soup = bs4.BeautifulSoup(f, 'lxml')

    if len(soup.contents) == 0:
        logger.warning('File is empty?')
        return []

    messages = []

    cur_sent_at: datetime.datetime | None = None
    cur_sender: str | None = None
    cur_text: str = ''

    if soup.body.p:
        loglines = soup.body.p.children
    else:
        loglines = soup.body.children

    for c in loglines:
        if c.name in {'font', 'span'} and cur_sent_at is None:
            # Get timestamp
            cur_sent_at = datetime_sent(chat_start, parse_timestamp(c))

            # Get sender
            if c.b:
                assert cur_sender is None
                cur_sender = (
                    c.b.get_text()
                    .strip()
                    .removesuffix(':')
                    .removeprefix('***')
                    .removesuffix('[m]')
                )

        elif c.name in {None, 'span', 'font'}:
            cur_text += c.get_text()

        elif c.name == 'a':
            cur_text += '<' + c['href'] + '>'

        elif c.name == 'br':
            if cur_sender:
                messages.append(
                    Message(cur_sent_at, cur_sender, cur_text.strip(), chat_id),
                )
            cur_sent_at = None
            cur_sender = None
            cur_text = ''

        elif c.name == 'b':
            # Indicates system message. Ignore
            pass

        elif c.name in {'h1', 'h3'}:
            pass  # Ignore log header

        else:
            assert False, c

    return messages


def parse_messages_in_chat_folder(chat_folder_path: Path) -> list[Message]:
    messages = []
    server = args.purple_folder.parent.name
    receipient = args.purple_folder.name
    chat_id = f'{server} - {receipient}'
    for file_path in sorted(chat_folder_path.iterdir()):
        messages.extend(parse_messages_in_chat_file(file_path, chat_id))

    messages.sort()
    return messages