osu!lazer file recovery utility

save_me.py

'''
Tool for the very common situation where you mess with the osu lazer database
files (the ones you're told not to touch) without making a backup and suffer
the deserved consequences.

The interesting part here is that the files themselves are still present on
disk, but the lazer client is unable to perceive them (the existence of all
objects is tracked through client.realm). So the strategy here is to look for
all files after a certain cutoff date, pattern match for the desired file type,
then move them to one location, where they can be re-imported directly through
the client.

Requies osrparse, which is available from pip.
'''

import datetime
import os
import re
import shutil
import urllib.request

from osrparse import Replay

OSU_BLOB_PATH = "~/.local/share/osu/files"
CUTOFF_DATE = datetime.datetime(2025, 1, 16) # change as needed

def execute_walk(directory, callback):
    directory = os.path.expanduser(directory)
    for root, _, files in os.walk(directory):
        for file in files:
            file_path = os.path.join(root, file)
            callback(file_path)

beatmapsets = dict()

def find_beatmapsets(file_path):
    global beatmapsets
    mod_time = os.path.getmtime(file_path)
    mod_datetime = datetime.datetime.fromtimestamp(mod_time)
    if mod_datetime <= CUTOFF_DATE:
        return
    lines = []
    try:
        with open(file_path, 'r', encoding='utf-8') as f:
            lines = [s.strip() for s in f.readlines()]
    except UnicodeDecodeError:
        return
    if not lines[0].startswith('osu file format'):
        return
    # TODO: get metadata
    metadata_line = lines.index('[Metadata]')
    field_pattern = r'^(.*?):(.*)'
    title = None
    artist = None
    mapset_id = None
    for i in range(metadata_line+1, len(lines)):
        if lines[i] == '':
            continue
        m = re.match(field_pattern, lines[i])
        if m is None:
            break
        match m.group(1):
            case 'Title':
                title = m.group(2)
            case 'Artist':
                artist = m.group(2)
            case 'BeatmapSetID':
                mapset_id = int(m.group(2))
    if mapset_id is not None:
        beatmapsets[mapset_id] = (title, artist)
    # readable_time = mod_datetime.strftime('%Y-%m-%d %H:%M:%S')
    # print(f"Found {title} [{version}]: Last modified on {readable_time}")

def recover_beatmapsets():
    execute_walk(OSU_BLOB_PATH, find_beatmapsets)
    if not os.path.exists('maps/'):
        os.mkdir('maps')
    for id, (title, artist) in beatmapsets.items():
        print(f'Fetching {id}: {title} ({artist})')
        b_url = f'https://catboy.best/d/{id}'
        b_path = os.path.join('maps', f'{id} {artist} - {title}.osz')
        if not os.path.exists(b_path):
            urllib.request.urlretrieve(b_url, b_path)
    print(len(beatmapsets))

replays = set()

replay_count = 0
def find_replays(file_path):
    global replays, replay_count
    mod_time = os.path.getmtime(file_path)
    mod_datetime = datetime.datetime.fromtimestamp(mod_time)
    if mod_datetime <= CUTOFF_DATE:
        return
    try:
        replay = Replay.from_path(file_path)
    except:
        return
    replay_count += 1
    replays.add(file_path)

def recover_replays():
    execute_walk(OSU_BLOB_PATH, find_replays)
    if not os.path.exists('replays/'):
        os.mkdir('replays')
    for replay_path in replays:
        basename = os.path.basename(replay_path) + '.osr'
        new_path = os.path.join('replays', basename)
        # replace
        shutil.copy(replay_path, new_path)
    print(replay_count)

# comment/uncomment depending on what you need
recover_beatmapsets()
recover_replays()