/touhou-replay-decoder

touhou project replay file decoder.

Primary LanguagePython

Touhou project replay file decoder

The project is alomst a Python implementation for threp by Fluorohydride. But it include more flexiable API and bug fix (and more slow! Thanks Python).

It support th10(東方風神録 ～ Mountain of Faith) and th11(東方地霊殿 ～ Subterranean Animism), while threp support more game, my own statistical only need this two so I have not port others.

API

API.py

cut_scan(origin_dir, target_dir) will transform all *.rpy files included in origin_dir to python pickled file into target_dir.

Every pickled file represents a dictionary, you can use pickle.load(pickle_file_path) to load it:

with open(your_pickle_path, 'rb') as f:
    info = pickle.load(f)

Attributions

  • info['stages']: How many stages replay include.
  • info['character'],info['ctype'],info['rank'],info['clear']: In a th10 replay, (0,0,0,8) means that it's a Reimu A play in a easy whole(so not the extra) game. For detail you can check th10.th10type and th11.th11type.
  • info['code']: Indicate what the replay come from, 0x72303174 for th10, and 0x72313174 for th11.
  • info['meta']: The binary data of the header of the decoded replay file.
  • info['stages']: is a dictionary, info['stages'][0] is a another dictionary correspond to first stage included in the replay.
stage = info['stages'][0]
  • stage['score']: the score got in this stage.
  • stage['frame']: the number of frames in the stage.
  • stage['llength']: the replay section binary data length in the stage.
  • stage['index']: indicate the index of start and end of the three parts(header,replay,tail) in this stage.
  • stage['bin']: Dictionary. The binary data part of the stage.
replay = stage['bin']['replay']
replay[:100]
#bytearray(b'\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00@\x00@\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00@\x00\x00\x00\x00\x00P\x00\x10\x00\x00\x00P\x00\x00\x00\x00\x00P\x00\x00\x00\x00\x00P\x00\x00\x00')

The replay is a sequence consist of binary blocks(per frame), every block coded into 6 bytes.

from utils import replay_to_binary_seq
seq = replay_to_binary_seq(replay)
seq[10:20]
'''
['000000000000000000000000010000000000000000000000',
 '000000000000000000000000010000000000000000000000',
 '000000000000000000000000010000000000000000000000',
 '000000000001000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000',
 '000000000000000000000000010100000000000000000000']
'''

The 48 bits record the keyboard state in that frame:

    000000000000000000000000000100000000000000000000
                           ***** ***                pressing ctrl, right, left, down, up, shift,x,z
    000000000001000000000000000101000000000000000000
           ***** ***                                press ctrl, right, left, down, up, shift,x,z
    000000000000000000000000000001000000000000010000
                                           ***** ***release ctrl, right,left,down,up,shift,x,z

For detail, you can check the document string of replay_to_binary_seq included in utils.py.