1
0
Fork 0
mirror of https://github.com/zeldaret/oot.git synced 2024-12-29 08:16:11 +00:00
oot/tools/decompile_data.py

117 lines
2.9 KiB
Python
Raw Normal View History

2020-03-17 04:31:30 +00:00
#! /usr/bin/python3
import os
import re
import struct
2020-03-17 04:31:30 +00:00
script_dir = os.path.dirname(os.path.realpath(__file__))
root_dir = script_dir + "/../"
data_dir = root_dir + "data/"
asm_dir = root_dir + "asm/"
2020-03-17 04:31:30 +00:00
def try_text(text_bytes):
bad_bytes = 0
for byte in text_bytes:
if byte < 32:
bad_bytes += 1
# Arbitrary string detection heuristic
#if bad_bytes / len(text_bytes) >= 0.3:
# return None
2020-03-17 04:31:30 +00:00
try:
text = text_bytes.decode("EUC-JP")
except UnicodeDecodeError:
return None
text = text.strip(" \0")
if len(text) > 0 and bool(re.search('[a-zA-Z]', text)):
text = text.replace("\\x00", "")
text = text.replace("\n", "\\n")
text = text.replace("\"", "\\\"")
return text
def is_zeros(stuff):
for piece in stuff:
if piece.strip() != "0x00000000":
return False
return True
2020-03-17 04:31:30 +00:00
def try_float(word):
if (word[:3] == "0x3") or (word[:3] == "0x4") or \
(word[:3] == "0xB") or (word[:3] == "0xC"):
return struct.unpack('!f', bytes.fromhex(word[2:10]))[0]
def quick_convert(words_string):
words = words_string.split(",")
byte_array = b""
for word in words:
data = word.strip()[2:]
byte_array += bytearray.fromhex(data)
return byte_array.decode("EUC-JP")
2020-03-17 04:31:30 +00:00
def word_convert(byte_string):
try:
words = byte_string.split(",")
byte_array = b""
for word in words:
data = word.strip()[2:]
byte_array += bytearray.fromhex(data)
except ValueError:
return byte_string
if byte_array[0] == 63 and byte_array[-1] == 45:
return byte_string
if data == '0A0A0000':
return "\\n\\n"
if len(words) > 1 and not is_zeros(words[1:]):
2020-03-17 04:31:30 +00:00
res = try_text(byte_array)
if res is not None:
return " .asciz \"" + res + "\"\n .balign 4\n"
if len(words) == 1 or is_zeros(words[1:]):
res = str(try_float(words[0].strip()))
if res is not None:
return " .float " + res + "\n"
2020-03-17 04:31:30 +00:00
return byte_string
def handle_match(match):
in_str = match.group()[6:]
ret = word_convert(in_str)
if ret == in_str:
return match.group()
return ret
def process_data_file(file_path):
with open(file_path) as f:
file_text = f.read()
new_file_text = re.sub(" \\.word 0x.*", handle_match, file_text)
if new_file_text != file_text:
with open(file_path, "w") as f:
f.write(new_file_text)
return True
return False
def main():
i = 0
for root, dirs, files in os.walk(asm_dir):
2020-03-17 04:31:30 +00:00
for file in files:
#if i == 10:
# return
if file.endswith(".s"):
2020-03-17 04:31:30 +00:00
path = os.path.join(root, file)
if process_data_file(path):
print("Processed " + path)
i += 1
main()