1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
|
import struct
def padto(n, m):
return (n+m-1) & ~(m-1)
def resourceclass(rtype):
return globals().get(rtype.strip() + 'Resource', Resource)
class TypeList:
def __init__(self, type, fmap, fdata, namebase, start, count):
self.type = type
self.fmap = fmap
self.fdata = fdata
self.namebase = namebase
self.start = start
self.count = count
self.ids = None
def resources(self):
if self.ids is None:
ResourceCls = resourceclass(self.type)
d = {}
self.fmap.seek(self.start)
for resid, resname, resattr, resofshi, resofslo in [
struct.unpack(">HHBBHxxxx", self.fmap.read(12))
for i in range(self.count)]:
if resname == 0xffff:
name = None
else:
self.fmap.seek(self.namebase + resname)
namelen, = struct.unpack(">B", self.fmap.read(1))
name = self.fmap.read(namelen)
assert resid not in d
d[resid] = ResourceCls(self.type, resid, name, resattr,
self.fdata, resofslo + (resofshi<<16))
self.ids = d
return self.ids
def __getitem__(self, id):
return self.resources()[id]
def keys(self):
return list(self.resources().keys())
def values(self):
return list(self.resources().values())
def items(self):
return list(self.resources().items())
def namedict(self):
return dict([(r.name, r) for r in list(self.resources().values()) if r.name is not None])
class MacBinary:
def __init__(self, f):
if type(f) is type(''):
f = open(f, 'rb')
self.f = f
self.f.seek(0x53)
self.dataforksize, self.resforksize = struct.unpack(">ll", self.f.read(8))
self.loadresources()
def getdata(self):
self.f.seek(0x80)
return self.f.read(self.dataforksize)
def loadresources(self):
f = Subfile(self.f, padto(0x80 + self.dataforksize, 0x80), self.resforksize)
ofsdata, ofsmap, lendata, lenmap = struct.unpack(">llll", f.read(16))
fdata = Subfile(f, ofsdata, lendata)
fmap = Subfile(f, ofsmap, lenmap)
fmap.seek(24)
ofstype, ofsname = struct.unpack(">HH", fmap.read(4))
self.dtypes = {}
fmap.seek(ofstype)
numtypes, = struct.unpack(">H", fmap.read(2))
numtypes = numtypes + 1
for rtype, num, ofsref in [struct.unpack(">4sHH", fmap.read(8))
for i in range(numtypes)]:
assert rtype not in self.dtypes
self.dtypes[rtype] = TypeList(rtype, fmap, fdata, ofsname,
ofstype + ofsref, num + 1)
def __getitem__(self, rtype):
return self.dtypes[rtype]
def types(self):
return self.dtypes
def keys(self):
return list(self.dtypes.keys())
def values(self):
return list(self.dtypes.values())
def items(self):
return list(self.dtypes.items())
class Subfile:
def __init__(self, f, start, length):
if start < 0:
raise ValueError('negative position')
if isinstance(f, Subfile):
if start + length > f.length:
raise ValueError('subfile out of bounds')
f, start = f.f, f.start+start
self.f = f
self.start = start
self.length = length
self.position = 0
def read(self, size=None):
if size is None or self.position + size > self.length:
size = self.length - self.position
if size <= 0:
return ''
self.f.seek(self.start + self.position)
self.position = self.position + size
return self.f.read(size)
def seek(self, npos):
if npos < 0:
raise ValueError('negative position')
self.position = npos
class Resource:
def __init__(self, type, id, name, attr, srcfile, srcofs):
self.type = type
self.id = id
self.name = name
self.attr = attr
self.srcfile = srcfile
self.srcofs = srcofs
def subfile(self):
self.srcfile.seek(self.srcofs)
length, = struct.unpack(">l", self.srcfile.read(4))
return Subfile(self.srcfile, self.srcofs + 4, length)
def load(self):
return self.subfile().read()
class RGBImage:
def __init__(self, w, h, data):
assert len(data) == 3*w*h
self.w = w
self.h = h
self.data = data
def loadcolormap(f):
size, = struct.unpack(">xxxxxxH", f.read(8))
size = size + 1
d = {}
for index, r, g, b in [struct.unpack(">HHHH", f.read(8)) for i in range(size)]:
assert index not in d, 'duplicate color index'
d[index] = r/256.0, g/256.0, b/256.0
return d
def image2rgb(image):
# returns (w, h, data)
h = len(image)
result1 = []
for line in image:
for r, g, b in line:
result1.append(chr(int(r)) + chr(int(g)) + chr(int(b)))
return len(image[0]), len(image), ''.join(result1)
class clutResource(Resource):
# a color table
def gettable(self):
return loadcolormap(self.subfile())
class ppatResource(Resource):
# a pattern
def getimage(self):
f = self.subfile()
pattype, patmap, patdata = struct.unpack(">Hll", f.read(10))
if pattype != 1:
raise ValueError('Pattern type not supported')
f.seek(patmap)
(rowBytes, h, w, packType, packSize,
pixelType, pixelSize, cmpCount, cmpSize, pmTable) = (
struct.unpack(">xxxxHxxxxHHxxHlxxxxxxxxHHHHxxxxlxxxx", f.read(50)))
isBitmap = (rowBytes & 0x8000) != 0
rowBytes &= 0x3FFF
if packType != 0:
raise ValueError('packed image not supported')
if pixelType != 0 or cmpCount != 1:
raise ValueError('direct RGB image not supported')
assert cmpSize == pixelSize and pixelSize in [1,2,4,8]
f.seek(pmTable)
colormap = loadcolormap(f)
bits_per_pixel = pixelSize
pixels_per_byte = 8 // bits_per_pixel
image = []
f.seek(patdata)
for y in range(h):
line = f.read(rowBytes)
imgline = []
for x in range(w):
n = x//pixels_per_byte
idx = ((ord(line[n]) >> ((pixels_per_byte - 1 - x%pixels_per_byte) * bits_per_pixel))
& ((1<<bits_per_pixel)-1))
imgline.append(colormap[idx])
image.append(imgline)
return image
class LEVLResource(Resource):
# bub & bob level
WIDTH = 32
HEIGHT = 25
MONSTERS = 30
WALLS = { 1:'#', 0:' '}
WINDS = { 0:' ', 1:'>', 2:'<', 3:'v', 4:'^', 5:'x', 0x66:' '}
FLAGS = ['flag0', 'letter', 'fire', 'lightning', 'water', 'top', 'flag6', 'flag7']
def getlevel(self, mnstrlist):
f = self.subfile()
result = {}
walls = []
for y in range(self.HEIGHT):
line = f.read(self.WIDTH//8)
line = [self.WALLS[(ord(line[x//8]) >> (x%8)) & 1]
for x in range(self.WIDTH)]
walls.append(''.join(line))
result['walls'] = '\n'.join(walls)
winds = []
for y in range(self.HEIGHT):
line = f.read(self.WIDTH)
line = [self.WINDS[ord(v)] for v in line]
winds.append(''.join(line))
result['winds'] = '\n'.join(winds)
monsters = []
for i in range(self.MONSTERS):
x,y,monster_type,f1,f2,f3 = struct.unpack(">BBBBBB", f.read(6))
if monster_type != 0:
assert f1 == 0, f1
cls = mnstrlist[monster_type-1]
monsters.append(cls(x=x, y=y, dir=f2, player=f3))
result['monsters'] = monsters
result['level'], = struct.unpack('>H', f.read(2))
for i in range(8):
result[self.FLAGS[i]] = ord(f.read(1))
return result
|