import wave
import struct

# the format of BBC Micro tape files is:
#   - a 0 bit is 1200 Hz (sampled at 44100Hz, wavelength is 36 samples)
#   - a 1 bit is 2400 Hz (ditto, 18 samples)
#  Leader tone is 2400 Hz
#
#  

# ref http://beebwiki.mdfs.net/Acorn_cassette_format

def physical_layer_read(filename):
	thames = wave.open(filename, 'rb')
	freq = thames.getframerate()

	result = []

	if thames.getsampwidth()!=2:
		raise Exception('Sorry, sample width is weird: %d' % (thames.getsampwidth(),))

	contents = thames.readframes(thames.getnframes())

	prev = None
	count = 0

	# Because 2400Hz tones fill two cycles in the space
	# where 1200Hz tones fill one, we will see them appear
	# twice. If this flag is set, we just output a 1 (for
	# the 2400Hz tone), so if we see another 2400Hz cycle
	# we shouldn't output another 1.
	in_2400 = False

	for i in range(0, len(contents)/2):
		value = struct.unpack('<h', contents[i*2:i*2+2])[0]
		if value<0:
			current = 'low'
		elif value>0:
			current = 'high'
		else:
			continue

		if prev=='low' and current=='high':
			if count < 27:
				# Short wavelength = high frequency
				if not in_2400:
					result.append(1)
					in_2400 = True
				else:
					in_2400 = False
			else:
				# Long wavelength = low frequency
				result.append(0)
				in_2400 = False
			count = 0
		else:
			count += 1

		prev = current

	return result

def data_layer_read(bits):

	skipcount = 0
	result = []

	i = 0
	while i<len(bits)-10:

		if bits[i]==0 and bits[i+9]==1:

			if skipcount > 32:
				# there's been a lead tone,
				# so start a new block
				result.append('')
			skipcount = 0

			value = 0

			for k in range(i+1, i+9):
				value = value >> 1
				value += bits[k]*128

			result[-1] += chr(value)

			i += 10

		else:
			i += 1
			skipcount += 1

	return result

def block_layer_read(block):
	result = { 
		'filename': '',
	}

	if block[0]!='*':
		# don't print error for glitches
		if len(block)>32:
			print 'warning: magic number missing'
		return

	i = 1
	while block[i]!=chr(0):
		result['filename'] += block[i]
		i += 1

	i += 1 # ? off by one ?
	i += 4 # skip load address
	i += 4 # skip exec address

	result['serial'] = ord(block[i+1])<<8 | ord(block[i])
	i += 2

	result['length'] = ord(block[i+1])<<8 | ord(block[i])
	i += 2

	result['flags'] = ord(block[i])
	i += 1

	i += 2 # skip CRC
	i += 4 # skip something else (?)

	result['data'] = block[i:-2]

	return result

def main():
	bits = physical_layer_read('bbc2.wav')
	blocks = data_layer_read(bits)

	files = []
	for block in blocks:
		details = block_layer_read(block)

		if details is None:
			continue

		print details

		print details['filename'], details['serial']
		if details['serial'] == 0:
			files.append('')

		files[-1]+=details['data']

	for i in range(len(files)):
		filename = 'thames%d.dat' % (i,)
		open(filename, 'wb').write(files[i])

if __name__=='__main__':
	main()