1 Star 0 Fork 27

浪子阿杰/youku_fetch_video

加入 Gitee
与超过 1200万 开发者一起发现、参与优秀开源项目,私有仓库也完全免费 :)
免费加入
文件
该仓库未声明开源许可证文件(LICENSE),使用请关注具体项目描述及其代码上游依赖。
克隆/下载
flv_join.py 8.96 KB
一键复制 编辑 原始数据 按行查看 历史
Garethxiao 提交于 2013-07-11 13:59 . create project
#!/usr/bin/env python
import struct
from cStringIO import StringIO
TAG_TYPE_METADATA = 18
##################################################
# AMF0
##################################################
AMF_TYPE_NUMBER = 0x00
AMF_TYPE_BOOLEAN = 0x01
AMF_TYPE_STRING = 0x02
AMF_TYPE_OBJECT = 0x03
AMF_TYPE_MOVIECLIP = 0x04
AMF_TYPE_NULL = 0x05
AMF_TYPE_UNDEFINED = 0x06
AMF_TYPE_REFERENCE = 0x07
AMF_TYPE_MIXED_ARRAY = 0x08
AMF_TYPE_END_OF_OBJECT = 0x09
AMF_TYPE_ARRAY = 0x0A
AMF_TYPE_DATE = 0x0B
AMF_TYPE_LONG_STRING = 0x0C
AMF_TYPE_UNSUPPORTED = 0x0D
AMF_TYPE_RECORDSET = 0x0E
AMF_TYPE_XML = 0x0F
AMF_TYPE_CLASS_OBJECT = 0x10
AMF_TYPE_AMF3_OBJECT = 0x11
class ECMAObject:
def __init__(self, max_number):
self.max_number = max_number
self.data = []
self.map = {}
def put(self, k, v):
self.data.append((k, v))
self.map[k] = v
def get(self, k):
return self.map[k]
def set(self, k, v):
for i in range(len(self.data)):
if self.data[i][0] == k:
self.data[i] = (k, v)
break
else:
raise KeyError(k)
self.map[k] = v
def keys(self):
return self.map.keys()
def __str__(self):
return 'ECMAObject<'+repr(self.map)+'>'
def __eq__(self, other):
return self.max_number == other.max_number and self.data == other.data
def read_amf_number(stream):
return struct.unpack('>d', stream.read(8))[0]
def read_amf_boolean(stream):
b = read_byte(stream)
assert b in (0, 1)
return bool(b)
def read_amf_string(stream):
xx = stream.read(2)
if xx == '':
# dirty fix for the invalid Qiyi flv
return None
n = struct.unpack('>H', xx)[0]
s = stream.read(n)
assert len(s) == n
return s.decode('utf-8')
def read_amf_object(stream):
obj = {}
while True:
k = read_amf_string(stream)
if not k:
assert read_byte(stream) == AMF_TYPE_END_OF_OBJECT
break
v = read_amf(stream)
obj[k] = v
return obj
def read_amf_mixed_array(stream):
max_number = read_uint(stream)
mixed_results = ECMAObject(max_number)
while True:
k = read_amf_string(stream)
if k is None:
# dirty fix for the invalid Qiyi flv
break
if not k:
assert read_byte(stream) == AMF_TYPE_END_OF_OBJECT
break
v = read_amf(stream)
mixed_results.put(k, v)
assert len(mixed_results.data) == max_number
return mixed_results
def read_amf_array(stream):
n = read_uint(stream)
v = []
for i in range(n):
v.append(read_amf(stream))
return v
amf_readers = {
AMF_TYPE_NUMBER: read_amf_number,
AMF_TYPE_BOOLEAN: read_amf_boolean,
AMF_TYPE_STRING: read_amf_string,
AMF_TYPE_OBJECT: read_amf_object,
AMF_TYPE_MIXED_ARRAY: read_amf_mixed_array,
AMF_TYPE_ARRAY: read_amf_array,
}
def read_amf(stream):
return amf_readers[read_byte(stream)](stream)
def write_amf_number(stream, v):
stream.write(struct.pack('>d', v))
def write_amf_boolean(stream, v):
if v:
stream.write('\x01')
else:
stream.write('\x00')
def write_amf_string(stream, s):
s = s.encode('utf-8')
stream.write(struct.pack('>H', len(s)))
stream.write(s)
def write_amf_object(stream, o):
for k in o:
write_amf_string(stream, k)
write_amf(stream, o[k])
write_amf_string(stream, '')
write_byte(stream, AMF_TYPE_END_OF_OBJECT)
def write_amf_mixed_array(stream, o):
write_uint(stream, o.max_number)
for k, v in o.data:
write_amf_string(stream, k)
write_amf(stream, v)
write_amf_string(stream, '')
write_byte(stream, AMF_TYPE_END_OF_OBJECT)
def write_amf_array(stream, o):
write_uint(stream, len(o))
for v in o:
write_amf(stream, v)
amf_writers_tags = {
float: AMF_TYPE_NUMBER,
bool: AMF_TYPE_BOOLEAN,
unicode: AMF_TYPE_STRING,
dict: AMF_TYPE_OBJECT,
ECMAObject: AMF_TYPE_MIXED_ARRAY,
list: AMF_TYPE_ARRAY,
}
amf_writers = {
AMF_TYPE_NUMBER: write_amf_number,
AMF_TYPE_BOOLEAN: write_amf_boolean,
AMF_TYPE_STRING: write_amf_string,
AMF_TYPE_OBJECT: write_amf_object,
AMF_TYPE_MIXED_ARRAY: write_amf_mixed_array,
AMF_TYPE_ARRAY: write_amf_array,
}
def write_amf(stream, v):
if isinstance(v, ECMAObject):
tag = amf_writers_tags[ECMAObject]
else:
tag = amf_writers_tags[type(v)]
write_byte(stream, tag)
amf_writers[tag](stream, v)
##################################################
# FLV
##################################################
def read_int(stream):
return struct.unpack('>i', stream.read(4))[0]
def read_uint(stream):
return struct.unpack('>I', stream.read(4))[0]
def write_uint(stream, n):
stream.write(struct.pack('>I', n))
def read_byte(stream):
return ord(stream.read(1))
def write_byte(stream, b):
stream.write(chr(b))
def read_unsigned_medium_int(stream):
x1, x2, x3 = struct.unpack('BBB', stream.read(3))
return (x1 << 16) | (x2 << 8) | x3
def read_tag(stream):
# header size: 15 bytes
header = stream.read(15)
if len(header) == 4:
return
x = struct.unpack('>IBBBBBBBBBBB', header)
previous_tag_size = x[0]
data_type = x[1]
body_size = (x[2] << 16) | (x[3] << 8) | x[4]
assert body_size < 1024*1024*128, 'tag body size too big (> 128MB)'
timestamp = (x[5] << 16) | (x[6] << 8) | x[7]
timestamp += x[8] << 24
assert x[9:] == (0, 0, 0)
body = stream.read(body_size)
return (data_type, timestamp, body_size, body, previous_tag_size)
#previous_tag_size = read_uint(stream)
#data_type = read_byte(stream)
#body_size = read_unsigned_medium_int(stream)
#assert body_size < 1024*1024*128, 'tag body size too big (> 128MB)'
#timestamp = read_unsigned_medium_int(stream)
#timestamp += read_byte(stream) << 24
#assert read_unsigned_medium_int(stream) == 0
#body = stream.read(body_size)
#return (data_type, timestamp, body_size, body, previous_tag_size)
def write_tag(stream, tag):
data_type, timestamp, body_size, body, previous_tag_size = tag
write_uint(stream, previous_tag_size)
write_byte(stream, data_type)
write_byte(stream, body_size>>16 & 0xff)
write_byte(stream, body_size>>8 & 0xff)
write_byte(stream, body_size & 0xff)
write_byte(stream, timestamp>>16 & 0xff)
write_byte(stream, timestamp>>8 & 0xff)
write_byte(stream, timestamp & 0xff)
write_byte(stream, timestamp>>24 & 0xff)
stream.write('\0\0\0')
stream.write(body)
def read_flv_header(stream):
assert stream.read(3) == 'FLV'
header_version = read_byte(stream)
assert header_version == 1
type_flags = read_byte(stream)
assert type_flags == 5
data_offset = read_uint(stream)
assert data_offset == 9
def write_flv_header(stream):
stream.write('FLV')
write_byte(stream, 1)
write_byte(stream, 5)
write_uint(stream, 9)
def read_meta_data(stream):
meta_type = read_amf(stream)
meta = read_amf(stream)
return meta_type, meta
def read_meta_tag(tag):
data_type, timestamp, body_size, body, previous_tag_size = tag
assert data_type == TAG_TYPE_METADATA
assert timestamp == 0
assert previous_tag_size == 0
return read_meta_data(StringIO(body))
def write_meta_data(stream, meta_type, meta_data):
assert isinstance(meta_type, basesting)
write_amf(meta_type)
write_amf(meta_data)
def write_meta_tag(stream, meta_type, meta_data):
buffer = StringIO()
write_amf(buffer, meta_type)
write_amf(buffer, meta_data)
body = buffer.getvalue()
write_tag(stream, (TAG_TYPE_METADATA, 0, len(body), body, 0))
##################################################
# main
##################################################
def guess_output(inputs):
import os.path
inputs = map(os.path.basename, inputs)
n = min(map(len, inputs))
for i in reversed(range(1, n)):
if len(set(s[:i] for s in inputs)) == 1:
return inputs[0][:i] + '.flv'
return 'output.flv'
def concat_flvs(flvs, output=None):
assert flvs, 'no flv file found'
import os.path
if not output:
output = guess_output(flvs)
elif os.path.isdir(output):
output = os.path.join(output, guess_output(flvs))
print 'Joining %s into %s' % (', '.join(flvs), output)
ins = [open(flv, 'rb') for flv in flvs]
for stream in ins:
read_flv_header(stream)
meta_tags = map(read_tag, ins)
metas = map(read_meta_tag, meta_tags)
meta_types, metas = zip(*metas)
assert len(set(meta_types)) == 1
meta_type = meta_types[0]
# must merge fields: duration
# TODO: check other meta info, update other meta info
total_duration = sum(meta.get('duration') for meta in metas)
meta_data = metas[0]
meta_data.set('duration', total_duration)
out = open(output, 'wb')
write_flv_header(out)
write_meta_tag(out, meta_type, meta_data)
timestamp_start = 0
for stream in ins:
while True:
tag = read_tag(stream)
if tag:
data_type, timestamp, body_size, body, previous_tag_size = tag
timestamp += timestamp_start
tag = data_type, timestamp, body_size, body, previous_tag_size
write_tag(out, tag)
else:
break
timestamp_start = timestamp
write_uint(out, previous_tag_size)
return output
def usage():
print 'python flv_join.py --output target.flv flv...'
def main():
import sys, getopt
try:
opts, args = getopt.getopt(sys.argv[1:], "ho:", ["help", "output="])
except getopt.GetoptError, err:
usage()
sys.exit(1)
output = None
for o, a in opts:
if o in ("-h", "--help"):
usage()
sys.exit()
elif o in ("-o", "--output"):
output = a
else:
usage()
sys.exit(1)
if not args:
usage()
sys.exit(1)
concat_flvs(args, output)
if __name__ == '__main__':
main()
马建仓 AI 助手
尝试更多
代码解读
代码找茬
代码优化
1
https://gitee.com/662871/youku_fetch_video.git
git@gitee.com:662871/youku_fetch_video.git
662871
youku_fetch_video
youku_fetch_video
master

搜索帮助