def __init__(self, hp):

self.hp = hp

self.target_path_origin = self.hp.direct_dir + '-origin'

self.aws_access_id = os.environ.get("access_id")

self.aws_secret_key = os.environ.get("secret_access_key")

self.aws_bucket_name = os.environ.get("bucket_name")

def load_from_s3(self):

# set aws credentials

s3r = boto3.resource('s3', aws_access_key_id=self.aws_access_id,

aws_secret_access_key=self.aws_secret_key)

bucket = s3r.Bucket(self.aws_bucket_name)

# mkdir

os.makedirs(self.hp.base_dir, exist_ok=True)

os.makedirs(self.hp.data_dir, exist_ok=True)

os.makedirs(self.hp.direct_dir, exist_ok=True)

os.makedirs(self.target_path_origin, exist_ok=True)

for object in bucket.objects.filter(Prefix = self.hp.direct_dir):

bucket.download_file(object.key, object.key)

def m4a_to_wav(self):

m4a_list = os.listdir(self.hp.direct_dir)

m4a_list.sort()

for m4a in m4a_list:

track = AudioSegment.from_file(os.path.join(self.hp.direct_dir, m4a), format='m4a')

wav = re.sub(r'[^0-9]', '', m4a[:-3])

wav += '.wav'

file_handle = track.export(os.path.join(self.target_path_origin, wav), format='wav')

os.system('rm -rf {}'.format(self.hp.direct_dir))

def down_sampling(self):

os.makedirs(self.hp.direct_dir, exist_ok=True)

wav_origin_list = os.listdir(self.target_path_origin)

wav_origin_list.sort()

for wav in wav_origin_list:

sample_rate, _ = sio.wavfile.read(self.target_path_origin + '/' + wav)

if sample_rate != self.hp.sampling_rate:

os.makedirs(self.hp.direct_dir, exist_ok=True)

os.system('ffmpeg -i {} -ac 1 -ar {} {} -y'.format(self.target_path_origin + '/' + wav, str(self.hp.sampling_rate), self.hp.direct_dir + '/' + wav))

os.system('rm -rf {}'.format(self.target_path_origin))

def s3_target_load(self):

self.load_from_s3()

files = os.listdir(self.hp.direct_dir)

if files[int(len(files)/2)][-3:] == 'm4a':

self.m4a_to_wav()