import argparse

import os

import hdf5_getters

import numpy as np

from glob import glob

parser = argparse.ArgumentParser(description='Million Song Dataset Genre Classification')

parser.add_argument('-ic', '--input-classes', type=str, help='Input Tag traum classes path',

default="data/TagTraum/msd_tagtraum_cd2c.cls")

parser.add_argument('-msd', '--msd', type=str, help='MSD /data path',

default="data/MSD")

parser.add_argument('-o', '--output', type=str, help='Output data file path',

default="extractedLetters/output.csv")

args = vars(parser.parse_args())

print(args)

INPUTCLASSES = args["input_classes"]

MSD = args["msd"]

OUTPUT = args['output']

# Bring TagTraum's genres into memory

genres = {}

with open(INPUTCLASSES, 'r') as f:

for line in f:

[track_id, track_genre] = line.strip().split(maxsplit=1)

genres[track_id] = track_genre

# Feature names

featnames = ''

featnames += 'genre,track_id,artist_name,title,loudness,tempo,time_signature,key,mode,duration,'

featnames += 'avg_timbre1,avg_timbre2,avg_timbre3,avg_timbre4,avg_timbre5,avg_timbre6,'

featnames += 'avg_timbre7,avg_timbre8,avg_timbre9,avg_timbre10,avg_timbre11,avg_timbre12,'

featnames += 'var_timbre1,var_timbre2,var_timbre3,var_timbre4,var_timbre5,var_timbre6,'

featnames += 'var_timbre7,var_timbre8,var_timbre9,var_timbre10,var_timbre11,var_timbre12,'

featnames += 'min_timbre1,min_timbre2,min_timbre3,min_timbre4,min_timbre5,min_timbre6,'

featnames += 'min_timbre7,min_timbre8,min_timbre9,min_timbre10,min_timbre11,min_timbre12,'

featnames += 'max_timbre1,max_timbre2,max_timbre3,max_timbre4,max_timbre5,max_timbre6,'

featnames += 'max_timbre7,max_timbre8,max_timbre9,max_timbre10,max_timbre11,max_timbre12,'

featnames += 'ptp_timbre1,ptp_timbre2,ptp_timbre3,ptp_timbre4,ptp_timbre5,ptp_timbre6,'

featnames += 'ptp_timbre7,ptp_timbre8,ptp_timbre9,ptp_timbre10,ptp_timbre11,ptp_timbre12,'

featnames += 'mu_timbre1,mu_timbre2,mu_timbre3,mu_timbre4,mu_timbre5,mu_timbre6,'

featnames += 'mu_timbre7,mu_timbre8,mu_timbre9,mu_timbre10,mu_timbre11,mu_timbre12,'

featnames += 'med_timbre1,med_timbre2,med_timbre3,med_timbre4,med_timbre5,med_timbre6,'

featnames += 'med_timbre7,med_timbre8,med_timbre9,med_timbre10,med_timbre11,med_timbre12,'

featnames += 'std_timbre1,std_timbre2,std_timbre3,std_timbre4,std_timbre5,std_timbre6,'

featnames += 'std_timbre7,std_timbre8,std_timbre9,std_timbre10,std_timbre11,std_timbre12,'

featnames += 'avg_pitches1,avg_pitches2,avg_pitches3,avg_pitches4,avg_pitches5,avg_pitches6,'

featnames += 'avg_pitches7,avg_pitches8,avg_pitches9,avg_pitches10,avg_pitches11,avg_pitches12,'

featnames += 'var_pitches1,var_pitches2,var_pitches3,var_pitches4,var_pitches5,var_pitches6,'

featnames += 'var_pitches7,var_pitches8,var_pitches9,var_pitches10,var_pitches11,var_pitches12,'

featnames += 'min_pitches1,min_pitches2,min_pitches3,min_pitches4,min_pitches5,min_pitches6,'

featnames += 'min_pitches7,min_pitches8,min_pitches9,min_pitches10,min_pitches11,min_pitches12,'

featnames += 'max_pitches1,max_pitches2,max_pitches3,max_pitches4,max_pitches5,max_pitches6,'

featnames += 'max_pitches7,max_pitches8,max_pitches9,max_pitches10,max_pitches11,max_pitches12,'

featnames += 'ptp_pitches1,ptp_pitches2,ptp_pitches3,ptp_pitches4,ptp_pitches5,ptp_pitches6,'

featnames += 'ptp_pitches7,ptp_pitches8,ptp_pitches9,ptp_pitches10,ptp_pitches11,ptp_pitches12,'

featnames += 'mu_pitches1,mu_pitches2,mu_pitches3,mu_pitches4,mu_pitches5,mu_pitches6,'

featnames += 'mu_pitches7,mu_pitches8,mu_pitches9,mu_pitches10,mu_pitches11,mu_pitches12,'

featnames += 'med_pitches1,med_pitches2,med_pitches3,med_pitches4,med_pitches5,med_pitches6,'

featnames += 'med_pitches7,med_pitches8,med_pitches9,med_pitches10,med_pitches11,med_pitches12,'

featnames += 'std_pitches1,std_pitches2,std_pitches3,std_pitches4,std_pitches5,std_pitches6,'

featnames += 'std_pitches7,std_pitches8,std_pitches9,std_pitches10,std_pitches11,std_pitches12,'

featnames += 'avg_loudness_max,var_loudness_max,min_loudness_max,max_loudness_max,'

featnames += 'ptp_loudness_max,mu_loudness_max,med_loudness_max,std_loudness_max,'

featnames += 'avg_loudness_max_time,var_loudness_max_time,min_loudness_max_time,max_loudness_max_time,'

featnames += 'ptp_loudness_max_time,mu_loudness_max_time,med_loudness_max_time,std_loudness_max_time,'

featnames += 'avg_loudness_start,var_loudness_start,min_loudness_start,max_loudness_start,'

featnames += 'ptp_loudness_start,mu_loudness_start,med_loudness_start,std_loudness_start,'

featnames += 'hotttnesss,danceability,end_of_fade_in,energy,start_of_fade_out,year\n'

num_feats = len(featnames.split(','))

def get_statistical_feats(feat):

return sf

def get_feats(h5):

return f

output_filename = OUTPUT

if not output_filename.endswith('.csv'):

output_filename = output_filename + ".csv"

exists = os.path.isfile(output_filename)

# Generate output file

with open(output_filename, 'a', encoding='utf-8') as output:

if not exists:

output.write(featnames)

# Go through all files in ./data/MSD

files = [y for x in os.walk('./data/MSD') for y in glob(os.path.join(x[0], '*.h5'))]

for file in files:

filename = os.path.splitext(os.path.basename(file))[0]

# Get track's features

if filename in genres:

h5 = hdf5_getters.open_h5_file_read(file)

feats = [genres[filename], filename]

feats.extend(get_feats(h5))

# Close h5 and write into output file

h5.close()

assert len(feats) == num_feats, 'feat length problem, len(feats)='+str(len(feats))

output.write(','.join(feats) + '\n')