n-grams/music.py

import os, sys
import random
import datetime
import glob
import markov
import pickle


corpus = []
song = []

#TODO: convert these into inputs
# lengthofsong = 10 Should we control this? Setting it to random now
timesignature = ['3/4','4/4','1/8','C'] #Sometimes the letter “C” (meaning common time) will be used in place of 4/4. 
#Both C and 4/4 indicate that there are four quarter note beats in each measure.
keysignature = "G"
key_enforced = False
key_enforced = True #Set to true if user wants in specific key
print("hello")
# get the list of filenames (abc files downloaded from http://www.norbeck.nu/abc/)
# getdirs = []
# dirs = ["hn201612/i/*.abc", "hn201612/s/*.abc"]
# dirs = ["data/*.abc"]
# dirs = ["data"]
# for dir1 in dirs:
# 	for filename in glob.iglob(dir1):
# 		getdirs += [filename]

selected_timeSign = '3/4'
#Finds all absolute paths in directory
#https://stackoverflow.com/questions/9816816/get-absolute-paths-of-all-files-in-a-directory
def abs_paths(dir):
    for dir_path,_,filenames in os.walk(dir):
        for f in filenames:
            yield os.path.abspath(os.path.join(dir_path, f))
# ex_filename = "hn201612/i/hnsong1.abc"
# parsing on file to extract songs and add them to corpus
for filename in abs_paths("n-grams/data/beginner"):
	with open(filename) as f:
		print(filename)
		lines = f.readlines()
		last = len(lines)
		accepted = False
		for index, line in enumerate(lines):
			if (line.find("|") < 0 and index - 1 == last):
				# if the next line does not have pipes add song to corpus and then set song variable empty again
				if accepted and key_enforced and key_accepted:
					corpus.append(song)
					accepted = False
					key_accepted = False
				if accepted:
					corpus.append(song)
					accepted = False
				song = []
			else:
				if line.find("|") > -1:
					# a line should be split on "|" and copied to the corpus if it has pipes
					sline = line.split("|")
					# add the list of measures to the song
					song += [x.strip("\r\n") for x in sline if len(x.strip("\r\n")) > 0]
					last = index
				elif "M:" in line:
					#time signature
					if selected_timeSign == "4/4":
						if "4/4" in line or "C|" in line:
							accepted = True
					elif selected_timeSign in line:
						accepted = True
				elif line.find("K:") and key_enforced:
					#key signature
					if keysignature in line:
						key_accepted = True
					
print("Training on {} songs...".format(len(corpus)))

# MARKOV PART
# n-gram length for markov model
n = 1

model = markov.generate_model_from_token_lists(corpus, n)


# save pickle
with open('markov_chain.pickle', 'wb') as handle:
	pickle.dump(model, handle)


def nextword(word):
	return markov.generate(model, 3, seed=word, max_iterations=1)


def writesong(songlength, first):
	song = [first]
	for i in range(songlength):
		song += nextword(str(song[-1]))
	return song

# choose a random song length from list of song lengths in corpus
lengthofsong = random.choice([len(x) for x in corpus if len(x) > 10])
print("Song length will be {}".format(lengthofsong))
song_len = [len(x) for x in corpus if len(x)>10]
song_len.sort()
print("Song lengths",song_len)
firstnote = markov.generate(model, n, max_iterations=3)[0]
# print "first note: {}".format(firstnote)

print("Here is the song in abc format:")
song = writesong(lengthofsong, firstnote)
dob = datetime.datetime.now().strftime('%H%M')

print(dob)
print(song)

# make song file
songname = "n-grams/gen_songs_abc/gen_song_{}.abc".format(dob)
print("\n\nYou can find the song in {}".format(songname))
lastpart = lengthofsong - lengthofsong%4

# hack to include dictionary at the beginning of every abc file
# will add a more sophisticated way to generate the values in the future
title = "Markov Song {}".format(dob)
songbeginning = ['X:1','T:' + title, 'R:song', 'C:Visakh Ajith', 'Z:id:hn-song-111', 'M:3/4', 'L:1/8', 'Q:1/4=120', 'K:G'
]
songbeginning = [x+"\n" for x in songbeginning]

# convert song to abc format and write to file
newsong = open(songname, 'w')
newsong.writelines(songbeginning)
for i in range(lastpart):
	newsong.write(" | ".join(song[i:i+3]) + "\n")
newsong.write(" | ".join(song[lastpart:lengthofsong]))

#abc2ly markov.abc
# lilypond -fpng markov.ly