importing a million song dataset file and converting to a dataframe

import msd to dataframe.py

import pandas as pd

#open and split file then convert to df
lines = [line.strip().split("\t") for line in open("P:\\A.tsv.a.txt", "r")]
df=pd.DataFrame(lines)

#pull out columns for further split
cols=range(18,22)+range(33,42)
arrays=df.loc[1:5,cols].values

#split - doesnt work too many levels
#out=[[item.split(",") for item in array] for array in arrays]
#df2=pd.DataFrame(out)

#split column at a time
col=df[20]
#print col
res=[item.split(",") for item in col]
df2=pd.DataFrame(res)
#problem is inconsitent number of point stored in each item