small code for csv2ffm

map_col = lambda dat,col: col+"-"+dat.map(str)
gen_hash_item = lambda field, feat: '{0}:{1}:1'.format(field,hashstr(feat))
def gen_hash_row(feats,label):
    result = []
    for idx, item in enumerate(feats):
        val = item.split('-')[-1]
        if val != 'nan':
                result.append(gen_hash_item(idx,item))
    lbl = 1
    if label == 0:
        lbl = -1
    return str(lbl) + ' ' + ' '.join(result)+'\n'

merge_dat_val = merge_dat.drop(['display_id','clicked'],axis=1)
cols = merge_dat_val.columns
features = []
for col in merge_dat_val.columns:
    features.append(map_col(merge_dat_val[col],col))
features = np.array(features).T
with open(output_name,'w') as f_tr:
    i = 0;
    for item,label in zip(features,targets):
        if(i%200000==0):
            print i
        row = gen_hash_row(item,label)
        f_tr.write(row)
        i+=1