import sys, os
from tqdm import tqdm
import json
metalst = sys.argv[1]
wav_dir = sys.argv[2]
dataset_lst = sys.argv[3]
wav_res_ref_text = sys.argv[4]

f = open(metalst)
lines = f.readlines()
f.close()

datasets = []
with open(dataset_lst, "r") as f:
    for line in f:
        data = json.loads(line)
        datasets.append(os.path.join(wav_dir, '../', data['audio']))
f_w = open(wav_res_ref_text, 'w')
for idx, line in tqdm(enumerate(lines)):
    assert len(line.strip().split('||')) == 2
    utt, infer_text = line.strip().split('||')
    infer_text = infer_text.replace("|", "")
    if utt.endswith(".wav"):
        utt = utt[:-4]
    prompt_wav = datasets[idx]
    if not os.path.exists(os.path.join(wav_dir, utt + '.wav')):
        continue

    if not os.path.isabs(prompt_wav):
        prompt_wav = os.path.join(os.path.dirname(metalst), prompt_wav)


    out_line = '|'.join([os.path.join(wav_dir, utt + '.wav'), prompt_wav, infer_text])
    f_w.write(out_line + '\n')
f_w.close()