import os 
df = pd.read_excel("Annotation_CARIMAM_apo_2023_07_06.xlsx", 'annot_click_apo', usecols='A:E').dropna()
df = df[~df['File'].str.startswith('/nfs/NAS4/')]
df = df[~df['File'].str.startswith('/nfs/NAS3/')]
mask = ((df['File'].str.startswith('LOT2/BERMUDE'))|(df.File.str.startswith('LOT2/GUYANNE')) | (df.File.str.startswith('LOT2/ANG')))
df_test = df[mask]
df_train = df[~mask]

#import ipdb; ipdb.set_trace()
for idx, row in df_test.iterrows():
    file = row['File']
    label = int(row['positif_negatif'] != 'n')
    file = file if file.startswith('/nfs') else os.path.join("/nfs/NAS6/SABIOD/SITE/CARIMAM/DATA/", file)
    click_start, click_end = row[['pos_start', 'pos_end']]
    pos_click = click_start if click_start == click_end else np.random.uniform(click_start, click_end)
    pos_click = int(sf.info(file).samplerate * pos_click)
    nb_ctxt_frame = 512
    click_margin = nb_ctxt_frame * 2 * int(sf.info(file).samplerate * 0.001)
    sig, sr = sf.read(file, start=max(0, pos_click-click_margin), stop=pos_click+click_margin)
    if (pos_click-click_margin) < 0:
    	sig = np.pad(sig, [np.abs(pos_click-click_margin), 0])
    else (len(sig))
    print(len(sig)/sr)
    os.makedirs(os.path.dirname(("audio_test_article/%s")%row['File']), exist_ok=True)
    sf.write(("audio_test_article/%s-%s-%s.WAV")%(row['File'],click_start, label), sig, sr)
df_test.to_csv("df_test.csv")