import pandas as pd
from rich import print
from datasets import DatasetDict, Value, ClassLabel, Dataset,  Features


def load_dataset(input_url, label_url):
    input_df = pd.read_csv(input_url)
    label_df = pd.read_csv(label_url, header=None)

    input_df.columns = ['id', 'context', 'answerA', 'answerB', 'answerC']
    label_df.columns = ['id', 'label']

    features = Features({
        'context':  Value('string'), 
        'answerA':  Value('string'), 
        'answerB':  Value('string'), 
        'answerC':  Value('string'), 
        'label':  ClassLabel(num_classes=3, names=['A', 'B', 'C']), 
    })
    dataset = Dataset.from_dict({
        'context': input_df['context'], 
        'answerA': input_df['answerA'], 
        'answerB': input_df['answerB'], 
        'answerC': input_df['answerC'], 
        'label': label_df['label']
    }, features=features)
    return dataset


input_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Training%20%20Data/subtaskB_data_all.csv'
label_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Training%20%20Data/subtaskB_answers_all.csv'
train_dataset = load_dataset(input_url, label_url)

input_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Dev%20Data/subtaskB_dev_data.csv'
label_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Dev%20Data/subtaskB_gold_answers.csv'
validation_dataset = load_dataset(input_url, label_url)

input_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Test%20Data/subtaskB_test_data.csv'
label_url = 'https://raw.githubusercontent.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation/master/ALL%20data/Test%20Data/subtaskB_gold_answers.csv'
test_dataset = load_dataset(input_url, label_url)

dataset = DatasetDict({
    'train': train_dataset, 
    'validation': validation_dataset, 
    'test': test_dataset
})
print(dataset)
dataset.push_to_hub('extraordinarylab/comve_taskb')