# Tests the following end-to-end:
#
# 1. Comet is imported
# 2. Conflicting modules (i.e., TensorFlow) are not imported
# 3. Overridden methods are called (train_init, train_model, etc.) and run without error
#
# This test runs in an isolated environment to ensure TensorFlow imports are not leaked
# from previous tests.

import argparse
import os
import sys
import tempfile
from unittest.mock import Mock, patch

# Comet must be imported before the libraries it wraps
import comet_ml  # noqa

from ludwig.api import LudwigModel
from ludwig.constants import BATCH_SIZE, TRAINER
from ludwig.contribs.comet import CometCallback

# Bad key will ensure Comet is initialized, but nothing is uploaded externally.
os.environ["COMET_API_KEY"] = "key"

# Add tests dir to the import path
PATH_HERE = os.path.abspath(os.path.dirname(__file__))
PATH_ROOT = os.path.join(PATH_HERE, "..", "..", "..")
sys.path.insert(0, os.path.abspath(PATH_ROOT))

from tests.integration_tests.utils import category_feature, generate_data, image_feature  # noqa

parser = argparse.ArgumentParser()
parser.add_argument("--csv-filename", required=True)


def run(csv_filename):
    with tempfile.TemporaryDirectory() as tmpdir:
        # Image Inputs
        image_dest_folder = os.path.join(tmpdir, "generated_images")

        # Inputs & Outputs
        input_features = [image_feature(folder=image_dest_folder)]
        output_features = [category_feature(output_feature=True)]
        data_csv = generate_data(input_features, output_features, csv_filename)

        config = {
            "input_features": input_features,
            "output_features": output_features,
            "combiner": {"type": "concat", "output_size": 14},
            TRAINER: {"epochs": 2, BATCH_SIZE: 128},
        }

        callback = CometCallback()
        model = LudwigModel(config, callbacks=[callback])

        # Wrap these methods so we can check that they were called
        callback.on_train_init = Mock(side_effect=callback.on_train_init)
        callback.on_train_start = Mock(side_effect=callback.on_train_start)

        with patch("comet_ml.Experiment.log_asset_data") as mock_log_asset_data:
            # Training with csv
            _, _, _ = model.train(dataset=data_csv, output_directory=os.path.join(tmpdir, "output"))
            model.predict(dataset=data_csv)

    # Verify that the experiment was created successfully
    assert callback.cometml_experiment is not None

    # Check that these methods were called at least once
    callback.on_train_init.assert_called()
    callback.on_train_start.assert_called()

    # Check that we ran `train_model`, which calls into `log_assert_data`, successfully
    mock_log_asset_data.assert_called()


if __name__ == "__main__":
    args = parser.parse_args()
    run(args.csv_filename)