"""

Load the 3-omics and perform subtype detecion from the HCC dataset

tsv files used in the original study are available in the ./data folder of this project.

However, theses files must be decompressed using this function in linux:

gzip -d *.gz.

"""

# Python import needed

from simdeep.simdeep_boosting import SimDeepBoosting

from simdeep.config import PATH_THIS_FILE

from collections import OrderedDict

from os.path import isfile

from sys import exit

def main():

    """ Main function excecuted """

    path_data = PATH_THIS_FILE + "/../data/"

    # Testing if the files were decompressed in the good repository

    try:

        assert(isfile(path_data + "meth.tsv"))

        assert(isfile(path_data + "rna.tsv"))

        assert(isfile(path_data + "mir.tsv"))

    except AssertionError:

        print('gz files in {0} must be decompressed !\n exiting...'.format(path_data))

        exit(1)

    # Tsv files used in the original study in the appropriate order

    tsv_files = OrderedDict([

        ('MIR', 'mir.tsv'),

        ('METH', 'meth.tsv'),

        ('RNA', 'rna.tsv'),

    ])

    # File with survival event

    survival_tsv = 'survival.tsv'

    # As test dataset we will use the rna.tsv only

    tsv_test = {'RNA': 'rna.tsv'}

    # because it is the same data, we should use the same survival file

    test_survival = 'survival.tsv'

    PROJECT_NAME = 'HCC_dataset'

    EPOCHS = 10

    SEED = 10045

    nb_it = 3

    nb_threads = 2

    survival_flag = {

        'patient_id': 'Samples',

        'survival': 'days',

        'event': 'event'}

    import ray

    ray.init(num_cpus=3)

    normalization = {

        'NB_FEATURES_TO_KEEP': 100, # variance selection features. 0 is all the feature

        'TRAIN_RANK_NORM': True,

        'TRAIN_CORR_REDUCTION': True,

        'TRAIN_CORR_RANK_NORM': True,

        'TRAIN_ROBUST_SCALE': False,

    }

    # Instanciate a DeepProg instance

    boosting = SimDeepBoosting(

        nb_threads=nb_threads,

        nb_it=nb_it,

        split_n_fold=3,

        survival_tsv=survival_tsv,

        training_tsv=tsv_files,

        path_data=path_data,

        project_name=PROJECT_NAME,

        path_results=path_data,

        epochs=EPOCHS,

        survival_flag=survival_flag,

        distribute=True,

        cluster_method="mixture",

        use_autoencoders=True,

        feature_surv_analysis=True,

        normalization=normalization,

        seed=SEED)

    boosting.fit()

    # predict labels of the training

    boosting.predict_labels_on_full_dataset()

    boosting.compute_clusters_consistency_for_full_labels()

    boosting.evalutate_cluster_performance()

    boosting.collect_cindex_for_test_fold()

    boosting.collect_cindex_for_full_dataset()

    boosting.compute_pvalue_for_merged_test_fold()

    boosting.compute_feature_scores_per_cluster()

    boosting.write_feature_score_per_cluster()

    # Finally, load test set

    boosting.load_new_test_dataset(

        tsv_test,

        'test_RNA_only',

        test_survival,

    )

    boosting.predict_labels_on_test_dataset()

    boosting.compute_c_indexes_for_test_dataset()

    boosting.compute_clusters_consistency_for_test_labels()

    # Experimental method to plot the test dataset amongst the class kernel densities

    boosting.plot_supervised_kernel_for_test_sets()

    boosting.plot_supervised_predicted_labels_for_test_sets()

    #All the parameters are attributes of the SimDeep instance:

    # boosting.labels

    # boosting.test_labels

    # boosting.test_labels_proba

    # ... etc...

    # Close clusters and free memory

    ray.shutdown()

if __name__ == "__main__":

    main()