baecce5a6fbe4576ca8348a93dc56f07e7ba84f6,niftynet/contrib/csv_reader/csv_reader.py,CSVReader,_parse_csv,#CSVReader#Any#Any#,72

Before Change

        df.index = df["subject_id"]
        assert set(df.index) == set(self.subject_ids)
        df = df.loc[self.subject_ids]
        if to_ohe and len(df.columns[1:])==1:
            _dims = len(list(df[df.columns[1]].unique()))
            _indexable_output = self.to_ohe(df["label"].values, _dims)
            return df, _indexable_output, _dims

After Change

    def _parse_csv(self, path_to_csv, to_ohe):
        tf.logging.warning("This method will read your entire csv into memory")
        df = pd.read_csv(path_to_csv, index_col=0, header=None)
        if set(df.index) != set(self.subject_ids):
            print(set(self.subject_ids) - set(df.index))
            tf.logging.fatal("csv file provided at: {} does not have all the subject_ids".format(path_to_csv))
            raise Exception
        if to_ohe and len(df.columns)==1:
            _dims = len(list(df[1].unique()))
            _indexable_output = self.to_ohe(df[1].values, _dims)
            return df, _indexable_output, _dims
        elif not to_ohe and len(df.columns==1):
            _dims = 1
            _indexable_output = self.to_categorical(df[1].values, df[1].unique())
            return df, _indexable_output, _dims
        elif not to_ohe:
            _dims = len(df.columns)
            _indexable_output = list(df.values)

Instances