Source code for tmtk.clinical.Variable

from ..utils import Mappings, path_converter, ReservedKeywordException, is_not_a_value

import pandas as pd

class VarID:
    Clinical variable identifier. Contains logic to convert to string for
    jstree json.

    def __new__(cls, *args, **kwargs):
        if len(args) == 1:
            # highdim or tags
            if (len(args[0]) == 32 and '_' not in args[0]) or 'tags_id_' in args[0]:
                return args[0]

        return super(VarID, cls).__new__(cls)

    def __init__(self, *args):

        if len(args) == 1 and '__' in args[0]:
            l = args[0].rsplit('__', 1)
            if '_' in l[1]:
                l += l.pop(1).split('_')
            args = l

        elif len(args) == 1 and type(args[0]) == tuple:
            args = args[0]

        self.filename = args[0]
        self.column = args[1]
        self.category = args[2] if len(args) > 2 else None

    def __eq__(self, other):
        return hash(self) == hash(other)

    def __hash__(self):
        return hash(self.tuple)

    def __repr__(self):
        if self.category:
            return 'VarID({!r}, {}, {})'.format(*self.tuple)
            return 'VarID({!r}, {})'.format(*self.tuple)

    def __str__(self):
        if self.category:
            return '{}__{}_{}'.format(*self.tuple)
            return '{}__{}'.format(*self.tuple)

    def __getitem__(self, key):
        return self.tuple[key]

    def __iter__(self):
        for item in self.filename, self.column, self.category:
            if item:
                yield item

    def tuple(self):
        return tuple(self)

    def parent(self):
        return self.filename, self.column

    def create_category(self, i):
        Create a category ID and return it.

        :param i: Integer
        :return: new VarID
        assert not self.category, 'VarID already has category.'
        return VarID(self.filename, self.column, int(i))

[docs]class Variable: """ Base class for clinical variables """ VIS_DATE = 'LAD' VIS_TEXT = 'LAT' VIS_NUMERIC = 'LAN' VIS_CATEGORICAL = 'LAC' def __init__(self, datafile, column: int = None, clinical_parent=None): self.datafile = datafile self.filename = self.column = column self._zero_column = column - 1 self.parent = clinical_parent def __repr__(self): return '{} {!r}: {}'.format(self.__class__.__name__, self.var_id, self.concept_path) @property def values(self): """ :return: All values as found in the datafile. """ return self.datafile.df.iloc[:, self._zero_column] @values.setter def values(self, series: pd.Series): self.datafile.df.iloc[:, self._zero_column] = series @property def unique_values(self): """ :return: Unique set of values in the datafile. """ return self.values.unique() @property def var_id(self): """ :return: Variable identifier tuple (, column). """ return VarID(, self.column) @property def is_numeric_in_datafile(self): """ True if the datafile contains only numerical items. :return: bool. """ try: set(map(float, self.values)) return True except (ValueError, TypeError): return False @property def min(self): if self.is_numeric_in_datafile: return min(set(map(float, self.values))) @property def max(self): if self.is_numeric_in_datafile: return max(set(map(float, self.values))) @property def is_numeric(self): """ True if transmart-batch will load this concept as numerical. This includes information from word mapping and column mapping. :return: bool. """ if self.forced_categorical: return False if not self.is_in_wordmap: return self.is_numeric_in_datafile else: try: set(map(float, self.mapped_values)) return True except (ValueError, TypeError): return False @property def is_empty(self): """ Check if variable is fully empty. :return: bool. """ return self.values.apply(is_not_a_value).all() @property def concept_path(self): """ Concept path after conversions by transmart-batch. Combination of self.category_code and self.data_label. Cannot be set. :return: str. """ cp = self.parent.ColumnMapping.get_concept_path(self.var_id) return path_converter(cp) @property def category_code(self): """ The second column of the column mapping file for this variable. This combines with self.data_label to create self.concept_path. :return: str. """ return self.parent.ColumnMapping.select_row(self.var_id)[1] @category_code.setter def category_code(self, value): self.parent.ColumnMapping.set_concept_path(self.var_id, path=value) @property def column_map_data(self): """Column mapping row as dictionary where keys are short descriptors. :return: dict. """ row = self.parent.ColumnMapping.select_row(self.var_id) data_args = {} for i, s in enumerate(Mappings.column_mapping_s): data_args.update({s: row[i] if len(row) > i else None}) return data_args @property def data_label(self): """ Variable data label. :return: str. """ return self.parent.ColumnMapping.select_row(self.var_id)[3] @data_label.setter def data_label(self, value): self.parent.ColumnMapping.set_concept_path(self.var_id, label=value) @property def word_map_dict(self): """ A dictionary with word mapped categoricals. Keys are items in the datafile, values are what they will be mapped to through the word mapping file. Unmapped items are also added as key, value pair. :return: dict. """ values = set(self.values) d = dict(zip(values, values)) d.update(self.parent.WordMapping.get_word_map(self.var_id)) return d @word_map_dict.setter def word_map_dict(self, d): self.parent.WordMapping.set_word_map(self.var_id, d) @property def mapped_values(self): """ Data items after word mapping. :return: list. """ if self.is_in_wordmap: return else: return self.values @property def forced_categorical(self): """Check if forced categorical by entering 'CATEGORICAL' in data type column. Can be changed by setting this to True or False. :return: bool. """ return self.column_type == 'CATEGORICAL' @forced_categorical.setter def forced_categorical(self, value: bool): self.column_type = 'CATEGORICAL' if bool(value) else '' @property def is_in_wordmap(self): """ Check if variable is represented in word mapping file. :return: bool. """ return tuple(self.var_id) in self.parent.WordMapping.df.index
[docs] def word_mapped_not_present(self): """ Gets the values that are in the word map but not in the data file. :return: set. """ if not self.is_in_wordmap: return set() mapped_value_column = self.parent.WordMapping.df.columns[2] t_index = tuple(self.var_id) mapped_values = self.parent.WordMapping.df.loc[t_index, mapped_value_column] if type(mapped_values) != str: mapped_values = set(mapped_values) else: mapped_values = {mapped_values} return mapped_values - set(self.values)
@property def header(self): return self.datafile.df.columns[self._zero_column] @property def visual_attributes(self): if self.column_type == 'DATE': return self.VIS_DATE elif self.column_type == 'TEXT': return self.VIS_TEXT elif self.is_numeric: return self.VIS_NUMERIC else: return self.VIS_CATEGORICAL @property def reference_column(self): return self.parent.ColumnMapping.select_row(self.var_id)[4] @reference_column.setter def reference_column(self, value): self.parent.ColumnMapping.set_reference_column(self.var_id, value) @property def concept_code(self): return self.parent.ColumnMapping.select_row(self.var_id)[5] @concept_code.setter def concept_code(self, value): self.parent.ColumnMapping.set_concept_code(self.var_id, value) @property def column_type(self): """ Column data type setting can be found in modifiers file for MODIFIER vars, else it is in the DataType column of column mapping. If it is not found, it will be either numerical or categorical based on the datafile values. """ if self.data_label == 'MODIFIER': return self.parent.Modifiers.df.loc[self.modifier_code, self.parent.Modifiers.df.columns[3]] else: try: return self.parent.ColumnMapping.select_row(self.var_id)[6] except IndexError: return None @column_type.setter def column_type(self, value): self.parent.ColumnMapping.set_column_type(self.var_id, value) @property def modifier_code(self): """ Requires implementation, always returns '@'.""" return self.parent.ColumnMapping.select_row(self.var_id)[6] if self.data_label == 'MODIFIER' else '@' def _get_all(self, label: str): """ Will look for keyword variables in the same data file based a label and check whether it applies to self. The reference column can be used to specify to which columns a keyword applies. Providing a comma separated list of column indices is supported. :param str label: data label. :return list: a list of variables. """ vars_ = self.parent.find_variables_by_label(label, self.var_id.filename) inclusion_criteria = (None,, '') return [var for var in vars_ if var.reference_column in inclusion_criteria or str(self.column) in str(var.reference_column).split(',')] def _get_one_or_none(self, label: str): """ Will look for a keyword variable that applies to self. Will raise ReservedKeywordException if more than 1 is found. :param str label: data label. :return: variable. """ vars_ = self._get_all(label) if len(vars_) > 2: raise ReservedKeywordException('Multiple {} found for {}'.format(label, self)) elif vars_: return vars_[0] @property def subj_id(self): subj_id = self._get_one_or_none('SUBJ_ID') if subj_id: return subj_id else: raise ReservedKeywordException('No SUBJ_ID found for {}'.format(self)) @property def start_date(self): return self._get_one_or_none('START_DATE') @property def end_date(self): return self._get_one_or_none('END_DATE') @property def trial_visit(self): return self._get_one_or_none('TRIAL_VISIT_LABEL') @property def modifiers(self): """ Returns a list of all modifier variable that apply to this variable. The data label for these variables have to be 'MODIFIER' and the fifth column (reference column) has to either be empty or the column this variable has. :return: list of modifier variables. """ return self._get_all('MODIFIER')