Merge pull request #7153 from janezd/owsplit

Split: Move widget from Prototypes

Author: ajdapretnar

Orange/widgets/data/icons/Split.svg

@@ -0,0 +1,241 @@
+from functools import partial
+
+import numpy as np
+
+from AnyQt.QtCore import Qt
+
+from orangewidget.settings import Setting
+
+from Orange.widgets import gui
+from Orange.widgets.settings import ContextSetting, DomainContextHandler
+from Orange.widgets.widget import OWWidget, Msg, Output, Input
+from Orange.widgets.utils.itemmodels import DomainModel
+from Orange.widgets.utils.widgetpreview import WidgetPreview
+from Orange.data import \
+    Table, Domain, DiscreteVariable, StringVariable, ContinuousVariable
+from Orange.data.util import SharedComputeValue, get_unique_names
+
+
+def get_substrings(values, delimiter):
+    return sorted({ss.strip() for s in values for ss in s.split(delimiter)}
+                  - {""})
+
+
+class SplitColumnBase:
+    def __init__(self, data, attr, delimiter):
+        self.attr = attr
+        self.delimiter = delimiter
+        column = set(data.get_column(self.attr))
+        self.new_values = tuple(get_substrings(column, self.delimiter))
+
+    def __eq__(self, other):
+        return self.attr == other.attr \
+               and self.delimiter == other.delimiter \
+               and self.new_values == other.new_values
+
+    def __hash__(self):
+        return hash((self.attr, self.delimiter, self.new_values))
+
+
+class SplitColumnOneHot(SplitColumnBase):
+    InheritEq = True
+
+    def __call__(self, data):
+        column = data.get_column(self.attr)
+        values = [{ss.strip() for ss in s.split(self.delimiter)}
+                  for s in column]
+        return {v: np.array([i for i, xs in enumerate(values) if v in xs],
+                            dtype=int)
+                for v in self.new_values}
+
+
+class SplitColumnCounts(SplitColumnBase):
+    InheritEq = True
+
+    def __call__(self, data):
+        column = data.get_column(self.attr)
+        values = [[ss.strip() for ss in s.split(self.delimiter)]
+                  for s in column]
+        return {v: np.array([xs.count(v) for xs in values], dtype=float)
+                for v in self.new_values}
+
+
+class StringEncodingBase(SharedComputeValue):
+    def __init__(self, fn, new_feature):
+        super().__init__(fn)
+        self.new_feature = new_feature
+
+    def __eq__(self, other):
+        return super().__eq__(other) and self.new_feature == other.new_feature
+
+    def __hash__(self):
+        return super().__hash__() ^ hash(self.new_feature)
+
+    def compute(self, data, shared_data):
+        raise NotImplementedError  # silence pylint
+
+class OneHotStrings(StringEncodingBase):
+    InheritEq = True
+
+    def compute(self, data, shared_data):
+        indices = shared_data[self.new_feature]
+        col = np.zeros(len(data))
+        col[indices] = 1
+        return col
+
+
+class CountStrings(StringEncodingBase):
+    InheritEq = True
+
+    def compute(self, data, shared_data):
+        return shared_data[self.new_feature]
+
+
+class DiscreteEncoding:
+    def __init__(self, variable, delimiter, onehot, value):
+        self.variable = variable
+        self.delimiter = delimiter
+        self.onehot = onehot
+        self.value = value
+
+    def __call__(self, data):
+        column = data.get_column(self.variable).astype(float)
+        col = np.zeros(len(column))
+        col[np.isnan(column)] = np.nan
+        for val_idx, value in enumerate(self.variable.values):
+            parts = value.split(self.delimiter)
+            if self.onehot:
+                col[column == val_idx] = int(self.value in parts)
+            else:
+                col[column == val_idx] = parts.count(self.value)
+        return col
+
+    def __eq__(self, other):
+        return self.variable == other.variable \
+               and self.value == other.value \
+               and self.delimiter == other.delimiter \
+               and self.onehot == other.onehot
+
+    def __hash__(self):
+        return hash((self.variable, self.value, self.delimiter, self.onehot))
+
+
+class OWSplit(OWWidget):
+    name = "Split"
+    description = "Split text or categorical variables into indicator variables"
+    category = "Transform"
+    icon = "icons/Split.svg"
+    keywords = ["text to columns", "word encoding", "questionnaire", "survey",
+                "term", "word presence", "word counts", "categorical encoding",
+                "indicator variables"]
+    priority = 700
+
+    class Inputs:
+        data = Input("Data", Table)
+
+    class Outputs:
+        data = Output("Data", Table)
+
+    class Warning(OWWidget.Warning):
+        no_disc = Msg("Data contains only numeric variables.")
+
+    want_main_area = False
+    resizing_enabled = False
+
+    Categorical, Numerical, Counts = range(3)
+    OutputLabels = ("Categorical (No, Yes)", "Numerical (0, 1)", "Counts")
+
+    settingsHandler = DomainContextHandler()
+    attribute = ContextSetting(None)
+    delimiter = ContextSetting(";")
+    output_type = ContextSetting(Categorical)
+    auto_apply = Setting(True)
+
+    def __init__(self):
+        super().__init__()
+        self.data = None
+
+        variable_select_box = gui.vBox(self.controlArea, "Variable")
+
+        gui.comboBox(variable_select_box, self, "attribute",
+                     orientation=Qt.Horizontal, searchable=True,
+                     callback=self.apply.deferred,
+                     model=DomainModel(valid_types=(StringVariable,
+                                                    DiscreteVariable)))
+        le = gui.lineEdit(
+                variable_select_box, self, "delimiter", "Delimiter: ",
+                orientation=Qt.Horizontal, callback=self.apply.deferred,
+                controlWidth=20)
+        le.box.layout().addStretch(1)
+        le.setAlignment(Qt.AlignCenter)
+
+        gui.radioButtonsInBox(
+            self.controlArea, self, "output_type", self.OutputLabels,
+            box="Output Values",
+            callback=self.apply.deferred)
+
+        gui.auto_apply(self.buttonsArea, self, commit=self.apply)
+
+    @Inputs.data
+    def set_data(self, data):
+        self.closeContext()
+        self.data = data
+
+        model = self.controls.attribute.model()
+        model.set_domain(data.domain if data is not None else None)
+        self.Warning.no_disc(shown=data is not None and not model)
+        if not model:
+            self.attribute = None
+            self.data = None
+            return
+        self.attribute = model[0]
+        self.openContext(data)
+        self.apply.now()
+
+    @gui.deferred
+    def apply(self):
+        if self.attribute is None:
+            self.Outputs.data.send(None)
+            return
+        var = self.data.domain[self.attribute]
+        values, computer = self._get_compute_value(var)
+        new_columns = self._get_new_columns(values, computer)
+        new_domain = Domain(
+            self.data.domain.attributes + new_columns,
+            self.data.domain.class_vars, self.data.domain.metas
+        )
+        extended_data = self.data.transform(new_domain)
+        self.Outputs.data.send(extended_data)
+
+    def _get_compute_value(self, var):
+        if var.is_discrete:
+            values = get_substrings(var.values, self.delimiter)
+            computer = partial(
+                DiscreteEncoding,
+                var, self.delimiter, self.output_type != self.Counts)
+        else:
+            if self.output_type == self.Counts:
+                sc = SplitColumnCounts(self.data, var, self.delimiter)
+                computer = partial(CountStrings, sc)
+            else:
+                sc = SplitColumnOneHot(self.data, var, self.delimiter)
+                computer = partial(OneHotStrings, sc)
+            values = sc.new_values
+        return values, computer
+
+    def _get_new_columns(self, values, computer):
+        names = get_unique_names(self.data.domain, values, equal_numbers=False)
+        if self.output_type == self.Categorical:
+            return tuple(
+                DiscreteVariable(
+                    name, ("No", "Yes"), compute_value=computer(value))
+                for value, name in zip(values, names))
+        else:
+            return tuple(
+                ContinuousVariable(
+                    name, compute_value=computer(value))
+                for value, name in zip(values, names))
+
+
+if __name__ == "__main__":  # pragma: no cover
+    WidgetPreview(OWSplit).run(Table.from_file("tests/orange-in-education.tab"))