def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.__scoreidx = 0
self.__strategy = UnivariateFeatureSelect.Fixed
self.__k = 10
self.__p = 75.0
box = QGroupBox(title="Score", flat=True)
box.setLayout(QVBoxLayout())
self.__cb = cb = QComboBox(self, )
self.__cb.currentIndexChanged.connect(self.setScoreIndex)
self.__cb.activated.connect(self.edited)
box.layout().addWidget(cb)
self.layout().addWidget(box)
box = QGroupBox(title="Strategy", flat=True)
self.__group = group = QButtonGroup(self, exclusive=True)
self.__spins = {}
form = QFormLayout()
fixedrb = QRadioButton("Fixed:", checked=True)
group.addButton(fixedrb, UnivariateFeatureSelect.Fixed)
kspin = QSpinBox(
minimum=1,
value=self.__k,
enabled=self.__strategy == UnivariateFeatureSelect.Fixed)
kspin.valueChanged[int].connect(self.setK)
kspin.editingFinished.connect(self.edited)
self.__spins[UnivariateFeatureSelect.Fixed] = kspin
form.addRow(fixedrb, kspin)
percrb = QRadioButton("Percentile:")
group.addButton(percrb, UnivariateFeatureSelect.Percentile)
pspin = QDoubleSpinBox(
minimum=0.0,
maximum=100.0,
singleStep=0.5,
value=self.__p,
suffix="%",
enabled=self.__strategy == UnivariateFeatureSelect.Percentile)
pspin.valueChanged[float].connect(self.setP)
pspin.editingFinished.connect(self.edited)
self.__spins[UnivariateFeatureSelect.Percentile] = pspin
# Percentile controls disabled for now.
pspin.setEnabled(False)
percrb.setEnabled(False)
form.addRow(percrb, pspin)
# form.addRow(QRadioButton("FDR"), QDoubleSpinBox())
# form.addRow(QRadioButton("FPR"), QDoubleSpinBox())
# form.addRow(QRadioButton("FWE"), QDoubleSpinBox())
self.__group.buttonClicked.connect(self.__on_buttonClicked)
box.setLayout(form)
self.layout().addWidget(box)
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.__scoreidx = 0
self.__strategy = UnivariateFeatureSelect.Fixed
self.__k = 10
self.__p = 75.0
box = QGroupBox(title="Score", flat=True)
box.setLayout(QVBoxLayout())
self.__cb = cb = QComboBox(self, )
self.__cb.currentIndexChanged.connect(self.setScoreIndex)
self.__cb.activated.connect(self.edited)
box.layout().addWidget(cb)
self.layout().addWidget(box)
box = QGroupBox(title="Strategy", flat=True)
self.__group = group = QButtonGroup(self, exclusive=True)
self.__spins = {}
form = QFormLayout()
fixedrb = QRadioButton("Fixed:", checked=True)
group.addButton(fixedrb, UnivariateFeatureSelect.Fixed)
kspin = QSpinBox(
minimum=1, value=self.__k,
enabled=self.__strategy == UnivariateFeatureSelect.Fixed
)
kspin.valueChanged[int].connect(self.setK)
kspin.editingFinished.connect(self.edited)
self.__spins[UnivariateFeatureSelect.Fixed] = kspin
form.addRow(fixedrb, kspin)
percrb = QRadioButton("Percentile:")
group.addButton(percrb, UnivariateFeatureSelect.Percentile)
pspin = QDoubleSpinBox(
minimum=0.0, maximum=100.0, singleStep=0.5,
value=self.__p, suffix="%",
enabled=self.__strategy == UnivariateFeatureSelect.Percentile
)
pspin.valueChanged[float].connect(self.setP)
pspin.editingFinished.connect(self.edited)
self.__spins[UnivariateFeatureSelect.Percentile] = pspin
# Percentile controls disabled for now.
pspin.setEnabled(False)
percrb.setEnabled(False)
form.addRow(percrb, pspin)
# form.addRow(QRadioButton("FDR"), QDoubleSpinBox())
# form.addRow(QRadioButton("FPR"), QDoubleSpinBox())
# form.addRow(QRadioButton("FWE"), QDoubleSpinBox())
self.__group.buttonClicked.connect(self.__on_buttonClicked)
box.setLayout(form)
self.layout().addWidget(box)
class CUR(BaseEditor):
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.rank = 10
self.max_error = 1
form = QFormLayout()
self.rspin = QSpinBox(minimum=2, value=self.rank)
self.rspin.valueChanged[int].connect(self.setR)
self.rspin.editingFinished.connect(self.edited)
self.espin = QDoubleSpinBox(
minimum=0.1, maximum=100.0, singleStep=0.1,
value=self.max_error)
self.espin.valueChanged[float].connect(self.setE)
self.espin.editingFinished.connect(self.edited)
form.addRow("Rank:", self.rspin)
form.addRow("Relative error:", self.espin)
self.layout().addLayout(form)
def setParameters(self, params):
self.setR(params.get("rank", 10))
self.setE(params.get("max_error", 1))
def parameters(self):
return {"rank": self.rank, "max_error": self.max_error}
def setR(self, rank):
if self.rank != rank:
self.rank = rank
self.rspin.setValue(rank)
self.changed.emit()
def setE(self, max_error):
if self.max_error != max_error:
self.max_error = max_error
self.espin.setValue(max_error)
self.changed.emit()
@staticmethod
def createinstance(params):
rank = params.get("rank", 10)
max_error = params.get("max_error", 1)
return ProjectCUR(rank=rank, max_error=max_error)
def __repr__(self):
return "Rank: {}, Relative error: {}".format(self.rspin.value(),
self.espin.value())
class CUR(BaseEditor):
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.rank = 10
self.max_error = 1
form = QFormLayout()
self.rspin = QSpinBox(minimum=2, maximum=1000000, value=self.rank)
self.rspin.valueChanged[int].connect(self.setR)
self.rspin.editingFinished.connect(self.edited)
self.espin = QDoubleSpinBox(minimum=0.1,
maximum=100.0,
singleStep=0.1,
value=self.max_error)
self.espin.valueChanged[float].connect(self.setE)
self.espin.editingFinished.connect(self.edited)
form.addRow("Rank:", self.rspin)
form.addRow("Relative error:", self.espin)
self.layout().addLayout(form)
def setParameters(self, params):
self.setR(params.get("rank", 10))
self.setE(params.get("max_error", 1))
def parameters(self):
return {"rank": self.rank, "max_error": self.max_error}
def setR(self, rank):
if self.rank != rank:
self.rank = rank
self.rspin.setValue(rank)
self.changed.emit()
def setE(self, max_error):
if self.max_error != max_error:
self.max_error = max_error
self.espin.setValue(max_error)
self.changed.emit()
@staticmethod
def createinstance(params):
rank = params.get("rank", 10)
max_error = params.get("max_error", 1)
return ProjectCUR(rank=rank, max_error=max_error)
def __repr__(self):
return "Rank: {}, Relative error: {}".format(self.rspin.value(),
self.espin.value())
class BinarizeEditor(ScBaseEditor):
DEFAULT_CONDITION = Binarize.GreaterOrEqual
DEFAULT_THRESHOLD = 1
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self._threshold = self.DEFAULT_THRESHOLD
self.setLayout(QVBoxLayout())
form = QFormLayout()
self.cond_cb = QComboBox()
self.cond_cb.addItems(["Greater or Equal", "Greater"])
self.cond_cb.currentIndexChanged.connect(self.changed)
self.cond_cb.activated.connect(self.edited)
self.thr_spin = QDoubleSpinBox(minimum=0,
singleStep=0.5,
decimals=1,
value=self._threshold)
self.thr_spin.valueChanged[float].connect(self._set_threshold)
self.thr_spin.editingFinished.connect(self.edited)
form.addRow("Condition:", self.cond_cb)
form.addRow("Threshold:", self.thr_spin)
self.layout().addLayout(form)
def _set_threshold(self, t):
if self._threshold != t:
self._threshold = t
self.thr_spin.setValue(t)
self.changed.emit()
def setParameters(self, params):
cond = params.get("condition", self.DEFAULT_CONDITION)
self.cond_cb.setCurrentIndex(enum_to_index(Binarize.Condition, cond))
self._set_threshold(params.get("threshold", self.DEFAULT_THRESHOLD))
def parameters(self):
cond = index_to_enum(Binarize.Condition, self.cond_cb.currentIndex())
return {"condition": cond, "threshold": self._threshold}
@staticmethod
def createinstance(params):
condition = params.get("condition", BinarizeEditor.DEFAULT_CONDITION)
threshold = params.get("threshold", BinarizeEditor.DEFAULT_THRESHOLD)
return Binarize(condition, threshold)
def __repr__(self):
return "Condition: {}, Threshold: {}".format(
self.cond_cb.currentText(), self.thr_spin.value())
def __init__(self, *args, **kwargs):
super().__init__(*args)
layout = QStackedLayout(self)
self.double_spin = QDoubleSpinBox()
self.double_spin.valueChanged.connect(self.double_value_changed)
self.double_spin.editingFinished.connect(self.double_editing_finished)
layout.addWidget(self.double_spin)
self.int_spin = QSpinBox()
self.int_spin.setMaximum(10**4)
self.int_spin.valueChanged.connect(self.int_value_changed)
self.int_spin.editingFinished.connect(self.int_editing_finished)
layout.addWidget(self.int_spin)
self.setValue(kwargs.get('value', 0.))
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.rank = 10
self.max_error = 1
form = QFormLayout()
self.rspin = QSpinBox(minimum=2, maximum=1000000, value=self.rank)
self.rspin.valueChanged[int].connect(self.setR)
self.rspin.editingFinished.connect(self.edited)
self.espin = QDoubleSpinBox(
minimum=0.1, maximum=100.0, singleStep=0.1,
value=self.max_error)
self.espin.valueChanged[float].connect(self.setE)
self.espin.editingFinished.connect(self.edited)
form.addRow("Rank:", self.rspin)
form.addRow("Relative error:", self.espin)
self.layout().addLayout(form)
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self._threshold = self.DEFAULT_THRESHOLD
self.setLayout(QVBoxLayout())
form = QFormLayout()
self.cond_cb = QComboBox()
self.cond_cb.addItems(["Greater or Equal", "Greater"])
self.cond_cb.currentIndexChanged.connect(self.changed)
self.cond_cb.activated.connect(self.edited)
self.thr_spin = QDoubleSpinBox(minimum=0,
singleStep=0.5,
decimals=1,
value=self._threshold)
self.thr_spin.valueChanged[float].connect(self._set_threshold)
self.thr_spin.editingFinished.connect(self.edited)
form.addRow("Condition:", self.cond_cb)
form.addRow("Threshold:", self.thr_spin)
self.layout().addLayout(form)
def __init__(self, *args, **kwargs):
super().__init__(*args)
layout = QStackedLayout(self)
self.double_spin = QDoubleSpinBox()
self.double_spin.valueChanged.connect(self.double_value_changed)
self.double_spin.editingFinished.connect(self.double_editing_finished)
layout.addWidget(self.double_spin)
self.int_spin = QSpinBox()
self.int_spin.setMaximum(10 ** 4)
self.int_spin.valueChanged.connect(self.int_value_changed)
self.int_spin.editingFinished.connect(self.int_editing_finished)
layout.addWidget(self.int_spin)
self.setValue(kwargs.get('value', 0.))
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.__strategy = RandomFeatureSelectEditor.Fixed
self.__k = 10
self.__p = 75.0
box = QGroupBox(title="Strategy", flat=True)
self.__group = group = QButtonGroup(self, exclusive=True)
self.__spins = {}
form = QFormLayout()
fixedrb = QRadioButton("Fixed", checked=True)
group.addButton(fixedrb, RandomFeatureSelectEditor.Fixed)
kspin = QSpinBox(
minimum=1,
value=self.__k,
enabled=self.__strategy == RandomFeatureSelectEditor.Fixed,
)
kspin.valueChanged[int].connect(self.setK)
kspin.editingFinished.connect(self.edited)
self.__spins[RandomFeatureSelectEditor.Fixed] = kspin
form.addRow(fixedrb, kspin)
percrb = QRadioButton("Percentage")
group.addButton(percrb, RandomFeatureSelectEditor.Percentage)
pspin = QDoubleSpinBox(
minimum=0.0,
maximum=100.0,
singleStep=0.5,
value=self.__p,
suffix="%",
enabled=self.__strategy == RandomFeatureSelectEditor.Percentage,
)
pspin.valueChanged[float].connect(self.setP)
pspin.editingFinished.connect(self.edited)
self.__spins[RandomFeatureSelectEditor.Percentage] = pspin
form.addRow(percrb, pspin)
self.__group.buttonClicked.connect(self.__on_buttonClicked)
box.setLayout(form)
self.layout().addWidget(box)
def __init__(self, parent=None, **kwargs):
super().__init__(parent, **kwargs)
self.setLayout(QVBoxLayout())
self.rank = 10
self.max_error = 1
form = QFormLayout()
self.rspin = QSpinBox(minimum=2, value=self.rank)
self.rspin.valueChanged[int].connect(self.setR)
self.rspin.editingFinished.connect(self.edited)
self.espin = QDoubleSpinBox(
minimum=0.1, maximum=100.0, singleStep=0.1,
value=self.max_error)
self.espin.valueChanged[float].connect(self.setE)
self.espin.editingFinished.connect(self.edited)
form.addRow("Rank:", self.rspin)
form.addRow("Relative error:", self.espin)
self.layout().addLayout(form)
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
inputs = [("Data", Orange.data.Table, "set_data"),
("Learner", Learner, "set_learner")]
outputs = [("Data", Orange.data.Table)]
DEFAULT_LEARNER = SimpleTreeLearner()
METHODS = [AsDefault(), impute.DoNotImpute(), impute.Average(),
impute.AsValue(), impute.Model(DEFAULT_LEARNER), impute.Random(),
impute.DropInstances(), impute.Default()]
DEFAULT, DO_NOT_IMPUTE, MODEL_BASED_IMPUTER, AS_INPUT = 0, 1, 4, 7
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(DO_NOT_IMPUTE)
variable_methods = settings.ContextSetting({})
autocommit = settings.Setting(False)
default_value = settings.Setting(0.)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.METHODS):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.METHODS):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_combo.currentIndexChanged.connect(self._on_value_changed)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
value=self.default_value
)
self.value_stack = value_stack = QStackedLayout()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addLayout(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_methods, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal, checkbox_label="Apply automatically")
box.layout().insertSpacing(0, 80)
box.layout().insertWidget(0, self.report_button)
self.data = None
self.modified = False
self.default_method = self.METHODS[self.default_method_index]
self.update_varview()
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
self.default_method = self.METHODS[self.default_method_index]
self.METHODS[self.DEFAULT].method = self.default_method
# update variable view
for index in map(self.varmodel.index, range(len(self.varmodel))):
self.varmodel.setData(index,
self.variable_methods.get(index.row(), self.METHODS[self.DEFAULT]),
Qt.UserRole)
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@check_sql_input
def set_data(self, data):
self.closeContext()
self.varmodel[:] = []
self.variable_methods = {}
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
self.update_varview()
self.unconditional_commit()
def set_learner(self, learner):
self.learner = learner or self.DEFAULT_LEARNER
imputer = self.METHODS[self.MODEL_BASED_IMPUTER]
imputer.learner = self.learner
button = self.default_button_group.button(self.MODEL_BASED_IMPUTER)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(self.MODEL_BASED_IMPUTER)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = self.MODEL_BASED_IMPUTER
self.commit()
def get_method_for_column(self, column_index):
"""Returns the imputation method for column by its index.
"""
if not isinstance(column_index, int):
column_index = column_index.row()
return self.variable_methods.get(column_index,
self.METHODS[self.DEFAULT])
def _invalidate(self):
self.modified = True
self.commit()
def commit(self):
data = self.data
if self.data is not None:
if not len(self.data):
self.send("Data", self.data)
self.modified = False
return
drop_mask = np.zeros(len(self.data), bool)
attributes = []
class_vars = []
self.warning()
with self.progressBar(len(self.varmodel)) as progress:
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, self.default_method)
if not method.supports_variable(var):
self.warning("Default method can not handle '{}'".
format(var.name))
elif isinstance(method, impute.DropInstances):
drop_mask |= method(self.data, var)
else:
var = method(self.data, var)
if isinstance(var, Orange.data.Variable):
var = [var]
if i < len(self.data.domain.attributes):
attributes.extend(var)
else:
class_vars.extend(var)
progress.advance()
domain = Orange.data.Domain(attributes, class_vars,
self.data.domain.metas)
data = self.data.from_table(domain, self.data[~drop_mask])
self.send("Data", data)
self.modified = False
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, None)
if method is not None:
specific.append("{} ({})".format(var.name, str(method)))
default = self.default_method.name
if specific:
self.report_items((
("Default method", default),
("Specific imputers", ", ".join(specific))
))
else:
self.report_items((("Method", default),))
def _on_var_selection_changed(self):
indexes = self.selection.selectedIndexes()
methods = set(self.get_method_for_column(i.row()).name for i in indexes)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
if len(methods) == 1:
method = methods.pop()
for i, m in enumerate(self.METHODS):
if method == m.name:
self.variable_button_group.button(i).setChecked(True)
elif self.variable_button_group.checkedButton() is not None:
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
for method, button in zip(self.METHODS,
self.variable_button_group.buttons()):
enabled = all(method.supports_variable(var) for var in
selected_vars)
button.setEnabled(enabled)
if not has_discrete:
self.value_stack.setEnabled(True)
self.value_stack.setCurrentWidget(self.value_double)
self._on_value_changed()
elif len(selected_vars) == 1:
self.value_stack.setEnabled(True)
self.value_stack.setCurrentWidget(self.value_combo)
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
self._on_value_changed()
else:
self.variable_button_group.button(self.AS_INPUT).setEnabled(False)
self.value_stack.setEnabled(False)
def set_method_for_current_selection(self, method_index):
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
if method_index == self.DEFAULT:
for index in indexes:
self.variable_methods.pop(index, None)
else:
method = self.METHODS[method_index].copy()
for index in indexes:
self.variable_methods[index.row()] = method
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(index, self.get_method_for_column(index.row()), Qt.UserRole)
def _on_value_selected(self):
self.variable_button_group.button(self.AS_INPUT).setChecked(True)
self._on_value_changed()
def _on_value_changed(self):
widget = self.value_stack.currentWidget()
if widget is self.value_combo:
value = self.value_combo.currentText()
else:
value = self.value_double.value()
self.default_value = value
self.METHODS[self.AS_INPUT].default = value
index = self.variable_button_group.checkedId()
if index == self.AS_INPUT:
self.set_method_for_current_selection(index)
def reset_variable_methods(self):
indexes = map(self.varmodel.index, range(len(self.varmodel)))
self.set_method_for_indexes(indexes, self.DEFAULT)
self.variable_button_group.button(self.DEFAULT).setChecked(True)
class AbsoluteRelativeSpinBox(QWidget):
editingFinished = pyqtSignal()
valueChanged = pyqtSignal()
def __init__(self, *args, **kwargs):
super().__init__(*args)
layout = QStackedLayout(self)
self.double_spin = QDoubleSpinBox()
self.double_spin.valueChanged.connect(self.double_value_changed)
self.double_spin.editingFinished.connect(self.double_editing_finished)
layout.addWidget(self.double_spin)
self.int_spin = QSpinBox()
self.int_spin.setMaximum(10 ** 4)
self.int_spin.valueChanged.connect(self.int_value_changed)
self.int_spin.editingFinished.connect(self.int_editing_finished)
layout.addWidget(self.int_spin)
self.setValue(kwargs.get('value', 0.))
def double_value_changed(self):
if self.double_spin.value() > 1:
self.layout().setCurrentIndex(1)
self.int_spin.setValue(self.double_spin.value())
self.valueChanged.emit()
def double_editing_finished(self):
if self.double_spin.value() <= 1.:
self.editingFinished.emit()
def int_value_changed(self):
if self.int_spin.value() == 0:
self.layout().setCurrentIndex(0)
self.double_spin.setValue(1. - self.double_spin.singleStep())
# There is no need to emit valueChanged signal.
def int_editing_finished(self):
if self.int_spin.value() > 0:
self.editingFinished.emit()
def value(self):
return self.int_spin.value() or self.double_spin.value()
def setValue(self, value):
if isinstance(value, int):
self.layout().setCurrentIndex(1)
self.int_spin.setValue(value)
else:
self.layout().setCurrentIndex(0)
self.double_spin.setValue(value)
def setSingleStep(self, step):
if isinstance(step, float):
self.double_spin.setSingleStep(step)
else:
self.int_spin.setSingleStep(step)
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self.learner = None # type: Optional[Learner]
self.default_learner = SimpleTreeLearner(min_instances=10,
max_depth=10)
self.modified = False
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QGridLayout(box)
box_layout.setContentsMargins(5, 0, 0, 0)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, (method, _) in enumerate(list(METHODS.items())[1:-1]):
imputer = self.create_imputer(method)
button = QRadioButton(imputer.name)
button.setChecked(method == self.default_method_index)
button_group.addButton(button, method)
box_layout.addWidget(button, i % 3, i // 3)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = ListViewSearch(
selectionMode=QListView.ExtendedSelection, uniformItemSizes=True)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for method in Method:
imputer = self.create_imputer(method)
button = QRadioButton(text=imputer.name)
button_group.addButton(button, method)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000.,
maximum=1000.,
singleStep=.1,
decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection)
method_layout.addStretch(2)
reset_button = QPushButton("Restore All to Default",
checked=False,
checkable=False,
clicked=self.reset_variable_state,
default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_apply(self.controlArea, self, "autocommit")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
self.info.set_input_summary(self.info.NoInput)
self.info.set_output_summary(self.info.NoOutput)
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
keywords = ["substitute", "missing"]
class Inputs:
data = Input("Data", Orange.data.Table)
learner = Input("Learner", Learner)
class Outputs:
data = Output("Data", Orange.data.Table)
class Error(OWWidget.Error):
imputation_failed = Msg("Imputation failed for '{}'")
model_based_imputer_sparse = \
Msg("Model based imputer does not work for sparse data")
class Warning(OWWidget.Warning):
cant_handle_var = Msg("Default method can not handle '{}'")
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(int(Method.Leave)) # type: int
# Per-variable imputation state (synced in storeSpecificSettings)
_variable_imputation_state = settings.ContextSetting(
{}) # type: VariableState
autocommit = settings.Setting(True)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self.learner = None # type: Optional[Learner]
self.default_learner = SimpleTreeLearner(min_instances=10,
max_depth=10)
self.modified = False
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QGridLayout(box)
box_layout.setContentsMargins(5, 0, 0, 0)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, (method, _) in enumerate(list(METHODS.items())[1:-1]):
imputer = self.create_imputer(method)
button = QRadioButton(imputer.name)
button.setChecked(method == self.default_method_index)
button_group.addButton(button, method)
box_layout.addWidget(button, i % 3, i // 3)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = ListViewSearch(
selectionMode=QListView.ExtendedSelection, uniformItemSizes=True)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for method in Method:
imputer = self.create_imputer(method)
button = QRadioButton(text=imputer.name)
button_group.addButton(button, method)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000.,
maximum=1000.,
singleStep=.1,
decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection)
method_layout.addStretch(2)
reset_button = QPushButton("Restore All to Default",
checked=False,
checkable=False,
clicked=self.reset_variable_state,
default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_apply(self.controlArea, self, "autocommit")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
self.info.set_input_summary(self.info.NoInput)
self.info.set_output_summary(self.info.NoOutput)
def create_imputer(self, method, *args):
# type: (Method, ...) -> impute.BaseImputeMethod
if method == Method.Model:
if self.learner is not None:
return impute.Model(self.learner)
else:
return impute.Model(self.default_learner)
elif method == Method.AsAboveSoBelow:
assert self.default_method_index != Method.AsAboveSoBelow
default = self.create_imputer(Method(self.default_method_index))
m = AsDefault()
m.method = default
return m
else:
return METHODS[method](*args)
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
assert index != Method.AsAboveSoBelow
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
# update variable view
self.update_varview()
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@Inputs.data
@check_sql_input
def set_data(self, data):
self.cancel()
self.closeContext()
self.varmodel[:] = []
self._variable_imputation_state = {} # type: VariableState
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
# restore per variable imputation state
self._restore_state(self._variable_imputation_state)
summary = len(data) if data else self.info.NoInput
details = format_summary_details(data) if data else ""
self.info.set_input_summary(summary, details)
self.update_varview()
self.unconditional_commit()
@Inputs.learner
def set_learner(self, learner):
self.cancel()
self.learner = learner or self.default_learner
imputer = self.create_imputer(Method.Model)
button = self.default_button_group.button(Method.Model)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(Method.Model)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = Method.Model
self.update_varview()
self.commit()
def get_method_for_column(self, column_index):
# type: (int) -> impute.BaseImputeMethod
"""
Return the imputation method for column by its index.
"""
assert 0 <= column_index < len(self.varmodel)
idx = self.varmodel.index(column_index, 0)
state = idx.data(StateRole)
if state is None:
state = (Method.AsAboveSoBelow, ())
return self.create_imputer(state[0], *state[1])
def _invalidate(self):
self.modified = True
if self.__task is not None:
self.cancel()
self.commit()
def commit(self):
self.cancel()
self.warning()
self.Error.imputation_failed.clear()
self.Error.model_based_imputer_sparse.clear()
summary = len(self.data) if self.data else self.info.NoOutput
detail = format_summary_details(self.data) if self.data else ""
self.info.set_output_summary(summary, detail)
if not self.data or not self.varmodel.rowCount():
self.Outputs.data.send(self.data)
self.modified = False
return
data = self.data
impute_state = [(i, var, self.get_method_for_column(i))
for i, var in enumerate(self.varmodel)]
# normalize to the effective method bypasing AsDefault
impute_state = [(i, var, m.method if isinstance(m, AsDefault) else m)
for i, var, m in impute_state]
def impute_one(method, var, data):
# Readability counts, pylint: disable=no-else-raise
# type: (impute.BaseImputeMethod, Variable, Table) -> Any
if isinstance(method, impute.Model) and data.is_sparse():
raise SparseNotSupported()
elif isinstance(method, impute.DropInstances):
return RowMask(method(data, var))
elif not method.supports_variable(var):
raise VariableNotSupported(var)
else:
return method(data, var)
futures = []
for _, var, method in impute_state:
f = self.executor.submit(impute_one, copy.deepcopy(method), var,
data)
futures.append(f)
w = qconcurrent.FutureSetWatcher(futures)
w.doneAll.connect(self.__commit_finish)
w.progressChanged.connect(self.__progress_changed)
self.__task = Task(futures, w)
self.progressBarInit()
self.setInvalidated(True)
@Slot()
def __commit_finish(self):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
futures = self.__task.futures
assert len(futures) == len(self.varmodel)
assert self.data is not None
def get_variable(variable, future, drop_mask) \
-> Optional[List[Orange.data.Variable]]:
# Returns a (potentially empty) list of variables,
# or None on failure that should interrupt the imputation
assert future.done()
try:
res = future.result()
except SparseNotSupported:
self.Error.model_based_imputer_sparse()
return [] # None?
except VariableNotSupported:
self.Warning.cant_handle_var(variable.name)
return []
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error for %s", variable.name, exc_info=True)
self.Error.imputation_failed(variable.name)
return None
if isinstance(res, RowMask):
drop_mask |= res.mask
newvar = variable
else:
newvar = res
if isinstance(newvar, Orange.data.Variable):
newvar = [newvar]
return newvar
def create_data(attributes, class_vars):
domain = Orange.data.Domain(attributes, class_vars,
self.data.domain.metas)
try:
return self.data.from_table(domain, self.data[~drop_mask])
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error", exc_info=True)
self.Error.imputation_failed("Unknown")
return None
self.__task = None
self.setInvalidated(False)
self.progressBarFinished()
attributes = []
class_vars = []
drop_mask = np.zeros(len(self.data), bool)
for i, (var, fut) in enumerate(zip(self.varmodel, futures)):
newvar = get_variable(var, fut, drop_mask)
if newvar is None:
data = None
break
if i < len(self.data.domain.attributes):
attributes.extend(newvar)
else:
class_vars.extend(newvar)
else:
data = create_data(attributes, class_vars)
self.Outputs.data.send(data)
self.modified = False
summary = len(data) if data else self.info.NoOutput
details = format_summary_details(data) if data else ""
self.info.set_output_summary(summary, details)
@Slot(int, int)
def __progress_changed(self, n, d):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
self.progressBarSet(100. * n / d)
def cancel(self):
self.__cancel(wait=False)
def __cancel(self, wait=False):
if self.__task is not None:
task, self.__task = self.__task, None
task.cancel()
task.watcher.doneAll.disconnect(self.__commit_finish)
task.watcher.progressChanged.disconnect(self.__progress_changed)
if wait:
concurrent.futures.wait(task.futures)
task.watcher.flush()
self.progressBarFinished()
self.setInvalidated(False)
def onDeleteWidget(self):
self.__cancel(wait=True)
super().onDeleteWidget()
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.get_method_for_column(i)
if not isinstance(method, AsDefault):
specific.append("{} ({})".format(var.name, str(method)))
default = self.create_imputer(Method.AsAboveSoBelow)
if specific:
self.report_items((("Default method", default.name),
("Specific imputers", ", ".join(specific))))
else:
self.report_items((("Method", default.name), ))
def _on_var_selection_changed(self):
# Method is well documented, splitting it is not needed for readability,
# thus pylint: disable=too-many-branches
indexes = self.selection.selectedIndexes()
defmethod = (Method.AsAboveSoBelow, ())
methods = [index.data(StateRole) for index in indexes]
methods = [m if m is not None else defmethod for m in methods]
methods = set(methods)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
fixed_value = None
value_stack_enabled = False
current_value_widget = None
if len(methods) == 1:
method_type, parameters = methods.pop()
for m in Method:
if method_type == m:
self.variable_button_group.button(m).setChecked(True)
if method_type == Method.Default:
(fixed_value, ) = parameters
elif self.variable_button_group.checkedButton() is not None:
# Uncheck the current button
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
assert self.variable_button_group.checkedButton() is None
# Update variable methods GUI enabled state based on selection.
for method in Method:
# use a default constructed imputer to query support
imputer = self.create_imputer(method)
enabled = all(
imputer.supports_variable(var) for var in selected_vars)
button = self.variable_button_group.button(method)
button.setEnabled(enabled)
# Update the "Value" edit GUI.
if not has_discrete:
# no discrete variables -> allow mass edit for all (continuous vars)
value_stack_enabled = True
current_value_widget = self.value_double
elif len(selected_vars) == 1:
# single discrete var -> enable and fill the values combo
value_stack_enabled = True
current_value_widget = self.value_combo
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
else:
# mixed type selection -> disable
value_stack_enabled = False
current_value_widget = None
self.variable_button_group.button(Method.Default).setEnabled(False)
self.value_stack.setEnabled(value_stack_enabled)
if current_value_widget is not None:
self.value_stack.setCurrentWidget(current_value_widget)
if fixed_value is not None:
# set current value
if current_value_widget is self.value_combo:
self.value_combo.setCurrentIndex(fixed_value)
elif current_value_widget is self.value_double:
self.value_double.setValue(fixed_value)
else:
assert False
def set_method_for_current_selection(self, method_index):
# type: (Method) -> None
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
# type: (List[QModelIndex], Method) -> None
if method_index == Method.AsAboveSoBelow:
for index in indexes:
self.varmodel.setData(index, None, StateRole)
elif method_index == Method.Default:
current = self.value_stack.currentWidget()
if current is self.value_combo:
value = self.value_combo.currentIndex()
else:
value = self.value_double.value()
for index in indexes:
state = (int(Method.Default), (value, ))
self.varmodel.setData(index, state, StateRole)
else:
state = (int(method_index), ())
for index in indexes:
self.varmodel.setData(index, state, StateRole)
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(index,
self.get_method_for_column(index.row()),
DisplayMethodRole)
def _on_value_selected(self):
# The fixed 'Value' in the widget has been changed by the user.
self.variable_button_group.button(Method.Default).setChecked(True)
self.set_method_for_current_selection(Method.Default)
def reset_variable_state(self):
indexes = list(map(self.varmodel.index, range(len(self.varmodel))))
self.set_method_for_indexes(indexes, Method.AsAboveSoBelow)
self.variable_button_group.button(
Method.AsAboveSoBelow).setChecked(True)
def _store_state(self):
# type: () -> VariableState
"""
Save the current variable imputation state
"""
state = {} # type: VariableState
for i, var in enumerate(self.varmodel):
index = self.varmodel.index(i)
m = index.data(StateRole)
if m is not None:
state[var_key(var)] = m
return state
def _restore_state(self, state):
# type: (VariableState) -> None
"""
Restore the variable imputation state from the saved state
"""
def check(state):
# check if state is a proper State
if isinstance(state, tuple) and len(state) == 2:
m, p = state
if isinstance(m, int) and isinstance(p, tuple) and \
0 <= m < len(Method):
return True
return False
for i, var in enumerate(self.varmodel):
m = state.get(var_key(var), None)
if check(m):
self.varmodel.setData(self.varmodel.index(i), m, StateRole)
def storeSpecificSettings(self):
self._variable_imputation_state = self._store_state()
super().storeSpecificSettings()
def __init__(self):
super().__init__()
# copy METHODS (some are modified by the widget)
self.methods = copy.deepcopy(OWImpute.METHODS)
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.methods):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.methods):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_methods, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal,
checkbox_label="Apply automatically")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
self.data = None
self.learner = None
self.modified = False
self.default_method = self.methods[self.default_method_index]
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self._state = None # type: Optional[_FilterData]
box = gui.widgetBox(self.controlArea, "Info")
self._info = QLabel(box, wordWrap=True)
self._info.setText("No data in input\n")
box.layout().addWidget(self._info)
box = gui.widgetBox(self.controlArea, "Filter Type", spacing=-1)
rbg = QButtonGroup(box, exclusive=True)
layout = QHBoxLayout()
layout.setContentsMargins(0, 0, 0, 0)
for id_ in [Cells, Genes, Data]:
name, _, tip = FilterInfo[id_]
b = QRadioButton(name,
toolTip=tip,
checked=id_ == self.selected_filter_type)
rbg.addButton(b, id_)
layout.addWidget(b, stretch=10, alignment=Qt.AlignCenter)
box.layout().addLayout(layout)
rbg.buttonClicked[int].connect(self.set_filter_type)
self.filter_metric_cb = gui.comboBox(
box,
self,
"selected_filter_metric",
callback=self._update_metric,
enabled=self.selected_filter_type != Data)
for id_ in [DetectionCount, TotalCounts]:
text, ttip = MeasureInfo[id_]
self.filter_metric_cb.addItem(text)
idx = self.filter_metric_cb.count() - 1
self.filter_metric_cb.setItemData(idx, ttip, Qt.ToolTipRole)
self.filter_metric_cb.setCurrentIndex(self.selected_filter_metric)
form = QFormLayout(labelAlignment=Qt.AlignLeft,
formAlignment=Qt.AlignLeft,
fieldGrowthPolicy=QFormLayout.AllNonFixedFieldsGrow)
self._filter_box = box = gui.widgetBox(
self.controlArea, "Filter", orientation=form) # type: QGroupBox
self.threshold_stacks = (
QStackedWidget(enabled=self.limit_lower_enabled),
QStackedWidget(enabled=self.limit_upper_enabled),
)
finfo = np.finfo(np.float64)
for filter_ in [Cells, Genes, Data]:
if filter_ in {Cells, Genes}:
minimum = 0.0
ndecimals = 1
metric = self.selected_filter_metric
else:
minimum = finfo.min
ndecimals = 3
metric = -1
spinlower = QDoubleSpinBox(
self,
minimum=minimum,
maximum=finfo.max,
decimals=ndecimals,
keyboardTracking=False,
)
spinupper = QDoubleSpinBox(
self,
minimum=minimum,
maximum=finfo.max,
decimals=ndecimals,
keyboardTracking=False,
)
lower, upper = self.thresholds.get((filter_, metric), (0, 0))
spinlower.setValue(lower)
spinupper.setValue(upper)
self.threshold_stacks[0].addWidget(spinlower)
self.threshold_stacks[1].addWidget(spinupper)
spinlower.valueChanged.connect(self._limitchanged)
spinupper.valueChanged.connect(self._limitchanged)
self.threshold_stacks[0].setCurrentIndex(self.selected_filter_type)
self.threshold_stacks[1].setCurrentIndex(self.selected_filter_type)
self.limit_lower_enabled_cb = cb = QCheckBox(
"Min", checked=self.limit_lower_enabled)
cb.toggled.connect(self.set_lower_limit_enabled)
cb.setAttribute(Qt.WA_LayoutUsesWidgetRect, True)
form.addRow(cb, self.threshold_stacks[0])
self.limit_upper_enabled_cb = cb = QCheckBox(
"Max", checked=self.limit_upper_enabled)
cb.toggled.connect(self.set_upper_limit_enabled)
cb.setAttribute(Qt.WA_LayoutUsesWidgetRect, True)
form.addRow(cb, self.threshold_stacks[1])
box = gui.widgetBox(self.controlArea, "Plot Options")
self._showpoints = gui.checkBox(box,
self,
"display_dotplot",
"Show data points",
callback=self._update_dotplot)
self.log_scale_cb = QCheckBox("Log scale",
checked=self.scale == Scale.Log1p.name)
self.log_scale_cb.toggled[bool].connect(
lambda state: self.set_filter_scale(Scale.Log1p
if state else Scale.Linear))
box.layout().addWidget(self.log_scale_cb)
self.controlArea.layout().addStretch(10)
gui.auto_commit(self.controlArea, self, "auto_commit", "Commit")
self._view = pg.GraphicsView()
self._view.enableMouse(False)
self._view.setAntialiasing(True)
self._plot = plot = ViolinPlot()
self._plot.setDataPointsVisible(self.display_dotplot)
self._plot.setSelectionMode(
(ViolinPlot.Low if self.limit_lower_enabled else 0)
| (ViolinPlot.High if self.limit_upper_enabled else 0))
self._plot.setRange(QRectF(-1., 0., 2., 1.))
self._plot.selectionEdited.connect(self._limitchanged_plot)
self._view.setCentralWidget(self._plot)
bottom = self._plot.getAxis("bottom") # type: pg.AxisItem
bottom.hide()
plot.setMouseEnabled(False, False)
plot.hideButtons()
self.mainArea.layout().addWidget(self._view)
# Coalescing commit timer
self._committimer = QTimer(self, singleShot=True)
self._committimer.timeout.connect(self.commit)
self.addAction(
QAction("Select All",
self,
shortcut=QKeySequence.SelectAll,
triggered=self._select_all))
self._setup_axes()
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self.learner = None # type: Optional[Learner]
self.default_learner = SimpleTreeLearner()
self.modified = False
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for method, _ in list(METHODS.items())[1:-1]:
imputer = self.create_imputer(method)
button = QRadioButton(imputer.name)
button.setChecked(method == self.default_method_index)
button_group.addButton(button, method)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection,
uniformItemSizes=True
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for method in Method:
imputer = self.create_imputer(method)
button = QRadioButton(text=imputer.name)
button_group.addButton(button, method)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_state, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal,
checkbox_label="Apply automatically")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
class Inputs:
data = Input("Data", Orange.data.Table)
learner = Input("Learner", Learner)
class Outputs:
data = Output("Data", Orange.data.Table)
class Error(OWWidget.Error):
imputation_failed = Msg("Imputation failed for '{}'")
model_based_imputer_sparse = Msg("Model based imputer does not work for sparse data")
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(int(Method.Leave)) # type: int
# Per-variable imputation state (synced in storeSpecificSettings)
_variable_imputation_state = settings.ContextSetting({}) # type: VariableState
autocommit = settings.Setting(True)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self.learner = None # type: Optional[Learner]
self.default_learner = SimpleTreeLearner()
self.modified = False
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for method, _ in list(METHODS.items())[1:-1]:
imputer = self.create_imputer(method)
button = QRadioButton(imputer.name)
button.setChecked(method == self.default_method_index)
button_group.addButton(button, method)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection,
uniformItemSizes=True
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for method in Method:
imputer = self.create_imputer(method)
button = QRadioButton(text=imputer.name)
button_group.addButton(button, method)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_state, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal,
checkbox_label="Apply automatically")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
def create_imputer(self, method, *args):
# type: (Method, ...) -> impute.BaseImputeMethod
if method == Method.Model:
if self.learner is not None:
return impute.Model(self.learner)
else:
return impute.Model(self.default_learner)
elif method == Method.AsAboveSoBelow:
assert self.default_method_index != Method.AsAboveSoBelow
default = self.create_imputer(Method(self.default_method_index))
m = AsDefault()
m.method = default
return m
else:
return METHODS[method](*args)
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
assert index != Method.AsAboveSoBelow
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
# update variable view
self.update_varview()
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@Inputs.data
@check_sql_input
def set_data(self, data):
self.closeContext()
self.varmodel[:] = []
self._variable_imputation_state = {} # type: VariableState
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
# restore per variable imputation state
self._restore_state(self._variable_imputation_state)
self.update_varview()
self.unconditional_commit()
@Inputs.learner
def set_learner(self, learner):
self.learner = learner or self.default_learner
imputer = self.create_imputer(Method.Model)
button = self.default_button_group.button(Method.Model)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(Method.Model)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = Method.Model
self.update_varview()
self.commit()
def get_method_for_column(self, column_index):
# type: (int) -> impute.BaseImputeMethod
"""
Return the imputation method for column by its index.
"""
assert 0 <= column_index < len(self.varmodel)
idx = self.varmodel.index(column_index, 0)
state = idx.data(StateRole)
if state is None:
state = (Method.AsAboveSoBelow, ())
return self.create_imputer(state[0], *state[1])
def _invalidate(self):
self.modified = True
if self.__task is not None:
self.cancel()
self.commit()
def commit(self):
self.cancel()
self.warning()
self.Error.imputation_failed.clear()
self.Error.model_based_imputer_sparse.clear()
if self.data is None or len(self.data) == 0 or len(self.varmodel) == 0:
self.Outputs.data.send(self.data)
self.modified = False
return
data = self.data
impute_state = [
(i, var, self.get_method_for_column(i))
for i, var in enumerate(self.varmodel)
]
# normalize to the effective method bypasing AsDefault
impute_state = [
(i, var, m.method if isinstance(m, AsDefault) else m)
for i, var, m in impute_state
]
def impute_one(method, var, data):
# type: (impute.BaseImputeMethod, Variable, Table) -> Any
if isinstance(method, impute.Model) and data.is_sparse():
raise SparseNotSupported()
elif isinstance(method, impute.DropInstances):
return RowMask(method(data, var))
elif not method.supports_variable(var):
raise VariableNotSupported(var)
else:
return method(data, var)
futures = []
for _, var, method in impute_state:
f = self.executor.submit(
impute_one, copy.deepcopy(method), var, data)
futures.append(f)
w = qconcurrent.FutureSetWatcher(futures)
w.doneAll.connect(self.__commit_finish)
w.progressChanged.connect(self.__progress_changed)
self.__task = Task(futures, w)
self.progressBarInit(processEvents=False)
self.setBlocking(True)
@Slot()
def __commit_finish(self):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
futures = self.__task.futures
assert len(futures) == len(self.varmodel)
assert self.data is not None
self.__task = None
self.setBlocking(False)
self.progressBarFinished()
data = self.data
attributes = []
class_vars = []
drop_mask = np.zeros(len(self.data), bool)
for i, (var, fut) in enumerate(zip(self.varmodel, futures)):
assert fut.done()
newvar = []
try:
res = fut.result()
except SparseNotSupported:
self.Error.model_based_imputer_sparse()
# ?? break
except VariableNotSupported:
self.warning("Default method can not handle '{}'".
format(var.name))
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error for %s", var, exc_info=True)
self.Error.imputation_failed(var.name)
attributes = class_vars = None
break
else:
if isinstance(res, RowMask):
drop_mask |= res.mask
newvar = var
else:
newvar = res
if isinstance(newvar, Orange.data.Variable):
newvar = [newvar]
if i < len(data.domain.attributes):
attributes.extend(newvar)
else:
class_vars.extend(newvar)
if attributes is None:
data = None
else:
domain = Orange.data.Domain(attributes, class_vars,
data.domain.metas)
try:
data = self.data.from_table(domain, data[~drop_mask])
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error", exc_info=True)
self.Error.imputation_failed("Unknown")
data = None
self.Outputs.data.send(data)
self.modified = False
@Slot(int, int)
def __progress_changed(self, n, d):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
self.progressBarSet(100. * n / d)
def cancel(self):
if self.__task is not None:
task, self.__task = self.__task, None
task.cancel()
task.watcher.doneAll.disconnect(self.__commit_finish)
task.watcher.progressChanged.disconnect(self.__progress_changed)
concurrent.futures.wait(task.futures)
task.watcher.flush()
self.progressBarFinished()
self.setBlocking(False)
def onDeleteWidget(self):
self.cancel()
super().onDeleteWidget()
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.get_method_for_column(i)
if not isinstance(method, AsDefault):
specific.append("{} ({})".format(var.name, str(method)))
default = self.create_imputer(Method.AsAboveSoBelow)
if specific:
self.report_items((
("Default method", default.name),
("Specific imputers", ", ".join(specific))
))
else:
self.report_items((("Method", default.name),))
def _on_var_selection_changed(self):
indexes = self.selection.selectedIndexes()
defmethod = (Method.AsAboveSoBelow, ())
methods = [index.data(StateRole) for index in indexes]
methods = [m if m is not None else defmethod for m in methods]
methods = set(methods)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
fixed_value = None
value_stack_enabled = False
current_value_widget = None
if len(methods) == 1:
method_type, parameters = methods.pop()
for m in Method:
if method_type == m:
self.variable_button_group.button(m).setChecked(True)
if method_type == Method.Default:
(fixed_value,) = parameters
elif self.variable_button_group.checkedButton() is not None:
# Uncheck the current button
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
assert self.variable_button_group.checkedButton() is None
# Update variable methods GUI enabled state based on selection.
for method in Method:
# use a default constructed imputer to query support
imputer = self.create_imputer(method)
enabled = all(imputer.supports_variable(var) for var in
selected_vars)
button = self.variable_button_group.button(method)
button.setEnabled(enabled)
# Update the "Value" edit GUI.
if not has_discrete:
# no discrete variables -> allow mass edit for all (continuous vars)
value_stack_enabled = True
current_value_widget = self.value_double
elif len(selected_vars) == 1:
# single discrete var -> enable and fill the values combo
value_stack_enabled = True
current_value_widget = self.value_combo
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
else:
# mixed type selection -> disable
value_stack_enabled = False
current_value_widget = None
self.variable_button_group.button(Method.Default).setEnabled(False)
self.value_stack.setEnabled(value_stack_enabled)
if current_value_widget is not None:
self.value_stack.setCurrentWidget(current_value_widget)
if fixed_value is not None:
# set current value
if current_value_widget is self.value_combo:
self.value_combo.setCurrentIndex(fixed_value)
elif current_value_widget is self.value_double:
self.value_double.setValue(fixed_value)
else:
assert False
def set_method_for_current_selection(self, method_index):
# type: (Method) -> None
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
# type: (List[QModelIndex], Method) -> None
if method_index == Method.AsAboveSoBelow:
for index in indexes:
self.varmodel.setData(index, None, StateRole)
elif method_index == Method.Default:
current = self.value_stack.currentWidget()
if current is self.value_combo:
value = self.value_combo.currentIndex()
else:
value = self.value_double.value()
for index in indexes:
state = (int(Method.Default), (value,))
self.varmodel.setData(index, state, StateRole)
else:
state = (int(method_index), ())
for index in indexes:
self.varmodel.setData(index, state, StateRole)
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(
index, self.get_method_for_column(index.row()),
DisplayMethodRole)
def _on_value_selected(self):
# The fixed 'Value' in the widget has been changed by the user.
self.variable_button_group.button(Method.Default).setChecked(True)
self.set_method_for_current_selection(Method.Default)
def reset_variable_state(self):
indexes = list(map(self.varmodel.index, range(len(self.varmodel))))
self.set_method_for_indexes(indexes, Method.AsAboveSoBelow)
self.variable_button_group.button(Method.AsAboveSoBelow).setChecked(True)
def _store_state(self):
# type: () -> VariableState
"""
Save the current variable imputation state
"""
state = {} # type: VariableState
for i, var in enumerate(self.varmodel):
index = self.varmodel.index(i)
m = index.data(StateRole)
if m is not None:
state[var_key(var)] = m
return state
def _restore_state(self, state):
# type: (VariableState) -> None
"""
Restore the variable imputation state from the saved state
"""
def check(state):
# check if state is a proper State
if isinstance(state, tuple) and len(state) == 2:
m, p = state
if isinstance(m, int) and isinstance(p, tuple) and \
0 <= m < len(Method):
return True
return False
for i, var in enumerate(self.varmodel):
m = state.get(var_key(var), None)
if check(m):
self.varmodel.setData(self.varmodel.index(i), m, StateRole)
def storeSpecificSettings(self):
self._variable_imputation_state = self._store_state()
super().storeSpecificSettings()
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self._counts = None # type: Optional[np.ndarray]
box = gui.widgetBox(self.controlArea, "Info")
self._info = QLabel(box, wordWrap=True)
self._info.setText("No data in input\n")
box.layout().addWidget(self._info)
box = gui.widgetBox(self.controlArea, "Filter Type")
rbg = QButtonGroup(box, exclusive=True)
for id_ in [Cells, Genes, Data]:
name, _, tip = FilterInfo[id_]
b = QRadioButton(
name, toolTip=tip, checked=id_ == self.selected_filter_type
)
box.layout().addWidget(b)
rbg.addButton(b, id_)
rbg.buttonClicked[int].connect(self.set_filter_type)
box = gui.widgetBox(self.controlArea, "View")
self._showpoints = gui.checkBox(
box, self, "display_dotplot", "Show data points",
callback=self._update_dotplot
)
form = QFormLayout(
labelAlignment=Qt.AlignLeft,
formAlignment=Qt.AlignLeft,
fieldGrowthPolicy=QFormLayout.AllNonFixedFieldsGrow
)
self._filter_box = box = gui.widgetBox(
self.controlArea, "Filter", orientation=form
) # type: QGroupBox
self.threshold_stacks = (
QStackedWidget(enabled=self.limit_lower_enabled),
QStackedWidget(enabled=self.limit_upper_enabled),
)
finfo = np.finfo(np.float64)
for filter_ in [Cells, Genes, Data]:
if filter_ in {Cells, Genes}:
minimum = 0.0
ndecimals = 1
else:
minimum = finfo.min
ndecimals = 3
spinlower = QDoubleSpinBox(
self, minimum=minimum, maximum=finfo.max, decimals=ndecimals,
keyboardTracking=False,
)
spinupper = QDoubleSpinBox(
self, minimum=minimum, maximum=finfo.max, decimals=ndecimals,
keyboardTracking=False,
)
lower, upper = self.thresholds[filter_]
spinlower.setValue(lower)
spinupper.setValue(upper)
self.threshold_stacks[0].addWidget(spinlower)
self.threshold_stacks[1].addWidget(spinupper)
spinlower.valueChanged.connect(self._limitchanged)
spinupper.valueChanged.connect(self._limitchanged)
self.threshold_stacks[0].setCurrentIndex(self.selected_filter_type)
self.threshold_stacks[1].setCurrentIndex(self.selected_filter_type)
self.limit_lower_enabled_cb = cb = QCheckBox(
"Min", checked=self.limit_lower_enabled
)
cb.toggled.connect(self.set_lower_limit_enabled)
cb.setAttribute(Qt.WA_LayoutUsesWidgetRect, True)
form.addRow(cb, self.threshold_stacks[0])
self.limit_upper_enabled_cb = cb = QCheckBox(
"Max", checked=self.limit_upper_enabled
)
cb.toggled.connect(self.set_upper_limit_enabled)
cb.setAttribute(Qt.WA_LayoutUsesWidgetRect, True)
form.addRow(cb, self.threshold_stacks[1])
self.controlArea.layout().addStretch(10)
gui.auto_commit(self.controlArea, self, "auto_commit", "Commit")
self._view = pg.GraphicsView()
self._view.enableMouse(False)
self._view.setAntialiasing(True)
self._plot = plot = ViolinPlot()
self._plot.setDataPointsVisible(self.display_dotplot)
self._plot.setSelectionMode(
(ViolinPlot.Low if self.limit_lower_enabled else 0) |
(ViolinPlot.High if self.limit_upper_enabled else 0)
)
self._plot.selectionEdited.connect(self._limitchanged_plot)
self._view.setCentralWidget(self._plot)
self._plot.setTitle("Detected genes")
left = self._plot.getAxis("left") # type: pg.AxisItem
left.setLabel("Detected genes")
bottom = self._plot.getAxis("bottom") # type: pg.AxisItem
bottom.hide()
plot.setMouseEnabled(False, False)
plot.hideButtons()
self.mainArea.layout().addWidget(self._view)
# Coalescing commit timer
self._committimer = QTimer(self, singleShot=True)
self._committimer.timeout.connect(self.commit)
self.addAction(
QAction("Select All", self, shortcut=QKeySequence.SelectAll,
triggered=self._select_all)
)
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
class Inputs:
data = Input("Data", Orange.data.Table)
learner = Input("Learner", Learner)
class Outputs:
data = Output("Data", Orange.data.Table)
class Error(OWWidget.Error):
imputation_failed = Msg("Imputation failed for '{}'")
model_based_imputer_sparse = Msg("Model based imputer does not work for sparse data")
DEFAULT_LEARNER = SimpleTreeLearner()
METHODS = [AsDefault(), impute.DoNotImpute(), impute.Average(),
impute.AsValue(), impute.Model(DEFAULT_LEARNER), impute.Random(),
impute.DropInstances(), impute.Default()]
DEFAULT, DO_NOT_IMPUTE, MODEL_BASED_IMPUTER, AS_INPUT = 0, 1, 4, 7
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(DO_NOT_IMPUTE)
variable_methods = settings.ContextSetting({})
autocommit = settings.Setting(True)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
# copy METHODS (some are modified by the widget)
self.methods = copy.deepcopy(OWImpute.METHODS)
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.methods):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.methods):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_methods, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal,
checkbox_label="Apply automatically")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
self.data = None
self.learner = None
self.modified = False
self.default_method = self.methods[self.default_method_index]
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
self.default_method = self.methods[self.default_method_index]
self.methods[self.DEFAULT].method = self.default_method
# update variable view
for index in map(self.varmodel.index, range(len(self.varmodel))):
method = self.variable_methods.get(
index.row(), self.methods[self.DEFAULT])
self.varmodel.setData(index, method, Qt.UserRole)
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@Inputs.data
@check_sql_input
def set_data(self, data):
self.closeContext()
self.varmodel[:] = []
self.variable_methods = {}
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
self.update_varview()
self.unconditional_commit()
@Inputs.learner
def set_learner(self, learner):
self.learner = learner or self.DEFAULT_LEARNER
imputer = self.methods[self.MODEL_BASED_IMPUTER]
imputer.learner = self.learner
button = self.default_button_group.button(self.MODEL_BASED_IMPUTER)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(self.MODEL_BASED_IMPUTER)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = self.MODEL_BASED_IMPUTER
self.update_varview()
self.commit()
def get_method_for_column(self, column_index):
"""Returns the imputation method for column by its index.
"""
if not isinstance(column_index, int):
column_index = column_index.row()
return self.variable_methods.get(column_index,
self.methods[self.DEFAULT])
def _invalidate(self):
self.modified = True
if self.__task is not None:
self.cancel()
self.commit()
def commit(self):
self.cancel()
self.warning()
self.Error.imputation_failed.clear()
self.Error.model_based_imputer_sparse.clear()
if self.data is None or len(self.data) == 0 or len(self.varmodel) == 0:
self.Outputs.data.send(self.data)
self.modified = False
return
data = self.data
impute_state = [
(i, var, self.variable_methods.get(i, self.default_method))
for i, var in enumerate(self.varmodel)
]
def impute_one(method, var, data):
# type: (impute.BaseImputeMethod, Variable, Table) -> Any
if isinstance(method, impute.Model) and data.is_sparse():
raise SparseNotSupported()
elif isinstance(method, impute.DropInstances):
return RowMask(method(data, var))
elif not method.supports_variable(var):
raise VariableNotSupported(var)
else:
return method(data, var)
futures = []
for _, var, method in impute_state:
f = self.executor.submit(
impute_one, copy.deepcopy(method), var, data)
futures.append(f)
w = qconcurrent.FutureSetWatcher(futures)
w.doneAll.connect(self.__commit_finish)
w.progressChanged.connect(self.__progress_changed)
self.__task = Task(futures, w)
self.progressBarInit(processEvents=False)
self.setBlocking(True)
@Slot()
def __commit_finish(self):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
futures = self.__task.futures
assert len(futures) == len(self.varmodel)
assert self.data is not None
self.__task = None
self.setBlocking(False)
self.progressBarFinished()
data = self.data
attributes = []
class_vars = []
drop_mask = np.zeros(len(self.data), bool)
for i, (var, fut) in enumerate(zip(self.varmodel, futures)):
assert fut.done()
newvar = []
try:
res = fut.result()
except SparseNotSupported:
self.Error.model_based_imputer_sparse()
# ?? break
except VariableNotSupported:
self.warning("Default method can not handle '{}'".
format(var.name))
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error for %s", var, exc_info=True)
self.Error.imputation_failed(var.name)
attributes = class_vars = None
break
else:
if isinstance(res, RowMask):
drop_mask |= res.mask
newvar = var
else:
newvar = res
if isinstance(newvar, Orange.data.Variable):
newvar = [newvar]
if i < len(data.domain.attributes):
attributes.extend(newvar)
else:
class_vars.extend(newvar)
if attributes is None:
data = None
else:
domain = Orange.data.Domain(attributes, class_vars,
data.domain.metas)
try:
data = self.data.from_table(domain, data[~drop_mask])
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error", exc_info=True)
self.Error.imputation_failed("Unknown")
data = None
self.Outputs.data.send(data)
self.modified = False
@Slot(int, int)
def __progress_changed(self, n, d):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
self.progressBarSet(100. * n / d)
def cancel(self):
if self.__task is not None:
task, self.__task = self.__task, None
task.cancel()
task.watcher.doneAll.disconnect(self.__commit_finish)
task.watcher.progressChanged.disconnect(self.__progress_changed)
concurrent.futures.wait(task.futures)
task.watcher.flush()
self.progressBarFinished()
self.setBlocking(False)
def onDeleteWidget(self):
self.cancel()
super().onDeleteWidget()
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, None)
if method is not None:
specific.append("{} ({})".format(var.name, str(method)))
default = self.default_method.name
if specific:
self.report_items((
("Default method", default),
("Specific imputers", ", ".join(specific))
))
else:
self.report_items((("Method", default),))
def _on_var_selection_changed(self):
indexes = self.selection.selectedIndexes()
methods = [self.get_method_for_column(i.row()) for i in indexes]
def method_key(method):
"""
Decompose method into its type and parameters.
"""
# The return value should be hashable and __eq__ comparable
if isinstance(method, AsDefault):
return AsDefault, (method.method,)
elif isinstance(method, impute.Model):
return impute.Model, (method.learner,)
elif isinstance(method, impute.Default):
return impute.Default, (method.default,)
else:
return type(method), None
methods = set(method_key(m) for m in methods)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
fixed_value = None
value_stack_enabled = False
current_value_widget = None
if len(methods) == 1:
method_type, parameters = methods.pop()
for i, m in enumerate(self.methods):
if method_type == type(m):
self.variable_button_group.button(i).setChecked(True)
if method_type is impute.Default:
(fixed_value,) = parameters
elif self.variable_button_group.checkedButton() is not None:
# Uncheck the current button
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
assert self.variable_button_group.checkedButton() is None
for method, button in zip(self.methods,
self.variable_button_group.buttons()):
enabled = all(method.supports_variable(var) for var in
selected_vars)
button.setEnabled(enabled)
if not has_discrete:
value_stack_enabled = True
current_value_widget = self.value_double
elif len(selected_vars) == 1:
value_stack_enabled = True
current_value_widget = self.value_combo
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
else:
value_stack_enabled = False
current_value_widget = None
self.variable_button_group.button(self.AS_INPUT).setEnabled(False)
self.value_stack.setEnabled(value_stack_enabled)
if current_value_widget is not None:
self.value_stack.setCurrentWidget(current_value_widget)
if fixed_value is not None:
if current_value_widget is self.value_combo:
self.value_combo.setCurrentIndex(fixed_value)
elif current_value_widget is self.value_double:
self.value_double.setValue(fixed_value)
else:
assert False
def set_method_for_current_selection(self, method_index):
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
if method_index == self.DEFAULT:
for index in indexes:
self.variable_methods.pop(index.row(), None)
elif method_index == OWImpute.AS_INPUT:
current = self.value_stack.currentWidget()
if current is self.value_combo:
value = self.value_combo.currentIndex()
else:
value = self.value_double.value()
for index in indexes:
method = impute.Default(default=value)
self.variable_methods[index.row()] = method
else:
method = self.methods[method_index]
for index in indexes:
self.variable_methods[index.row()] = method
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(index, self.get_method_for_column(index.row()), Qt.UserRole)
def _on_value_selected(self):
# The fixed 'Value' in the widget has been changed by the user.
self.variable_button_group.button(self.AS_INPUT).setChecked(True)
self.set_method_for_current_selection(self.AS_INPUT)
def reset_variable_methods(self):
indexes = list(map(self.varmodel.index, range(len(self.varmodel))))
self.set_method_for_indexes(indexes, self.DEFAULT)
self.variable_button_group.button(self.DEFAULT).setChecked(True)
def __init__(self):
super().__init__()
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.METHODS):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.METHODS):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_combo.currentIndexChanged.connect(self._on_value_changed)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
value=self.default_value
)
self.value_stack = value_stack = QStackedLayout()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addLayout(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_methods, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal, checkbox_label="Apply automatically")
box.layout().insertSpacing(0, 80)
box.layout().insertWidget(0, self.report_button)
self.data = None
self.modified = False
self.default_method = self.METHODS[self.default_method_index]
self.update_varview()
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
class Inputs:
data = Input("Data", Orange.data.Table)
learner = Input("Learner", Learner)
class Outputs:
data = Output("Data", Orange.data.Table)
class Error(OWWidget.Error):
imputation_failed = Msg("Imputation failed for '{}'")
model_based_imputer_sparse = Msg("Model based imputer does not work for sparse data")
DEFAULT_LEARNER = SimpleTreeLearner()
METHODS = [AsDefault(), impute.DoNotImpute(), impute.Average(),
impute.AsValue(), impute.Model(DEFAULT_LEARNER), impute.Random(),
impute.DropInstances(), impute.Default()]
DEFAULT, DO_NOT_IMPUTE, MODEL_BASED_IMPUTER, AS_INPUT = 0, 1, 4, 7
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(DO_NOT_IMPUTE)
variable_methods = settings.ContextSetting({})
autocommit = settings.Setting(True)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
# copy METHODS (some are modified by the widget)
self.methods = copy.deepcopy(OWImpute.METHODS)
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.methods):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(
selectionMode=QListView.ExtendedSelection
)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed
)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.methods):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected
)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000., maximum=1000., singleStep=.1, decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection
)
method_layout.addStretch(2)
reset_button = QPushButton(
"Restore All to Default", checked=False, checkable=False,
clicked=self.reset_variable_methods, default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(
self.controlArea, self, "autocommit", "Apply",
orientation=Qt.Horizontal, checkbox_label="Apply automatically")
box.layout().insertSpacing(0, 80)
box.layout().insertWidget(0, self.report_button)
self.data = None
self.learner = None
self.modified = False
self.default_method = self.methods[self.default_method_index]
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
self.default_method = self.methods[self.default_method_index]
self.methods[self.DEFAULT].method = self.default_method
# update variable view
for index in map(self.varmodel.index, range(len(self.varmodel))):
method = self.variable_methods.get(
index.row(), self.methods[self.DEFAULT])
self.varmodel.setData(index, method, Qt.UserRole)
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@Inputs.data
@check_sql_input
def set_data(self, data):
self.closeContext()
self.varmodel[:] = []
self.variable_methods = {}
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
self.update_varview()
self.unconditional_commit()
@Inputs.learner
def set_learner(self, learner):
self.learner = learner or self.DEFAULT_LEARNER
imputer = self.methods[self.MODEL_BASED_IMPUTER]
imputer.learner = self.learner
button = self.default_button_group.button(self.MODEL_BASED_IMPUTER)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(self.MODEL_BASED_IMPUTER)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = self.MODEL_BASED_IMPUTER
self.update_varview()
self.commit()
def get_method_for_column(self, column_index):
"""Returns the imputation method for column by its index.
"""
if not isinstance(column_index, int):
column_index = column_index.row()
return self.variable_methods.get(column_index,
self.methods[self.DEFAULT])
def _invalidate(self):
self.modified = True
if self.__task is not None:
self.cancel()
self.commit()
def commit(self):
self.cancel()
self.warning()
self.Error.imputation_failed.clear()
self.Error.model_based_imputer_sparse.clear()
if self.data is None or len(self.data) == 0 or len(self.varmodel) == 0:
self.Outputs.data.send(self.data)
self.modified = False
return
data = self.data
impute_state = [
(i, var, self.variable_methods.get(i, self.default_method))
for i, var in enumerate(self.varmodel)
]
def impute_one(method, var, data):
# type: (impute.BaseImputeMethod, Variable, Table) -> Any
if isinstance(method, impute.Model) and data.is_sparse():
raise SparseNotSupported()
elif isinstance(method, impute.DropInstances):
return RowMask(method(data, var))
elif not method.supports_variable(var):
raise VariableNotSupported(var)
else:
return method(data, var)
futures = []
for _, var, method in impute_state:
f = self.executor.submit(
impute_one, copy.deepcopy(method), var, data)
futures.append(f)
w = qconcurrent.FutureSetWatcher(futures)
w.doneAll.connect(self.__commit_finish)
w.progressChanged.connect(self.__progress_changed)
self.__task = Task(futures, w)
self.progressBarInit(processEvents=False)
self.setBlocking(True)
@Slot()
def __commit_finish(self):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
futures = self.__task.futures
assert len(futures) == len(self.varmodel)
assert self.data is not None
self.__task = None
self.setBlocking(False)
self.progressBarFinished()
data = self.data
attributes = []
class_vars = []
drop_mask = np.zeros(len(self.data), bool)
for i, (var, fut) in enumerate(zip(self.varmodel, futures)):
assert fut.done()
newvar = []
try:
res = fut.result()
except SparseNotSupported:
self.Error.model_based_imputer_sparse()
# ?? break
except VariableNotSupported:
self.warning("Default method can not handle '{}'".
format(var.name))
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error for %s", var, exc_info=True)
self.Error.imputation_failed(var.name)
attributes = class_vars = None
break
else:
if isinstance(res, RowMask):
drop_mask |= res.mask
newvar = var
else:
newvar = res
if isinstance(newvar, Orange.data.Variable):
newvar = [newvar]
if i < len(data.domain.attributes):
attributes.extend(newvar)
else:
class_vars.extend(newvar)
if attributes is None:
data = None
else:
domain = Orange.data.Domain(attributes, class_vars,
data.domain.metas)
try:
data = self.data.from_table(domain, data[~drop_mask])
except Exception: # pylint: disable=broad-except
log = logging.getLogger(__name__)
log.info("Error", exc_info=True)
self.Error.imputation_failed("Unknown")
data = None
self.Outputs.data.send(data)
self.modified = False
@Slot(int, int)
def __progress_changed(self, n, d):
assert QThread.currentThread() is self.thread()
assert self.__task is not None
self.progressBarSet(100. * n / d)
def cancel(self):
if self.__task is not None:
task, self.__task = self.__task, None
task.cancel()
task.watcher.doneAll.disconnect(self.__commit_finish)
task.watcher.progressChanged.disconnect(self.__progress_changed)
concurrent.futures.wait(task.futures)
task.watcher.flush()
self.progressBarFinished()
self.setBlocking(False)
def onDeleteWidget(self):
self.cancel()
super().onDeleteWidget()
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, None)
if method is not None:
specific.append("{} ({})".format(var.name, str(method)))
default = self.default_method.name
if specific:
self.report_items((
("Default method", default),
("Specific imputers", ", ".join(specific))
))
else:
self.report_items((("Method", default),))
def _on_var_selection_changed(self):
indexes = self.selection.selectedIndexes()
methods = [self.get_method_for_column(i.row()) for i in indexes]
def method_key(method):
"""
Decompose method into its type and parameters.
"""
# The return value should be hashable and __eq__ comparable
if isinstance(method, AsDefault):
return AsDefault, (method.method,)
elif isinstance(method, impute.Model):
return impute.Model, (method.learner,)
elif isinstance(method, impute.Default):
return impute.Default, (method.default,)
else:
return type(method), None
methods = set(method_key(m) for m in methods)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
fixed_value = None
value_stack_enabled = False
current_value_widget = None
if len(methods) == 1:
method_type, parameters = methods.pop()
for i, m in enumerate(self.methods):
if method_type == type(m):
self.variable_button_group.button(i).setChecked(True)
if method_type is impute.Default:
(fixed_value,) = parameters
elif self.variable_button_group.checkedButton() is not None:
# Uncheck the current button
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
assert self.variable_button_group.checkedButton() is None
for method, button in zip(self.methods,
self.variable_button_group.buttons()):
enabled = all(method.supports_variable(var) for var in
selected_vars)
button.setEnabled(enabled)
if not has_discrete:
value_stack_enabled = True
current_value_widget = self.value_double
elif len(selected_vars) == 1:
value_stack_enabled = True
current_value_widget = self.value_combo
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
else:
value_stack_enabled = False
current_value_widget = None
self.variable_button_group.button(self.AS_INPUT).setEnabled(False)
self.value_stack.setEnabled(value_stack_enabled)
if current_value_widget is not None:
self.value_stack.setCurrentWidget(current_value_widget)
if fixed_value is not None:
if current_value_widget is self.value_combo:
self.value_combo.setCurrentIndex(fixed_value)
elif current_value_widget is self.value_double:
self.value_double.setValue(fixed_value)
else:
assert False
def set_method_for_current_selection(self, method_index):
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
if method_index == self.DEFAULT:
for index in indexes:
self.variable_methods.pop(index.row(), None)
elif method_index == OWImpute.AS_INPUT:
current = self.value_stack.currentWidget()
if current is self.value_combo:
value = self.value_combo.currentIndex()
else:
value = self.value_double.value()
for index in indexes:
method = impute.Default(default=value)
self.variable_methods[index.row()] = method
else:
method = self.methods[method_index]
for index in indexes:
self.variable_methods[index.row()] = method
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(index, self.get_method_for_column(index.row()), Qt.UserRole)
def _on_value_selected(self):
# The fixed 'Value' in the widget has been changed by the user.
self.variable_button_group.button(self.AS_INPUT).setChecked(True)
self.set_method_for_current_selection(self.AS_INPUT)
def reset_variable_methods(self):
indexes = list(map(self.varmodel.index, range(len(self.varmodel))))
self.set_method_for_indexes(indexes, self.DEFAULT)
self.variable_button_group.button(self.DEFAULT).setChecked(True)
def _add_row(self, parameter: Optional[Parameter] = None):
row_id = len(self.__controls)
if parameter is None:
parameter = Parameter(f"p{row_id + 1}")
edit = QLineEdit(text=parameter.name)
edit.setFixedWidth(60)
edit.textChanged.connect(self.__on_text_changed)
button = gui.button(None,
self,
"×",
callback=self.__on_remove_button_clicked,
autoDefault=False,
width=34,
sizePolicy=(QSizePolicy.Maximum,
QSizePolicy.Maximum))
kwargs = {"minimum": -2147483647, "maximum": 2147483647}
init_spin = QDoubleSpinBox(decimals=4, **kwargs)
lower_spin = QDoubleSpinBox(**kwargs)
upper_spin = QDoubleSpinBox(**kwargs)
init_spin.setValue(parameter.initial)
lower_spin.setValue(parameter.lower)
upper_spin.setValue(parameter.upper)
lower_check = QCheckBox(checked=bool(parameter.use_lower))
upper_check = QCheckBox(checked=bool(parameter.use_upper))
lower_spin.setEnabled(lower_check.isChecked())
upper_spin.setEnabled(upper_check.isChecked())
init_spin.valueChanged.connect(self.__on_init_spin_changed)
lower_spin.valueChanged.connect(self.__on_lower_spin_changed)
upper_spin.valueChanged.connect(self.__on_upper_spin_changed)
lower_check.stateChanged.connect(self.__on_lower_check_changed)
upper_check.stateChanged.connect(self.__on_upper_check_changed)
controls = (button, edit, init_spin, lower_check, lower_spin,
upper_check, upper_spin)
n_rows = self.__layout.rowCount()
for i, control in enumerate(controls):
self.__layout.addWidget(control, n_rows, i)
self.__data.append(parameter)
self.__controls.append(controls)
self._set_labels_visible(True)
class AbsoluteRelativeSpinBox(QWidget):
editingFinished = pyqtSignal()
valueChanged = pyqtSignal()
def __init__(self, *args, **kwargs):
super().__init__(*args)
layout = QStackedLayout(self)
self.double_spin = QDoubleSpinBox()
self.double_spin.valueChanged.connect(self.double_value_changed)
self.double_spin.editingFinished.connect(self.double_editing_finished)
layout.addWidget(self.double_spin)
self.int_spin = QSpinBox()
self.int_spin.setMaximum(10**4)
self.int_spin.valueChanged.connect(self.int_value_changed)
self.int_spin.editingFinished.connect(self.int_editing_finished)
layout.addWidget(self.int_spin)
self.setValue(kwargs.get('value', 0.))
def double_value_changed(self):
if self.double_spin.value() > 1:
self.layout().setCurrentIndex(1)
self.int_spin.setValue(self.double_spin.value())
self.valueChanged.emit()
def double_editing_finished(self):
if self.double_spin.value() <= 1.:
self.editingFinished.emit()
def int_value_changed(self):
if self.int_spin.value() == 0:
self.layout().setCurrentIndex(0)
self.double_spin.setValue(1. - self.double_spin.singleStep())
# There is no need to emit valueChanged signal.
def int_editing_finished(self):
if self.int_spin.value() > 0:
self.editingFinished.emit()
def value(self):
return self.int_spin.value() or self.double_spin.value()
def setValue(self, value):
if isinstance(value, int):
self.layout().setCurrentIndex(1)
self.int_spin.setValue(value)
else:
self.layout().setCurrentIndex(0)
self.double_spin.setValue(value)
def setSingleStep(self, step):
if isinstance(step, float):
self.double_spin.setSingleStep(step)
else:
self.int_spin.setSingleStep(step)
class OWImpute(OWWidget):
name = "Impute"
description = "Impute missing values in the data table."
icon = "icons/Impute.svg"
priority = 2130
inputs = [("Data", Orange.data.Table, "set_data"),
("Learner", Learner, "set_learner")]
outputs = [("Data", Orange.data.Table)]
DEFAULT_LEARNER = SimpleTreeLearner()
METHODS = [
AsDefault(),
impute.DoNotImpute(),
impute.Average(),
impute.AsValue(),
impute.Model(DEFAULT_LEARNER),
impute.Random(),
impute.DropInstances(),
impute.Default()
]
DEFAULT, DO_NOT_IMPUTE, MODEL_BASED_IMPUTER, AS_INPUT = 0, 1, 4, 7
settingsHandler = settings.DomainContextHandler()
_default_method_index = settings.Setting(DO_NOT_IMPUTE)
variable_methods = settings.ContextSetting({})
autocommit = settings.Setting(False)
default_value = settings.Setting(0.)
want_main_area = False
resizing_enabled = False
def __init__(self):
super().__init__()
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = QGroupBox(title=self.tr("Default Method"), flat=False)
box_layout = QVBoxLayout(box)
main_layout.addWidget(box)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, method in enumerate(self.METHODS):
if not method.columns_only:
button = QRadioButton(method.name)
button.setChecked(i == self.default_method_index)
button_group.addButton(button, i)
box_layout.addWidget(button)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = QListView(selectionMode=QListView.ExtendedSelection)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
method_layout = QVBoxLayout()
horizontal_layout.addLayout(method_layout)
button_group = QButtonGroup()
for i, method in enumerate(self.METHODS):
button = QRadioButton(text=method.name)
button_group.addButton(button, i)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected)
self.value_combo.currentIndexChanged.connect(self._on_value_changed)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000.,
maximum=1000.,
singleStep=.1,
decimals=3,
value=self.default_value)
self.value_stack = value_stack = QStackedLayout()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addLayout(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection)
method_layout.addStretch(2)
reset_button = QPushButton("Restore All to Default",
checked=False,
checkable=False,
clicked=self.reset_variable_methods,
default=False,
autoDefault=False)
method_layout.addWidget(reset_button)
self.variable_button_group = button_group
box = gui.auto_commit(self.controlArea,
self,
"autocommit",
"Apply",
orientation=Qt.Horizontal,
checkbox_label="Apply automatically")
box.layout().insertSpacing(0, 80)
box.layout().insertWidget(0, self.report_button)
self.data = None
self.modified = False
self.default_method = self.METHODS[self.default_method_index]
self.update_varview()
@property
def default_method_index(self):
return self._default_method_index
@default_method_index.setter
def default_method_index(self, index):
if self._default_method_index != index:
self._default_method_index = index
self.default_button_group.button(index).setChecked(True)
self.default_method = self.METHODS[self.default_method_index]
self.METHODS[self.DEFAULT].method = self.default_method
# update variable view
for index in map(self.varmodel.index, range(len(self.varmodel))):
self.varmodel.setData(
index,
self.variable_methods.get(index.row(),
self.METHODS[self.DEFAULT]),
Qt.UserRole)
self._invalidate()
def set_default_method(self, index):
"""Set the current selected default imputation method.
"""
self.default_method_index = index
@check_sql_input
def set_data(self, data):
self.closeContext()
self.varmodel[:] = []
self.variable_methods = {}
self.modified = False
self.data = data
if data is not None:
self.varmodel[:] = data.domain.variables
self.openContext(data.domain)
self.update_varview()
self.unconditional_commit()
def set_learner(self, learner):
self.learner = learner or self.DEFAULT_LEARNER
imputer = self.METHODS[self.MODEL_BASED_IMPUTER]
imputer.learner = self.learner
button = self.default_button_group.button(self.MODEL_BASED_IMPUTER)
button.setText(imputer.name)
variable_button = self.variable_button_group.button(
self.MODEL_BASED_IMPUTER)
variable_button.setText(imputer.name)
if learner is not None:
self.default_method_index = self.MODEL_BASED_IMPUTER
self.commit()
def get_method_for_column(self, column_index):
"""Returns the imputation method for column by its index.
"""
if not isinstance(column_index, int):
column_index = column_index.row()
return self.variable_methods.get(column_index,
self.METHODS[self.DEFAULT])
def _invalidate(self):
self.modified = True
self.commit()
def commit(self):
data = self.data
if self.data is not None:
drop_mask = np.zeros(len(self.data), bool)
attributes = []
class_vars = []
self.warning()
with self.progressBar(len(self.varmodel)) as progress:
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, self.default_method)
if not method.supports_variable(var):
self.warning(
"Default method can not handle '{}'".format(
var.name))
elif isinstance(method, impute.DropInstances):
drop_mask |= method(self.data, var)
else:
var = method(self.data, var)
if isinstance(var, Orange.data.Variable):
var = [var]
if i < len(self.data.domain.attributes):
attributes.extend(var)
else:
class_vars.extend(var)
progress.advance()
domain = Orange.data.Domain(attributes, class_vars,
self.data.domain.metas)
data = self.data.from_table(domain, self.data[~drop_mask])
self.send("Data", data)
self.modified = False
def send_report(self):
specific = []
for i, var in enumerate(self.varmodel):
method = self.variable_methods.get(i, None)
if method is not None:
specific.append("{} ({})".format(var.name, str(method)))
default = self.default_method.name
if specific:
self.report_items((("Default method", default),
("Specific imputers", ", ".join(specific))))
else:
self.report_items((("Method", default), ))
def _on_var_selection_changed(self):
indexes = self.selection.selectedIndexes()
methods = set(
self.get_method_for_column(i.row()).name for i in indexes)
selected_vars = [self.varmodel[index.row()] for index in indexes]
has_discrete = any(var.is_discrete for var in selected_vars)
if len(methods) == 1:
method = methods.pop()
for i, m in enumerate(self.METHODS):
if method == m.name:
self.variable_button_group.button(i).setChecked(True)
elif self.variable_button_group.checkedButton() is not None:
self.variable_button_group.setExclusive(False)
self.variable_button_group.checkedButton().setChecked(False)
self.variable_button_group.setExclusive(True)
for method, button in zip(self.METHODS,
self.variable_button_group.buttons()):
enabled = all(
method.supports_variable(var) for var in selected_vars)
button.setEnabled(enabled)
if not has_discrete:
self.value_stack.setEnabled(True)
self.value_stack.setCurrentWidget(self.value_double)
self._on_value_changed()
elif len(selected_vars) == 1:
self.value_stack.setEnabled(True)
self.value_stack.setCurrentWidget(self.value_combo)
self.value_combo.clear()
self.value_combo.addItems(selected_vars[0].values)
self._on_value_changed()
else:
self.variable_button_group.button(self.AS_INPUT).setEnabled(False)
self.value_stack.setEnabled(False)
def set_method_for_current_selection(self, method_index):
indexes = self.selection.selectedIndexes()
self.set_method_for_indexes(indexes, method_index)
def set_method_for_indexes(self, indexes, method_index):
if method_index == self.DEFAULT:
for index in indexes:
self.variable_methods.pop(index, None)
else:
method = self.METHODS[method_index].copy()
for index in indexes:
self.variable_methods[index.row()] = method
self.update_varview(indexes)
self._invalidate()
def update_varview(self, indexes=None):
if indexes is None:
indexes = map(self.varmodel.index, range(len(self.varmodel)))
for index in indexes:
self.varmodel.setData(index,
self.get_method_for_column(index.row()),
Qt.UserRole)
def _on_value_selected(self):
self.variable_button_group.button(self.AS_INPUT).setChecked(True)
self._on_value_changed()
def _on_value_changed(self):
widget = self.value_stack.currentWidget()
if widget is self.value_combo:
value = self.value_combo.currentText()
else:
value = self.value_double.value()
self.default_value = value
self.METHODS[self.AS_INPUT].default = value
index = self.variable_button_group.checkedId()
if index == self.AS_INPUT:
self.set_method_for_current_selection(index)
def reset_variable_methods(self):
indexes = map(self.varmodel.index, range(len(self.varmodel)))
self.set_method_for_indexes(indexes, self.DEFAULT)
self.variable_button_group.button(self.DEFAULT).setChecked(True)
def __init__(self):
super().__init__()
self.data = None # type: Optional[Orange.data.Table]
self.learner = None # type: Optional[Learner]
self.default_learner = SimpleTreeLearner(min_instances=10,
max_depth=10)
self.modified = False
self.executor = qconcurrent.ThreadExecutor(self)
self.__task = None
main_layout = QVBoxLayout()
main_layout.setContentsMargins(10, 10, 10, 10)
self.controlArea.layout().addLayout(main_layout)
box = gui.vBox(None, "Default Method")
main_layout.addWidget(box)
box_layout = QGridLayout()
box_layout.setSpacing(8)
box.layout().addLayout(box_layout)
button_group = QButtonGroup()
button_group.buttonClicked[int].connect(self.set_default_method)
for i, (method, _) in enumerate(list(METHODS.items())[1:-1]):
imputer = self.create_imputer(method)
button = QRadioButton(imputer.name)
button.setChecked(method == self.default_method_index)
button_group.addButton(button, method)
box_layout.addWidget(button, i % 3, i // 3)
def set_default_time(datetime):
datetime = datetime.toSecsSinceEpoch()
if datetime != self.default_time:
self.default_time = datetime
if self.default_method_index == Method.Default:
self._invalidate()
hlayout = QHBoxLayout()
box.layout().addLayout(hlayout)
button = QRadioButton("Fixed values; numeric variables:")
button_group.addButton(button, Method.Default)
button.setChecked(Method.Default == self.default_method_index)
hlayout.addWidget(button)
locale = QLocale()
locale.setNumberOptions(locale.NumberOption.RejectGroupSeparator)
validator = QDoubleValidator()
validator.setLocale(locale)
self.numeric_value_widget = le = gui.lineEdit(
None,
self,
"default_numeric",
validator=validator,
alignment=Qt.AlignRight,
callback=self._invalidate,
enabled=self.default_method_index == Method.Default)
hlayout.addWidget(le)
hlayout.addWidget(QLabel(", time:"))
self.time_widget = gui.DateTimeEditWCalendarTime(self)
self.time_widget.setEnabled(
self.default_method_index == Method.Default)
self.time_widget.setKeyboardTracking(False)
self.time_widget.setContentsMargins(0, 0, 0, 0)
self.time_widget.set_datetime(
QDateTime.fromSecsSinceEpoch(self.default_time))
self.connect_control(
"default_time", lambda value: self.time_widget.set_datetime(
QDateTime.fromSecsSinceEpoch(value)))
self.time_widget.dateTimeChanged.connect(set_default_time)
hlayout.addWidget(self.time_widget)
self.default_button_group = button_group
box = QGroupBox(title=self.tr("Individual Attribute Settings"),
flat=False)
main_layout.addWidget(box)
horizontal_layout = QHBoxLayout(box)
main_layout.addWidget(box)
self.varview = ListViewSearch(
selectionMode=QListView.ExtendedSelection, uniformItemSizes=True)
self.varview.setItemDelegate(DisplayFormatDelegate())
self.varmodel = itemmodels.VariableListModel()
self.varview.setModel(self.varmodel)
self.varview.selectionModel().selectionChanged.connect(
self._on_var_selection_changed)
self.selection = self.varview.selectionModel()
horizontal_layout.addWidget(self.varview)
vertical_layout = QVBoxLayout(margin=0)
self.methods_container = QWidget(enabled=False)
method_layout = QVBoxLayout(margin=0)
self.methods_container.setLayout(method_layout)
button_group = QButtonGroup()
for method in Method:
imputer = self.create_imputer(method)
button = QRadioButton(text=imputer.name)
button_group.addButton(button, method)
method_layout.addWidget(button)
self.value_combo = QComboBox(
minimumContentsLength=8,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength,
activated=self._on_value_selected)
self.value_double = QDoubleSpinBox(
editingFinished=self._on_value_selected,
minimum=-1000.,
maximum=1000.,
singleStep=.1,
decimals=3,
)
self.value_stack = value_stack = QStackedWidget()
value_stack.addWidget(self.value_combo)
value_stack.addWidget(self.value_double)
method_layout.addWidget(value_stack)
button_group.buttonClicked[int].connect(
self.set_method_for_current_selection)
self.reset_button = QPushButton(
"Restore All to Default",
enabled=False,
default=False,
autoDefault=False,
clicked=self.reset_variable_state,
)
vertical_layout.addWidget(self.methods_container)
vertical_layout.addStretch(2)
vertical_layout.addWidget(self.reset_button)
horizontal_layout.addLayout(vertical_layout)
self.variable_button_group = button_group
box = gui.auto_apply(self.controlArea, self, "autocommit")
box.button.setFixedWidth(180)
box.layout().insertStretch(0)
self.info.set_input_summary(self.info.NoInput)
self.info.set_output_summary(self.info.NoOutput)
