class DropDownRadioBooleanFilter(QWidget, Control):
"""Container for multiple boolean filters
"""
def __init__(self, tree, dataset, master, parent=None):
QWidget.__init__(self, parent)
Control.__init__(self, tree, dataset, master)
self.setLayout(QHBoxLayout())
self.cb = QComboBox(self)
self.layout().addWidget(self.cb)
rblayout = QVBoxLayout()
self.radioButtons = [QRadioButton("Only", self),
QRadioButton("Excluded", self)
]
for b in self.radioButtons:
rblayout.addWidget(b)
self.radioButtons[0].setChecked(True)
self.layout().addLayout(rblayout)
self.options = []
self.setOptions(tree.subelements_top("Option"))
def setOptions(self, options):
self.cb.clear()
self.options = []
for option in options:
self.cb.addItem(option.displayName)
self.options.append(option)
for op, rb in zip(self.options[0].subelements_top("Option"),
self.radioButtons):
rb.setText(op.displayName)
rb.setChecked(getattr(op, "default", "false") == "true")
def value(self):
return {"excluded": "0" if self.radioButtons[0].isChecked() else "1"}
def query(self):
filter = self.options[self.cb.currentIndex()]
filter = biomart.FilterDescription(
self.tree.registry, "FilterDescription",
filter.attributes, filter.children)
return [("Filter", filter, self.value())]
def setControlValue(self, name, value):
for i, option in enumerate(self.options):
if option.internalName == name:
self.cb.setCurrentIndex(i)
if value == "Only":
self.radioButtons[0].setChecked(True)
class DropDownRadioBooleanFilter(QWidget, Control):
"""Container for multiple boolean filters
"""
def __init__(self, tree, dataset, master, parent=None):
QWidget.__init__(self, parent)
Control.__init__(self, tree, dataset, master)
self.setLayout(QHBoxLayout())
self.cb = QComboBox(self)
self.layout().addWidget(self.cb)
rblayout = QVBoxLayout()
self.radioButtons = [QRadioButton("Only", self),
QRadioButton("Excluded", self)
]
for b in self.radioButtons:
rblayout.addWidget(b)
self.radioButtons[0].setChecked(True)
self.layout().addLayout(rblayout)
self.options = []
self.setOptions(tree.subelements_top("Option"))
def setOptions(self, options):
self.cb.clear()
self.options = []
for option in options:
self.cb.addItem(option.displayName)
self.options.append(option)
for op, rb in zip(self.options[0].subelements_top("Option"),
self.radioButtons):
rb.setText(op.displayName)
rb.setChecked(getattr(op, "default", "false") == "true")
def value(self):
return {"excluded": "0" if self.radioButtons[0].isChecked() else "1"}
def query(self):
filter = self.options[self.cb.currentIndex()]
filter = biomart.FilterDescription(
self.tree.registry, "FilterDescription",
filter.attributes, filter.children)
return [("Filter", filter, self.value())]
def setControlValue(self, name, value):
for i, option in enumerate(self.options):
if option.internalName == name:
self.cb.setCurrentIndex(i)
if value == "Only":
self.radioButtons[0].setChecked(True)
class RecentPathsWComboMixin(RecentPathsWidgetMixin):
"""
Adds file combo handling to :obj:`RecentPathsWidgetMixin`.
The mixin constructs a combo box `self.file_combo` and provides a method
`set_file_list` for updating its content. The mixin also overloads the
inherited `add_path` and `select_file` to call `set_file_list`.
"""
def __init__(self):
super().__init__()
self.file_combo = \
QComboBox(self, sizeAdjustPolicy=QComboBox.AdjustToContents)
def add_path(self, filename):
"""Add (or move) a file name to the top of recent paths"""
super().add_path(filename)
self.set_file_list()
def select_file(self, n):
"""Move the n-th file to the top of the list"""
super().select_file(n)
self.set_file_list()
def set_file_list(self):
"""
Sets the items in the file list combo
"""
self._check_init()
self.file_combo.clear()
if not self.recent_paths:
self.file_combo.addItem("(none)")
self.file_combo.model().item(0).setEnabled(False)
else:
for i, recent in enumerate(self.recent_paths):
self.file_combo.addItem(recent.basename)
self.file_combo.model().item(i).setToolTip(recent.abspath)
if not os.path.exists(recent.abspath):
self.file_combo.setItemData(i, QBrush(Qt.red),
Qt.TextColorRole)
def workflowEnvChanged(self, key, value, oldvalue):
super().workflowEnvChanged(key, value, oldvalue)
if key == "basedir":
self.set_file_list()
class DropDownIdListFilter(QWidget, Control):
"""Container for multiple id list filters
"""
def __init__(self, tree, dataset, master, parent=None):
QWidget.__init__(self, parent)
Control.__init__(self, tree, dataset, master)
self.setLayout(QVBoxLayout())
self.setContentsMargins(0, 0, 0, 0)
self.cb = QComboBox()
self.idsEdit = QPlainTextEdit()
self.layout().addWidget(self.cb)
self.layout().addWidget(self.idsEdit)
self.options = []
self.setOptions(tree.subelements_top("Option"))
def setOptions(self, options):
self.cb.clear()
self.options = []
for option in options:
self.cb.addItem(option.displayName)
self.options.append(option)
def value(self):
return str(self.idsEdit.toPlainText()).split()
def query(self):
filter = self.options[self.cb.currentIndex()]
filter = biomart.FilterDescription(
self.tree.registry, "FilterDescription",
filter.attributes, filter.children)
return [("Filter", filter, self.value())]
def setControlValue(self, name, value):
if isinstance(value, list):
value = "\n".join(value)
for i, op in enumerate(self.options):
if name == op.internalName:
self.cb.setCurrentIndex(i)
self.idsEdit.setPlainText(value)
class DropDownIdListFilter(QWidget, Control):
"""Container for multiple id list filters
"""
def __init__(self, tree, dataset, master, parent=None):
QWidget.__init__(self, parent)
Control.__init__(self, tree, dataset, master)
self.setLayout(QVBoxLayout())
self.setContentsMargins(0, 0, 0, 0)
self.cb = QComboBox()
self.idsEdit = QPlainTextEdit()
self.layout().addWidget(self.cb)
self.layout().addWidget(self.idsEdit)
self.options = []
self.setOptions(tree.subelements_top("Option"))
def setOptions(self, options):
self.cb.clear()
self.options = []
for option in options:
self.cb.addItem(option.displayName)
self.options.append(option)
def value(self):
return str(self.idsEdit.toPlainText()).split()
def query(self):
filter = self.options[self.cb.currentIndex()]
filter = biomart.FilterDescription(
self.tree.registry, "FilterDescription",
filter.attributes, filter.children)
return [("Filter", filter, self.value())]
def setControlValue(self, name, value):
if isinstance(value, list):
value = "\n".join(value)
for i, op in enumerate(self.options):
if name == op.internalName:
self.cb.setCurrentIndex(i)
self.idsEdit.setPlainText(value)
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)
class ControlCombo(ControlBase, QWidget):
"""This class represents a wrapper to the combo box"""
def __init__(self, label='', default=None, helptext=None):
QWidget.__init__(self)
ControlBase.__init__(self, label, default, helptext)
##########################################################################
############ Functions ###################################################
##########################################################################
def init_form(self):
self._layout = QHBoxLayout()
self._combo = QComboBox(self.form)
if self._label is not None:
self._combolabel = QLabel(self.form)
self._layout.addWidget(self._combolabel)
self._combolabel.setAccessibleName('ControlCombo-label')
self.label = self._label
else:
self._combolabel = None
self._layout.addWidget(self._combo)
self.form.setLayout(self._layout)
self._combo.setSizePolicy(QSizePolicy.Expanding, QSizePolicy.Fixed)
self._layout.setContentsMargins(0, 0, 0, 0)
self.form.setContentsMargins(0, 0, 0, 0)
self.form.setMinimumHeight(38)
self.form.setMaximumHeight(38)
self.form.setSizePolicy(QSizePolicy.Preferred, QSizePolicy.Fixed)
self._combo.currentIndexChanged.connect(self._currentIndexChanged)
self._combo.activated.connect(self._activated)
self._combo.highlighted.connect(self._highlighted)
self._combo.editTextChanged.connect(self._editTextChanged)
self._items = {}
self._addingItem = False
def clear(self):
self._items = {}
self._value = None
self._combo.clear()
def add_item(self, label, value=None):
self._addingItem = True
if value is not None:
if not (value in self._items.values()):
self._combo.addItem(label)
else:
if not (label in self._items.keys()):
self._combo.addItem(label)
firstValue = False
if self._items == {}:
firstValue = True
if value is None:
self._items[str(label)] = label
else:
self._items[str(label)] = value
self._addingItem = False
if firstValue:
self.value = self._items[label]
def __add__(self, val):
if isinstance(val, tuple):
self.add_item(val[0], val[1])
else:
self.add_item(val)
return self
def get_item_index_by_name(self, item_name):
"""
Returns the index of the item containing the given name
:param item_name: item name in combo box
:type item_name: string
"""
return self._combo.findText(item_name)
def count(self):
return self._combo.count()
def show(self):
"""
Show the control
"""
QWidget.show(self)
def hide(self):
"""
Hide the control
"""
QWidget.hide(self)
##########################################################################
############ Events ######################################################
##########################################################################
def current_index_changed_event(self, index):
"""Called when the user chooses an item in the combobox and
the selected choice is different from the last one selected.
@index: item's index
"""
pass
def activated_event(self, index):
"""Called when the user chooses an item in the combobox.
Note that this signal happens even when the choice is not changed
@index: item's index
"""
pass
def highlighted_event(self, index):
pass
def edittext_changed_event(self, text):
pass
##########################################################################
############ PROPERTIES ##################################################
##########################################################################
@property
def form(self):
return self
@property
def current_index(self):
return self._combo.currentIndex()
@current_index.setter
def current_index(self, value):
self._combo.setCurrentIndex(value)
@property
def items(self):
return self._items.items()
@property
def value(self):
return self._value
@value.setter
def value(self, value):
for key, val in self.items:
if value == val:
index = self._combo.findText(key)
self._combo.setCurrentIndex(index)
if self._value != value:
self.changed_event()
self._value = val
@property
def text(self):
return str(self._combo.currentText())
@text.setter
def text(self, value):
for key, val in self.items:
if value == key:
self.value = val
break
@property
def label(self):
if self._combolabel:
return self._combolabel.text()
else:
return None
@label.setter
def label(self, value):
"""
Label of the control, if applies
@type value: string
"""
if self._combolabel:
self._combolabel.setText(value)
##########################################################################
############ Private functions ###########################################
##########################################################################
def _activated(self, index):
if not self._addingItem:
item = self._combo.currentText()
if len(item) >= 1:
ControlBase.value.fset(self, self._items[str(item)])
self.activated_event(index)
def _highlighted(self, index):
"""Called when an item in the combobox popup
list is highlighted by the user.
@index: item's index
"""
self.highlighted_event(index)
def _editTextChanged(self, text):
self.edittext_changed_event(text)
def _currentIndexChanged(self, index):
if not self._addingItem:
item = self._combo.currentText()
if len(item) >= 1:
ControlBase.value.fset(self, self._items[str(item)])
self.current_index_changed_event(index)
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()
class OWSql(OWWidget):
name = "SQL Table"
id = "orange.widgets.data.sql"
description = "Load data set from SQL."
icon = "icons/SQLTable.svg"
priority = 10
category = "Data"
keywords = ["data", "file", "load", "read"]
class Outputs:
data = Output(
"Data",
Table,
doc="Attribute-valued data set read from the input file.")
settings_version = 2
want_main_area = False
resizing_enabled = False
host = Setting(None)
port = Setting(None)
database = Setting(None)
schema = Setting(None)
username = ""
password = ""
table = Setting(None)
sql = Setting("")
guess_values = Setting(True)
download = Setting(False)
materialize = Setting(False)
materialize_table_name = Setting("")
class Information(OWWidget.Information):
data_sampled = Msg("Data description was generated from a sample.")
class Error(OWWidget.Error):
connection = Msg("{}")
no_backends = Msg("Please install a backend to use this widget")
missing_extension = Msg("Database is missing extension{}: {}")
def __init__(self):
super().__init__()
self.backend = None
self.data_desc_table = None
self.database_desc = None
vbox = gui.vBox(self.controlArea, "Server", addSpace=True)
box = gui.vBox(vbox)
self.backendmodel = BackendModel(Backend.available_backends())
self.backendcombo = QComboBox(box)
if len(self.backendmodel):
self.backendcombo.setModel(self.backendmodel)
else:
self.Error.no_backends()
box.setEnabled(False)
box.layout().addWidget(self.backendcombo)
self.servertext = QLineEdit(box)
self.servertext.setPlaceholderText('Server')
self.servertext.setToolTip('Server')
self.servertext.editingFinished.connect(self._load_credentials)
if self.host:
self.servertext.setText(self.host if not self.port else '{}:{}'.
format(self.host, self.port))
box.layout().addWidget(self.servertext)
self.databasetext = QLineEdit(box)
self.databasetext.setPlaceholderText('Database[/Schema]')
self.databasetext.setToolTip('Database or optionally Database/Schema')
if self.database:
self.databasetext.setText(
self.database if not self.schema else '{}/{}'.
format(self.database, self.schema))
box.layout().addWidget(self.databasetext)
self.usernametext = QLineEdit(box)
self.usernametext.setPlaceholderText('Username')
self.usernametext.setToolTip('Username')
box.layout().addWidget(self.usernametext)
self.passwordtext = QLineEdit(box)
self.passwordtext.setPlaceholderText('Password')
self.passwordtext.setToolTip('Password')
self.passwordtext.setEchoMode(QLineEdit.Password)
box.layout().addWidget(self.passwordtext)
self._load_credentials()
tables = gui.hBox(box)
self.tablemodel = TableModel()
self.tablecombo = QComboBox(
minimumContentsLength=35,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength)
self.tablecombo.setModel(self.tablemodel)
self.tablecombo.setToolTip('table')
tables.layout().addWidget(self.tablecombo)
self.tablecombo.activated[int].connect(self.select_table)
self.connectbutton = gui.button(tables,
self,
'↻',
callback=self.connect)
self.connectbutton.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
tables.layout().addWidget(self.connectbutton)
self.custom_sql = gui.vBox(box)
self.custom_sql.setVisible(False)
self.sqltext = QTextEdit(self.custom_sql)
self.sqltext.setPlainText(self.sql)
self.custom_sql.layout().addWidget(self.sqltext)
mt = gui.hBox(self.custom_sql)
cb = gui.checkBox(mt, self, 'materialize', 'Materialize to table ')
cb.setToolTip('Save results of the query in a table')
le = gui.lineEdit(mt, self, 'materialize_table_name')
le.setToolTip('Save results of the query in a table')
self.executebtn = gui.button(self.custom_sql,
self,
'Execute',
callback=self.open_table)
box.layout().addWidget(self.custom_sql)
gui.checkBox(box,
self,
"guess_values",
"Auto-discover discrete variables",
callback=self.open_table)
gui.checkBox(box,
self,
"download",
"Download data to local memory",
callback=self.open_table)
gui.rubber(self.buttonsArea)
QTimer.singleShot(0, self.connect)
def _load_credentials(self):
self._parse_host_port()
cm = self._credential_manager(self.host, self.port)
self.username = cm.username
self.password = cm.password
if self.username:
self.usernametext.setText(self.username)
if self.password:
self.passwordtext.setText(self.password)
def _save_credentials(self):
cm = self._credential_manager(self.host, self.port)
cm.username = self.username
cm.password = self.password
def _credential_manager(self, host, port):
return CredentialManager("SQL Table: {}:{}".format(host, port))
def error(self, id=0, text=""):
super().error(id, text)
err_style = 'QLineEdit {border: 2px solid red;}'
if 'server' in text or 'host' in text:
self.servertext.setStyleSheet(err_style)
else:
self.servertext.setStyleSheet('')
if 'role' in text:
self.usernametext.setStyleSheet(err_style)
else:
self.usernametext.setStyleSheet('')
if 'database' in text:
self.databasetext.setStyleSheet(err_style)
else:
self.databasetext.setStyleSheet('')
def _parse_host_port(self):
hostport = self.servertext.text().split(':')
self.host = hostport[0]
self.port = hostport[1] if len(hostport) == 2 else None
def connect(self):
self._parse_host_port()
self.database, _, self.schema = self.databasetext.text().partition('/')
self.username = self.usernametext.text() or None
self.password = self.passwordtext.text() or None
try:
if self.backendcombo.currentIndex() < 0:
return
backend = self.backendmodel[self.backendcombo.currentIndex()]
self.backend = backend(
dict(host=self.host,
port=self.port,
database=self.database,
user=self.username,
password=self.password))
self.Error.connection.clear()
self._save_credentials()
self.database_desc = OrderedDict(
(("Host", self.host), ("Port", self.port),
("Database", self.database), ("User name", self.username)))
self.refresh_tables()
self.select_table()
except BackendError as err:
error = str(err).split('\n')[0]
self.Error.connection(error)
self.database_desc = self.data_desc_table = None
self.tablecombo.clear()
def refresh_tables(self):
self.tablemodel.clear()
self.Error.missing_extension.clear()
if self.backend is None:
self.data_desc_table = None
return
self.tablemodel.append("Select a table")
self.tablemodel.extend(self.backend.list_tables(self.schema))
self.tablemodel.append("Custom SQL")
def select_table(self):
curIdx = self.tablecombo.currentIndex()
if self.tablecombo.itemText(curIdx) != "Custom SQL":
self.custom_sql.setVisible(False)
return self.open_table()
else:
self.custom_sql.setVisible(True)
self.data_desc_table = None
self.database_desc["Table"] = "(None)"
self.table = None
#self.Error.missing_extension(
# 's' if len(missing) > 1 else '',
# ', '.join(missing),
# shown=missing)
def open_table(self):
table = self.get_table()
self.data_desc_table = table
self.Outputs.data.send(table)
def get_table(self):
if self.tablecombo.currentIndex() <= 0:
if self.database_desc:
self.database_desc["Table"] = "(None)"
self.data_desc_table = None
return
if self.tablecombo.currentIndex() < self.tablecombo.count() - 1:
self.table = self.tablemodel[self.tablecombo.currentIndex()]
self.database_desc["Table"] = self.table
if "Query" in self.database_desc:
del self.database_desc["Query"]
else:
self.sql = self.table = self.sqltext.toPlainText()
if self.materialize:
import psycopg2
if not self.materialize_table_name:
self.Error.connection(
"Specify a table name to materialize the query")
return
try:
with self.backend.execute_sql_query(
"DROP TABLE IF EXISTS " +
self.materialize_table_name):
pass
with self.backend.execute_sql_query(
"CREATE TABLE " + self.materialize_table_name +
" AS " + self.table):
pass
with self.backend.execute_sql_query(
"ANALYZE " + self.materialize_table_name):
pass
self.table = self.materialize_table_name
except (psycopg2.ProgrammingError, BackendError) as ex:
self.Error.connection(str(ex))
return
try:
table = SqlTable(dict(host=self.host,
port=self.port,
database=self.database,
user=self.username,
password=self.password),
self.table,
backend=type(self.backend),
inspect_values=False)
except BackendError as ex:
self.Error.connection(str(ex))
return
self.Error.connection.clear()
sample = False
if table.approx_len() > LARGE_TABLE and self.guess_values:
confirm = QMessageBox(self)
confirm.setIcon(QMessageBox.Warning)
confirm.setText("Attribute discovery might take "
"a long time on large tables.\n"
"Do you want to auto discover attributes?")
confirm.addButton("Yes", QMessageBox.YesRole)
no_button = confirm.addButton("No", QMessageBox.NoRole)
sample_button = confirm.addButton("Yes, on a sample",
QMessageBox.YesRole)
confirm.exec()
if confirm.clickedButton() == no_button:
self.guess_values = False
elif confirm.clickedButton() == sample_button:
sample = True
self.Information.clear()
if self.guess_values:
QApplication.setOverrideCursor(QCursor(Qt.WaitCursor))
if sample:
s = table.sample_time(1)
domain = s.get_domain(inspect_values=True)
self.Information.data_sampled()
else:
domain = table.get_domain(inspect_values=True)
QApplication.restoreOverrideCursor()
table.domain = domain
if self.download:
if table.approx_len() > MAX_DL_LIMIT:
QMessageBox.warning(
self, 'Warning', "Data is too big to download.\n"
"Consider using the Data Sampler widget to download "
"a sample instead.")
self.download = False
elif table.approx_len() > AUTO_DL_LIMIT:
confirm = QMessageBox.question(
self, 'Question', "Data appears to be big. Do you really "
"want to download it to local memory?",
QMessageBox.Yes | QMessageBox.No, QMessageBox.No)
if confirm == QMessageBox.No:
self.download = False
if self.download:
table.download_data(MAX_DL_LIMIT)
table = Table(table)
return table
def send_report(self):
if not self.database_desc:
self.report_paragraph("No database connection.")
return
self.report_items("Database", self.database_desc)
if self.data_desc_table:
self.report_items("Data",
report.describe_data(self.data_desc_table))
@classmethod
def migrate_settings(cls, settings, version):
if version < 2:
# Until Orange version 3.4.4 username and password had been stored
# in Settings.
cm = cls._credential_manager(settings["host"], settings["port"])
cm.username = settings["username"]
cm.password = settings["password"]
class OWSql(OWWidget):
name = "SQL Table"
id = "orange.widgets.data.sql"
description = "Load data set from SQL."
icon = "icons/SQLTable.svg"
priority = 30
category = "Data"
keywords = ["data", "file", "load", "read", "SQL"]
class Outputs:
data = Output("Data", Table, doc="Attribute-valued data set read from the input file.")
settings_version = 2
want_main_area = False
resizing_enabled = False
host = Setting(None)
port = Setting(None)
database = Setting(None)
schema = Setting(None)
username = ""
password = ""
table = Setting(None)
sql = Setting("")
guess_values = Setting(True)
download = Setting(False)
materialize = Setting(False)
materialize_table_name = Setting("")
class Information(OWWidget.Information):
data_sampled = Msg("Data description was generated from a sample.")
class Error(OWWidget.Error):
connection = Msg("{}")
no_backends = Msg("Please install a backend to use this widget")
missing_extension = Msg("Database is missing extension{}: {}")
def __init__(self):
super().__init__()
self.backend = None
self.data_desc_table = None
self.database_desc = None
vbox = gui.vBox(self.controlArea, "Server", addSpace=True)
box = gui.vBox(vbox)
self.backends = BackendModel(Backend.available_backends())
self.backendcombo = QComboBox(box)
if len(self.backends):
self.backendcombo.setModel(self.backends)
else:
self.Error.no_backends()
box.setEnabled(False)
box.layout().addWidget(self.backendcombo)
self.servertext = QLineEdit(box)
self.servertext.setPlaceholderText('Server')
self.servertext.setToolTip('Server')
self.servertext.editingFinished.connect(self._load_credentials)
if self.host:
self.servertext.setText(self.host if not self.port else
'{}:{}'.format(self.host, self.port))
box.layout().addWidget(self.servertext)
self.databasetext = QLineEdit(box)
self.databasetext.setPlaceholderText('Database[/Schema]')
self.databasetext.setToolTip('Database or optionally Database/Schema')
if self.database:
self.databasetext.setText(
self.database if not self.schema else
'{}/{}'.format(self.database, self.schema))
box.layout().addWidget(self.databasetext)
self.usernametext = QLineEdit(box)
self.usernametext.setPlaceholderText('Username')
self.usernametext.setToolTip('Username')
box.layout().addWidget(self.usernametext)
self.passwordtext = QLineEdit(box)
self.passwordtext.setPlaceholderText('Password')
self.passwordtext.setToolTip('Password')
self.passwordtext.setEchoMode(QLineEdit.Password)
box.layout().addWidget(self.passwordtext)
self._load_credentials()
self.tables = TableModel()
tables = gui.hBox(box)
self.tablecombo = QComboBox(
minimumContentsLength=35,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength
)
self.tablecombo.setModel(self.tables)
self.tablecombo.setToolTip('table')
tables.layout().addWidget(self.tablecombo)
self.connect()
index = self.tablecombo.findText(str(self.table))
if index != -1:
self.tablecombo.setCurrentIndex(index)
# set up the callback to select_table in case of selection change
self.tablecombo.activated[int].connect(self.select_table)
self.connectbutton = gui.button(
tables, self, '↻', callback=self.connect)
self.connectbutton.setSizePolicy(
QSizePolicy.Fixed, QSizePolicy.Fixed)
tables.layout().addWidget(self.connectbutton)
self.custom_sql = gui.vBox(box)
self.custom_sql.setVisible(False)
self.sqltext = QTextEdit(self.custom_sql)
self.sqltext.setPlainText(self.sql)
self.custom_sql.layout().addWidget(self.sqltext)
mt = gui.hBox(self.custom_sql)
cb = gui.checkBox(mt, self, 'materialize', 'Materialize to table ')
cb.setToolTip('Save results of the query in a table')
le = gui.lineEdit(mt, self, 'materialize_table_name')
le.setToolTip('Save results of the query in a table')
self.executebtn = gui.button(
self.custom_sql, self, 'Execute', callback=self.open_table)
box.layout().addWidget(self.custom_sql)
gui.checkBox(box, self, "guess_values",
"Auto-discover categorical variables",
callback=self.open_table)
gui.checkBox(box, self, "download",
"Download data to local memory",
callback=self.open_table)
gui.rubber(self.buttonsArea)
QTimer.singleShot(0, self.select_table)
def _load_credentials(self):
self._parse_host_port()
cm = self._credential_manager(self.host, self.port)
self.username = cm.username
self.password = cm.password
if self.username:
self.usernametext.setText(self.username)
if self.password:
self.passwordtext.setText(self.password)
def _save_credentials(self):
cm = self._credential_manager(self.host, self.port)
cm.username = self.username or ''
cm.password = self.password or ''
def _credential_manager(self, host, port):
return CredentialManager("SQL Table: {}:{}".format(host, port))
def error(self, id=0, text=""):
super().error(id, text)
err_style = 'QLineEdit {border: 2px solid red;}'
if 'server' in text or 'host' in text:
self.servertext.setStyleSheet(err_style)
else:
self.servertext.setStyleSheet('')
if 'role' in text:
self.usernametext.setStyleSheet(err_style)
else:
self.usernametext.setStyleSheet('')
if 'database' in text:
self.databasetext.setStyleSheet(err_style)
else:
self.databasetext.setStyleSheet('')
def _parse_host_port(self):
hostport = self.servertext.text().split(':')
self.host = hostport[0]
self.port = hostport[1] if len(hostport) == 2 else None
def connect(self):
self._parse_host_port()
self.database, _, self.schema = self.databasetext.text().partition('/')
self.username = self.usernametext.text() or None
self.password = self.passwordtext.text() or None
try:
if self.backendcombo.currentIndex() < 0:
return
backend = self.backends[self.backendcombo.currentIndex()]
self.backend = backend(dict(
host=self.host,
port=self.port,
database=self.database,
user=self.username,
password=self.password
))
self.Error.connection.clear()
self._save_credentials()
self.database_desc = OrderedDict((
("Host", self.host), ("Port", self.port),
("Database", self.database), ("User name", self.username)
))
self.refresh_tables()
except BackendError as err:
error = str(err).split('\n')[0]
self.Error.connection(error)
self.database_desc = self.data_desc_table = None
self.tablecombo.clear()
def refresh_tables(self):
self.tables.clear()
self.Error.missing_extension.clear()
if self.backend is None:
self.data_desc_table = None
return
self.tables.append("Select a table")
self.tables.append("Custom SQL")
self.tables.extend(self.backend.list_tables(self.schema))
# Called on tablecombo selection change:
def select_table(self):
curIdx = self.tablecombo.currentIndex()
if self.tablecombo.itemText(curIdx) != "Custom SQL":
self.custom_sql.setVisible(False)
return self.open_table()
else:
self.custom_sql.setVisible(True)
self.data_desc_table = None
self.database_desc["Table"] = "(None)"
self.table = None
if len(str(self.sql)) > 14:
return self.open_table()
#self.Error.missing_extension(
# 's' if len(missing) > 1 else '',
# ', '.join(missing),
# shown=missing)
def open_table(self):
table = self.get_table()
self.data_desc_table = table
self.Outputs.data.send(table)
def get_table(self):
curIdx = self.tablecombo.currentIndex()
if curIdx <= 0:
if self.database_desc:
self.database_desc["Table"] = "(None)"
self.data_desc_table = None
return
if self.tablecombo.itemText(curIdx) != "Custom SQL":
self.table = self.tables[self.tablecombo.currentIndex()]
self.database_desc["Table"] = self.table
if "Query" in self.database_desc:
del self.database_desc["Query"]
what = self.table
else:
what = self.sql = self.sqltext.toPlainText()
self.table = "Custom SQL"
if self.materialize:
import psycopg2
if not self.materialize_table_name:
self.Error.connection(
"Specify a table name to materialize the query")
return
try:
with self.backend.execute_sql_query("DROP TABLE IF EXISTS " +
self.materialize_table_name):
pass
with self.backend.execute_sql_query("CREATE TABLE " +
self.materialize_table_name +
" AS " + self.sql):
pass
with self.backend.execute_sql_query("ANALYZE " + self.materialize_table_name):
pass
except (psycopg2.ProgrammingError, BackendError) as ex:
self.Error.connection(str(ex))
return
try:
table = SqlTable(dict(host=self.host,
port=self.port,
database=self.database,
user=self.username,
password=self.password),
what,
backend=type(self.backend),
inspect_values=False)
except BackendError as ex:
self.Error.connection(str(ex))
return
self.Error.connection.clear()
sample = False
if table.approx_len() > LARGE_TABLE and self.guess_values:
confirm = QMessageBox(self)
confirm.setIcon(QMessageBox.Warning)
confirm.setText("Attribute discovery might take "
"a long time on large tables.\n"
"Do you want to auto discover attributes?")
confirm.addButton("Yes", QMessageBox.YesRole)
no_button = confirm.addButton("No", QMessageBox.NoRole)
sample_button = confirm.addButton("Yes, on a sample",
QMessageBox.YesRole)
confirm.exec()
if confirm.clickedButton() == no_button:
self.guess_values = False
elif confirm.clickedButton() == sample_button:
sample = True
self.Information.clear()
if self.guess_values:
QApplication.setOverrideCursor(QCursor(Qt.WaitCursor))
if sample:
s = table.sample_time(1)
domain = s.get_domain(inspect_values=True)
self.Information.data_sampled()
else:
domain = table.get_domain(inspect_values=True)
QApplication.restoreOverrideCursor()
table.domain = domain
if self.download:
if table.approx_len() > MAX_DL_LIMIT:
QMessageBox.warning(
self, 'Warning', "Data is too big to download.\n"
"Consider using the Data Sampler widget to download "
"a sample instead.")
self.download = False
elif table.approx_len() > AUTO_DL_LIMIT:
confirm = QMessageBox.question(
self, 'Question', "Data appears to be big. Do you really "
"want to download it to local memory?",
QMessageBox.Yes | QMessageBox.No, QMessageBox.No)
if confirm == QMessageBox.No:
self.download = False
if self.download:
table.download_data(MAX_DL_LIMIT)
table = Table(table)
return table
def send_report(self):
if not self.database_desc:
self.report_paragraph("No database connection.")
return
self.report_items("Database", self.database_desc)
if self.data_desc_table:
self.report_items("Data",
report.describe_data(self.data_desc_table))
@classmethod
def migrate_settings(cls, settings, version):
if version < 2:
# Until Orange version 3.4.4 username and password had been stored
# in Settings.
cm = cls._credential_manager(settings["host"], settings["port"])
cm.username = settings["username"]
cm.password = settings["password"]
class FileWidget(QWidget):
on_open = pyqtSignal(str)
# TODO consider removing directory_aliases since it is not used any more
def __init__(self, dialog_title='', dialog_format='',
start_dir=os.path.expanduser('~/'),
icon_size=(12, 20), minimal_width=200,
browse_label='Browse', on_open=None,
reload_button=True, reload_label='Reload',
recent_files=None, directory_aliases=None,
allow_empty=True, empty_file_label='(none)'):
""" Creates a widget with a button for file loading and
an optional combo box for recent files and reload buttons.
Args:
dialog_title (str): The title of the dialog.
dialog_format (str): Formats for the dialog.
start_dir (str): A directory to start from.
icon_size (int, int): The size of buttons' icons.
on_open (callable): A callback function that accepts filepath as the only argument.
reload_button (bool): Whether to show reload button.
reload_label (str): The text displayed on the reload button.
recent_files (List[str]): List of recent files.
directory_aliases (dict): An {alias: dir} dictionary for fast directories' access.
allow_empty (bool): Whether empty path is allowed.
"""
super().__init__()
self.dialog_title = dialog_title
self.dialog_format = dialog_format
self.start_dir = start_dir
# Recent files should also contain `empty_file_label` so
# when (none) is selected this is stored in settings.
self.recent_files = recent_files if recent_files is not None else []
self.directory_aliases = directory_aliases or {}
self.allow_empty = allow_empty
self.empty_file_label = empty_file_label
if self.empty_file_label not in self.recent_files \
and (self.allow_empty or not self.recent_files):
self.recent_files.append(self.empty_file_label)
self.check_existence()
self.on_open.connect(on_open)
layout = QHBoxLayout(self)
layout.setContentsMargins(0, 0, 0, 0)
if recent_files is not None:
self.file_combo = QComboBox()
self.file_combo.setMinimumWidth(minimal_width)
self.file_combo.activated[int].connect(self.select)
self.update_combo()
layout.addWidget(self.file_combo)
self.browse_button = QPushButton(browse_label)
self.browse_button.setFocusPolicy(Qt.NoFocus)
self.browse_button.clicked.connect(self.browse)
self.browse_button.setIcon(self.style()
.standardIcon(QStyle.SP_DirOpenIcon))
self.browse_button.setIconSize(QSize(*icon_size))
self.browse_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
layout.addWidget(self.browse_button)
if reload_button:
self.reload_button = QPushButton(reload_label)
self.reload_button.setFocusPolicy(Qt.NoFocus)
self.reload_button.clicked.connect(self.reload)
self.reload_button.setIcon(self.style()
.standardIcon(QStyle.SP_BrowserReload))
self.reload_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
self.reload_button.setIconSize(QSize(*icon_size))
layout.addWidget(self.reload_button)
def browse(self, start_dir=None):
start_dir = start_dir or self.start_dir
path, _ = QFileDialog().getOpenFileName(self, self.dialog_title,
start_dir, self.dialog_format)
if path and self.recent_files is not None:
if path in self.recent_files:
self.recent_files.remove(path)
self.recent_files.insert(0, path)
self.update_combo()
if path:
self.open_file(path)
def select(self, n):
name = self.file_combo.currentText()
if name == self.empty_file_label:
del self.recent_files[n]
self.recent_files.insert(0, self.empty_file_label)
self.update_combo()
self.open_file(self.empty_file_label)
elif name in self.directory_aliases:
self.browse(self.directory_aliases[name])
elif n < len(self.recent_files):
name = self.recent_files[n]
del self.recent_files[n]
self.recent_files.insert(0, name)
self.update_combo()
self.open_file(self.recent_files[0])
def update_combo(self):
""" Sync combo values to the changes in self.recent_files. """
if self.recent_files is not None:
self.file_combo.clear()
for i, file in enumerate(self.recent_files):
# remove (none) when we have some files and allow_empty=False
if file == self.empty_file_label and \
not self.allow_empty and len(self.recent_files) > 1:
del self.recent_files[i]
else:
self.file_combo.addItem(os.path.split(file)[1])
for alias in self.directory_aliases.keys():
self.file_combo.addItem(alias)
def reload(self):
if self.recent_files:
self.select(0)
def check_existence(self):
if self.recent_files:
to_remove = []
for file in self.recent_files:
doc_path = os.path.join(get_sample_corpora_dir(), file)
exists = any(os.path.exists(f) for f in [file, doc_path])
if file != self.empty_file_label and not exists:
to_remove.append(file)
for file in to_remove:
self.recent_files.remove(file)
def open_file(self, path):
self.on_open.emit(path if path != self.empty_file_label else '')
def get_selected_filename(self):
if self.recent_files:
return self.recent_files[0]
else:
return self.empty_file_label
class OWPubmed(OWWidget):
name = 'Pubmed'
description = 'Fetch data from Pubmed.'
icon = 'icons/Pubmed.svg'
priority = 20
outputs = [(Output.CORPUS, Corpus)]
want_main_area = False
resizing_enabled = False
QT_DATE_FORMAT = 'yyyy-MM-dd'
PY_DATE_FORMAT = '%Y-%m-%d'
MIN_DATE = date(1800, 1, 1)
# Settings.
recent_emails = Setting([])
author = Setting('')
pub_date_from = Setting('')
pub_date_to = Setting('')
recent_keywords = Setting([])
last_advanced_query = Setting('')
num_records = Setting(1000)
# Text includes checkboxes.
includes_authors = Setting(True)
includes_title = Setting(True)
includes_mesh = Setting(True)
includes_abstract = Setting(True)
includes_url = Setting(True)
class Warning(OWWidget.Warning):
no_query = Msg('Please specify the keywords for this query.')
class Error(OWWidget.Error):
api_error = Msg('API error: {}.')
def __init__(self):
super().__init__()
self.output_corpus = None
self.pubmed_api = None
self.progress = None
self.email_is_valid = False
self.record_count = 0
self.download_running = False
# To hold all the controls. Makes access easier.
self.pubmed_controls = []
h_box = gui.hBox(self.controlArea)
label = gui.label(h_box, self, 'Email:')
label.setMaximumSize(label.sizeHint())
# Drop-down for recent emails.
self.email_combo = QComboBox(h_box)
self.email_combo.setMinimumWidth(150)
self.email_combo.setEditable(True)
self.email_combo.lineEdit().textChanged.connect(self.sync_email)
h_box.layout().addWidget(self.email_combo)
self.email_combo.activated[int].connect(self.select_email)
# RECORD SEARCH
self.search_tabs = gui.tabWidget(self.controlArea)
# --- Regular search ---
regular_search_box = gui.widgetBox(self.controlArea, addSpace=True)
# Author
self.author_input = gui.lineEdit(regular_search_box, self, 'author',
'Author:', orientation=Qt.Horizontal)
self.pubmed_controls.append(self.author_input)
h_box = gui.hBox(regular_search_box)
year_box = gui.widgetBox(h_box, orientation=Qt.Horizontal)
min_date = QDate.fromString(
self.MIN_DATE.strftime(self.PY_DATE_FORMAT),
self.QT_DATE_FORMAT
)
if not self.pub_date_from:
self.pub_date_from = self.MIN_DATE.strftime(self.PY_DATE_FORMAT)
if not self.pub_date_to:
self.pub_date_to = date.today().strftime(self.PY_DATE_FORMAT)
self.date_from = QDateEdit(
QDate.fromString(self.pub_date_from, self.QT_DATE_FORMAT),
displayFormat=self.QT_DATE_FORMAT,
minimumDate=min_date,
calendarPopup=True
)
self.date_to = QDateEdit(
QDate.fromString(self.pub_date_to, self.QT_DATE_FORMAT),
displayFormat=self.QT_DATE_FORMAT,
minimumDate=min_date,
calendarPopup=True
)
self.date_from.dateChanged.connect(
lambda date: setattr(self, 'pub_date_from',
date.toString(self.QT_DATE_FORMAT)))
self.date_to.dateChanged.connect(
lambda date: setattr(self, 'pub_date_to',
date.toString(self.QT_DATE_FORMAT)))
self.pubmed_controls.append(self.date_from)
self.pubmed_controls.append(self.date_to)
gui.label(year_box, self, 'From:')
year_box.layout().addWidget(self.date_from)
gui.label(year_box, self, 'to:')
year_box.layout().addWidget(self.date_to)
# Keywords.
h_box = gui.hBox(regular_search_box)
label = gui.label(h_box, self, 'Query:')
label.setMaximumSize(label.sizeHint())
self.keyword_combo = QComboBox(h_box)
self.keyword_combo.setMinimumWidth(150)
self.keyword_combo.setEditable(True)
h_box.layout().addWidget(self.keyword_combo)
self.keyword_combo.activated[int].connect(self.select_keywords)
self.pubmed_controls.append(self.keyword_combo)
tab_height = regular_search_box.sizeHint()
regular_search_box.setMaximumSize(tab_height)
# --- Advanced search ---
advanced_search_box = gui.widgetBox(self.controlArea, addSpace=True)
# Advanced search query.
h_box = gui.hBox(advanced_search_box)
self.advanced_query_input = QTextEdit(h_box)
h_box.layout().addWidget(self.advanced_query_input)
self.advanced_query_input.setMaximumSize(tab_height)
self.pubmed_controls.append(self.advanced_query_input)
gui.createTabPage(self.search_tabs, 'Regular search',
regular_search_box)
gui.createTabPage(self.search_tabs, 'Advanced search',
advanced_search_box)
# Search info label.
self.search_info_label = gui.label(
self.controlArea, self,
'Number of records found: /')
# Search for records button.
self.run_search_button = gui.button(
self.controlArea,
self,
'Find records',
callback=self.run_search,
tooltip='Performs a search for articles that fit the '
'specified parameters.')
self.pubmed_controls.append(self.run_search_button)
h_line = QFrame()
h_line.setFrameShape(QFrame.HLine)
h_line.setFrameShadow(QFrame.Sunken)
self.controlArea.layout().addWidget(h_line)
# RECORD RETRIEVAL
# Text includes box.
text_includes_box = gui.widgetBox(self.controlArea,
'Text includes', addSpace=True)
self.authors_checkbox = gui.checkBox(text_includes_box, self,
'includes_authors', 'Authors')
self.title_checkbox = gui.checkBox(text_includes_box, self,
'includes_title', 'Article title')
self.mesh_checkbox = gui.checkBox(text_includes_box, self,
'includes_mesh', 'Mesh headings')
self.abstract_checkbox = gui.checkBox(text_includes_box, self,
'includes_abstract', 'Abstract')
self.url_checkbox = gui.checkBox(text_includes_box, self,
'includes_url', 'URL')
self.pubmed_controls.append(self.authors_checkbox)
self.pubmed_controls.append(self.title_checkbox)
self.pubmed_controls.append(self.mesh_checkbox)
self.pubmed_controls.append(self.abstract_checkbox)
self.pubmed_controls.append(self.url_checkbox)
# Num. records.
h_box = gui.hBox(self.controlArea)
label = gui.label(h_box, self, 'Retrieve')
label.setMaximumSize(label.sizeHint())
self.num_records_input = gui.spin(h_box, self, 'num_records',
minv=1, maxv=100000)
self.max_records_label = gui.label(h_box, self, 'records from /.')
self.max_records_label.setMaximumSize(self.max_records_label
.sizeHint())
self.pubmed_controls.append(self.num_records_input)
# Download articles.
# Search for records button.
self.retrieve_records_button = gui.button(
self.controlArea,
self,
'Retrieve records',
callback=self.retrieve_records,
tooltip='Retrieves the specified documents.')
self.pubmed_controls.append(self.retrieve_records_button)
# Num. retrieved records info label.
self.retrieval_info_label = gui.label(
self.controlArea,
self,
'Number of records retrieved: /')
# Load the most recent emails.
self.set_email_list()
# Load the most recent queries.
self.set_keyword_list()
# Check the email and enable controls accordingly.
if self.recent_emails:
email = self.recent_emails[0]
self.email_is_valid = validate_email(email)
self.enable_controls()
def sync_email(self):
email = self.email_combo.currentText()
self.email_is_valid = validate_email(email)
self.enable_controls()
def enable_controls(self):
# Enable/disable controls accordingly.
for control in self.pubmed_controls:
control.setEnabled(self.email_is_valid)
if self.pubmed_api is None or self.pubmed_api.search_record_count == 0:
self.retrieve_records_button.setEnabled(False)
if not self.email_is_valid:
self.email_combo.setFocus()
def run_search(self):
self.Error.clear()
self.Warning.clear()
self.run_search_button.setEnabled(False)
self.retrieve_records_button.setEnabled(False)
# Add the email to history.
email = self.email_combo.currentText()
if email not in self.recent_emails:
self.recent_emails.insert(0, email)
# Check if the PubMed object is present.
if self.pubmed_api is None:
self.pubmed_api = Pubmed(
email=email,
progress_callback=self.api_progress_callback,
error_callback=self.api_error_callback,
)
if self.search_tabs.currentIndex() == 0:
# Get query parameters.
terms = self.keyword_combo.currentText().split()
authors = self.author_input.text().split()
error = self.pubmed_api._search_for_records(
terms, authors, self.pub_date_from, self.pub_date_to
)
if error is not None:
self.Error.api_error(str(error))
return
if self.keyword_combo.currentText() not in self.recent_keywords:
self.recent_keywords.insert(
0,
self.keyword_combo.currentText()
)
else:
query = self.advanced_query_input.toPlainText()
if not query:
self.Warning.no_query()
self.run_search_button.setEnabled(True)
self.retrieve_records_button.setEnabled(True)
return
error = self.pubmed_api._search_for_records(advanced_query=query)
if error is not None:
self.Error.api_error(str(error))
return
self.last_advanced_query = query
self.enable_controls()
self.update_search_info()
def retrieve_records(self):
self.Warning.clear()
self.Error.clear()
if self.pubmed_api is None:
return
if self.download_running:
self.download_running = False
self.run_search_button.setEnabled(True)
self.retrieve_records_button.setText('Retrieve records')
self.pubmed_api.stop_retrieving()
return
self.download_running = True
self.run_search_button.setEnabled(False)
self.output_corpus = None # Clear the old records.
# Change the button label.
self.retrieve_records_button.setText('Stop retrieving')
# Text fields.
text_includes_params = [
self.includes_authors,
self.includes_title,
self.includes_mesh,
self.includes_abstract,
self.includes_url,
True, # Publication date field; included always.
]
required_text_fields = [
field
for field_name, field
in zip(text_includes_params, PUBMED_TEXT_FIELDS)
if field_name
]
batch_size = min(Pubmed.MAX_BATCH_SIZE, self.num_records) + 1
with self.progressBar(self.num_records/batch_size) as progress:
self.progress = progress
self.output_corpus = self.pubmed_api._retrieve_records(
self.num_records,
required_text_fields
)
self.retrieve_records_button.setText('Retrieve records')
self.download_running = False
self.send(Output.CORPUS, self.output_corpus)
self.update_retrieval_info()
self.run_search_button.setEnabled(True)
def api_progress_callback(self, start_at=None):
if start_at is not None:
self.progress.count = start_at
else:
self.progress.advance()
def api_error_callback(self, error):
self.Error.api_error(str(error))
if self.progress is not None:
self.progress.finish()
def update_search_info(self):
max_records_count = min(
self.pubmed_api.MAX_RECORDS,
self.pubmed_api.search_record_count
)
self.search_info_label.setText(
'Number of retrievable records for '
'this search query: {} '.format(max_records_count)
)
self.max_records_label.setText(
'records from {}.'.format(max_records_count)
)
self.max_records_label.setMaximumSize(self.max_records_label
.sizeHint())
self.num_records_input.setMaximum(max_records_count)
self.retrieve_records_button.setFocus()
def update_retrieval_info(self):
document_count = 0
if self.output_corpus is not None:
document_count = len(self.output_corpus)
self.retrieval_info_label.setText(
'Number of records retrieved: {} '.format(document_count)
)
self.retrieval_info_label.setMaximumSize(
self.retrieval_info_label.sizeHint()
)
def select_email(self, n):
if n < len(self.recent_emails):
email = self.recent_emails[n]
del self.recent_emails[n]
self.recent_emails.insert(0, email)
if len(self.recent_emails) > 0:
self.set_email_list()
def set_email_list(self):
self.email_combo.clear()
for email in self.recent_emails:
self.email_combo.addItem(email)
def select_keywords(self, n):
if n < len(self.recent_keywords):
keywords = self.recent_keywords[n]
del self.recent_keywords[n]
self.recent_keywords.insert(0, keywords)
if len(self.recent_keywords) > 0:
self.set_keyword_list()
def set_keyword_list(self):
self.keyword_combo.clear()
if not self.recent_keywords:
# Sample queries.
self.recent_keywords.append('orchid')
self.recent_keywords.append('hypertension')
self.recent_keywords.append('blood pressure')
self.recent_keywords.append('radiology')
for keywords in self.recent_keywords:
self.keyword_combo.addItem(keywords)
def open_calendar(self, widget):
cal_dlg = CalendarDialog(self, 'Date picker')
if cal_dlg.exec_():
widget.setText(cal_dlg.picked_date)
class OWPubmed(OWWidget):
name = 'Pubmed'
description = 'Fetch data from Pubmed.'
icon = 'icons/Pubmed.svg'
priority = 140
outputs = [(Output.CORPUS, Corpus)]
want_main_area = False
resizing_enabled = False
QT_DATE_FORMAT = 'yyyy-MM-dd'
PY_DATE_FORMAT = '%Y-%m-%d'
MIN_DATE = date(1800, 1, 1)
# Settings.
recent_emails = Setting([])
author = Setting('')
pub_date_from = Setting('')
pub_date_to = Setting('')
recent_keywords = Setting([])
last_advanced_query = Setting('')
num_records = Setting(1000)
# Text includes checkboxes.
includes_authors = Setting(True)
includes_title = Setting(True)
includes_mesh = Setting(True)
includes_abstract = Setting(True)
includes_url = Setting(True)
class Warning(OWWidget.Warning):
no_query = Msg('Please specify the keywords for this query.')
class Error(OWWidget.Error):
api_error = Msg('API error: {}.')
def __init__(self):
super().__init__()
self.output_corpus = None
self.pubmed_api = None
self.progress = None
self.email_is_valid = False
self.record_count = 0
self.download_running = False
# To hold all the controls. Makes access easier.
self.pubmed_controls = []
h_box = gui.hBox(self.controlArea)
label = gui.label(h_box, self, 'Email:')
label.setMaximumSize(label.sizeHint())
# Drop-down for recent emails.
self.email_combo = QComboBox(h_box)
self.email_combo.setMinimumWidth(150)
self.email_combo.setEditable(True)
self.email_combo.lineEdit().textChanged.connect(self.sync_email)
h_box.layout().addWidget(self.email_combo)
self.email_combo.activated[int].connect(self.select_email)
# RECORD SEARCH
self.search_tabs = gui.tabWidget(self.controlArea)
# --- Regular search ---
regular_search_box = gui.widgetBox(self.controlArea, addSpace=True)
# Author
self.author_input = gui.lineEdit(regular_search_box,
self,
'author',
'Author:',
orientation=Qt.Horizontal)
self.pubmed_controls.append(self.author_input)
h_box = gui.hBox(regular_search_box)
year_box = gui.widgetBox(h_box, orientation=Qt.Horizontal)
min_date = QDate.fromString(
self.MIN_DATE.strftime(self.PY_DATE_FORMAT), self.QT_DATE_FORMAT)
if not self.pub_date_from:
self.pub_date_from = self.MIN_DATE.strftime(self.PY_DATE_FORMAT)
if not self.pub_date_to:
self.pub_date_to = date.today().strftime(self.PY_DATE_FORMAT)
self.date_from = QDateEdit(QDate.fromString(self.pub_date_from,
self.QT_DATE_FORMAT),
displayFormat=self.QT_DATE_FORMAT,
minimumDate=min_date,
calendarPopup=True)
self.date_to = QDateEdit(QDate.fromString(self.pub_date_to,
self.QT_DATE_FORMAT),
displayFormat=self.QT_DATE_FORMAT,
minimumDate=min_date,
calendarPopup=True)
self.date_from.dateChanged.connect(lambda date: setattr(
self, 'pub_date_from', date.toString(self.QT_DATE_FORMAT)))
self.date_to.dateChanged.connect(lambda date: setattr(
self, 'pub_date_to', date.toString(self.QT_DATE_FORMAT)))
self.pubmed_controls.append(self.date_from)
self.pubmed_controls.append(self.date_to)
gui.label(year_box, self, 'From:')
year_box.layout().addWidget(self.date_from)
gui.label(year_box, self, 'to:')
year_box.layout().addWidget(self.date_to)
# Keywords.
h_box = gui.hBox(regular_search_box)
label = gui.label(h_box, self, 'Query:')
label.setMaximumSize(label.sizeHint())
self.keyword_combo = QComboBox(h_box)
self.keyword_combo.setMinimumWidth(150)
self.keyword_combo.setEditable(True)
h_box.layout().addWidget(self.keyword_combo)
self.keyword_combo.activated[int].connect(self.select_keywords)
self.pubmed_controls.append(self.keyword_combo)
tab_height = regular_search_box.sizeHint()
regular_search_box.setMaximumSize(tab_height)
# --- Advanced search ---
advanced_search_box = gui.widgetBox(self.controlArea, addSpace=True)
# Advanced search query.
h_box = gui.hBox(advanced_search_box)
self.advanced_query_input = QTextEdit(h_box)
h_box.layout().addWidget(self.advanced_query_input)
self.advanced_query_input.setMaximumSize(tab_height)
self.pubmed_controls.append(self.advanced_query_input)
gui.createTabPage(self.search_tabs, 'Regular search',
regular_search_box)
gui.createTabPage(self.search_tabs, 'Advanced search',
advanced_search_box)
# Search info label.
self.search_info_label = gui.label(self.controlArea, self,
'Number of records found: /')
# Search for records button.
self.run_search_button = gui.button(
self.controlArea,
self,
'Find records',
callback=self.run_search,
tooltip='Performs a search for articles that fit the '
'specified parameters.')
self.pubmed_controls.append(self.run_search_button)
h_line = QFrame()
h_line.setFrameShape(QFrame.HLine)
h_line.setFrameShadow(QFrame.Sunken)
self.controlArea.layout().addWidget(h_line)
# RECORD RETRIEVAL
# Text includes box.
text_includes_box = gui.widgetBox(self.controlArea,
'Text includes',
addSpace=True)
self.authors_checkbox = gui.checkBox(text_includes_box, self,
'includes_authors', 'Authors')
self.title_checkbox = gui.checkBox(text_includes_box, self,
'includes_title', 'Article title')
self.mesh_checkbox = gui.checkBox(text_includes_box, self,
'includes_mesh', 'Mesh headings')
self.abstract_checkbox = gui.checkBox(text_includes_box, self,
'includes_abstract', 'Abstract')
self.url_checkbox = gui.checkBox(text_includes_box, self,
'includes_url', 'URL')
self.pubmed_controls.append(self.authors_checkbox)
self.pubmed_controls.append(self.title_checkbox)
self.pubmed_controls.append(self.mesh_checkbox)
self.pubmed_controls.append(self.abstract_checkbox)
self.pubmed_controls.append(self.url_checkbox)
# Num. records.
h_box = gui.hBox(self.controlArea)
label = gui.label(h_box, self, 'Retrieve')
label.setMaximumSize(label.sizeHint())
self.num_records_input = gui.spin(h_box,
self,
'num_records',
minv=1,
maxv=10000)
self.max_records_label = gui.label(h_box, self, 'records from /.')
self.max_records_label.setMaximumSize(
self.max_records_label.sizeHint())
self.pubmed_controls.append(self.num_records_input)
# Download articles.
# Search for records button.
self.retrieve_records_button = gui.button(
self.controlArea,
self,
'Retrieve records',
callback=self.retrieve_records,
tooltip='Retrieves the specified documents.')
self.pubmed_controls.append(self.retrieve_records_button)
# Num. retrieved records info label.
self.retrieval_info_label = gui.label(
self.controlArea, self, 'Number of records retrieved: /')
# Load the most recent emails.
self.set_email_list()
# Load the most recent queries.
self.set_keyword_list()
# Check the email and enable controls accordingly.
if self.recent_emails:
email = self.recent_emails[0]
self.email_is_valid = validate_email(email)
self.enable_controls()
def sync_email(self):
email = self.email_combo.currentText()
self.email_is_valid = validate_email(email)
self.enable_controls()
def enable_controls(self):
# Enable/disable controls accordingly.
for control in self.pubmed_controls:
control.setEnabled(self.email_is_valid)
if self.pubmed_api is None or self.pubmed_api.search_record_count == 0:
self.retrieve_records_button.setEnabled(False)
if not self.email_is_valid:
self.email_combo.setFocus()
def run_search(self):
self.Error.clear()
self.Warning.clear()
self.run_search_button.setEnabled(False)
self.retrieve_records_button.setEnabled(False)
# Add the email to history.
email = self.email_combo.currentText()
if email not in self.recent_emails:
self.recent_emails.insert(0, email)
# Check if the PubMed object is present.
if self.pubmed_api is None:
self.pubmed_api = Pubmed(
email=email,
progress_callback=self.api_progress_callback,
error_callback=self.api_error_callback,
)
if self.search_tabs.currentIndex() == 0:
# Get query parameters.
terms = self.keyword_combo.currentText().split()
authors = self.author_input.text().split()
error = self.pubmed_api._search_for_records(
terms, authors, self.pub_date_from, self.pub_date_to)
if error is not None:
self.Error.api_error(str(error))
return
if self.keyword_combo.currentText() not in self.recent_keywords:
self.recent_keywords.insert(0,
self.keyword_combo.currentText())
else:
query = self.advanced_query_input.toPlainText()
if not query:
self.Warning.no_query()
self.run_search_button.setEnabled(True)
self.retrieve_records_button.setEnabled(True)
return
error = self.pubmed_api._search_for_records(advanced_query=query)
if error is not None:
self.Error.api_error(str(error))
return
self.last_advanced_query = query
self.enable_controls()
self.update_search_info()
def retrieve_records(self):
self.Warning.clear()
self.Error.clear()
if self.pubmed_api is None:
return
if self.download_running:
self.download_running = False
self.run_search_button.setEnabled(True)
self.retrieve_records_button.setText('Retrieve records')
self.pubmed_api.stop_retrieving()
return
self.download_running = True
self.run_search_button.setEnabled(False)
self.output_corpus = None # Clear the old records.
# Change the button label.
self.retrieve_records_button.setText('Stop retrieving')
# Text fields.
text_includes_params = [
self.includes_authors,
self.includes_title,
self.includes_mesh,
self.includes_abstract,
self.includes_url,
True, # Publication date field; included always.
]
required_text_fields = [
field for field_name, field in zip(
text_includes_params, PUBMED_TEXT_FIELDS) if field_name
]
batch_size = min(Pubmed.MAX_BATCH_SIZE, self.num_records) + 1
with self.progressBar(self.num_records / batch_size) as progress:
self.progress = progress
self.output_corpus = self.pubmed_api._retrieve_records(
self.num_records, required_text_fields)
self.retrieve_records_button.setText('Retrieve records')
self.download_running = False
self.send(Output.CORPUS, self.output_corpus)
self.update_retrieval_info()
self.run_search_button.setEnabled(True)
def api_progress_callback(self, start_at=None):
if start_at is not None:
self.progress.count = start_at
else:
self.progress.advance()
def api_error_callback(self, error):
self.Error.api_error(str(error))
if self.progress is not None:
self.progress.finish()
def update_search_info(self):
max_records_count = min(self.pubmed_api.MAX_RECORDS,
self.pubmed_api.search_record_count)
self.search_info_label.setText(
'Number of retrievable records for '
'this search query: {} '.format(max_records_count))
self.max_records_label.setText(
'records from {}.'.format(max_records_count))
self.max_records_label.setMaximumSize(
self.max_records_label.sizeHint())
self.num_records_input.setMaximum(max_records_count)
self.retrieve_records_button.setFocus()
def update_retrieval_info(self):
document_count = 0
if self.output_corpus is not None:
document_count = len(self.output_corpus)
self.retrieval_info_label.setText(
'Number of records retrieved: {} '.format(document_count))
self.retrieval_info_label.setMaximumSize(
self.retrieval_info_label.sizeHint())
def select_email(self, n):
if n < len(self.recent_emails):
email = self.recent_emails[n]
del self.recent_emails[n]
self.recent_emails.insert(0, email)
if len(self.recent_emails) > 0:
self.set_email_list()
def set_email_list(self):
self.email_combo.clear()
for email in self.recent_emails:
self.email_combo.addItem(email)
def select_keywords(self, n):
if n < len(self.recent_keywords):
keywords = self.recent_keywords[n]
del self.recent_keywords[n]
self.recent_keywords.insert(0, keywords)
if len(self.recent_keywords) > 0:
self.set_keyword_list()
def set_keyword_list(self):
self.keyword_combo.clear()
if not self.recent_keywords:
# Sample queries.
self.recent_keywords.append('orchid')
self.recent_keywords.append('hypertension')
self.recent_keywords.append('blood pressure')
self.recent_keywords.append('radiology')
for keywords in self.recent_keywords:
self.keyword_combo.addItem(keywords)
def open_calendar(self, widget):
cal_dlg = CalendarDialog(self, 'Date picker')
if cal_dlg.exec_():
widget.setText(cal_dlg.picked_date)
def send_report(self):
if not self.pubmed_api:
return
max_records_count = min(self.pubmed_api.MAX_RECORDS,
self.pubmed_api.search_record_count)
if self.search_tabs.currentIndex() == 0:
terms = self.keyword_combo.currentText()
authors = self.author_input.text()
self.report_items(
(('Query', terms if terms else None),
('Authors', authors if authors else None),
('Date', 'from {} to {}'.format(self.pub_date_from,
self.pub_date_to)),
('Number of records retrieved', '{}/{}'.format(
len(self.output_corpus) if self.output_corpus else 0,
max_records_count))))
else:
query = self.advanced_query_input.toPlainText()
self.report_items(
(('Query', query if query else None),
('Number of records retrieved', '{}/{}'.format(
len(self.output_corpus) if self.output_corpus else 0,
max_records_count))))
class FileWidget(QWidget):
on_open = pyqtSignal(str)
# TODO consider removing directory_aliases since it is not used any more
def __init__(self,
dialog_title='',
dialog_format='',
start_dir=os.path.expanduser('~/'),
icon_size=(12, 20),
minimal_width=200,
browse_label='Browse',
on_open=None,
reload_button=True,
reload_label='Reload',
recent_files=None,
directory_aliases=None,
allow_empty=True,
empty_file_label='(none)'):
""" Creates a widget with a button for file loading and
an optional combo box for recent files and reload buttons.
Args:
dialog_title (str): The title of the dialog.
dialog_format (str): Formats for the dialog.
start_dir (str): A directory to start from.
icon_size (int, int): The size of buttons' icons.
on_open (callable): A callback function that accepts filepath as the only argument.
reload_button (bool): Whether to show reload button.
reload_label (str): The text displayed on the reload button.
recent_files (List[str]): List of recent files.
directory_aliases (dict): An {alias: dir} dictionary for fast directories' access.
allow_empty (bool): Whether empty path is allowed.
"""
super().__init__()
self.dialog_title = dialog_title
self.dialog_format = dialog_format
self.start_dir = start_dir
# Recent files should also contain `empty_file_label` so
# when (none) is selected this is stored in settings.
self.recent_files = recent_files if recent_files is not None else []
self.directory_aliases = directory_aliases or {}
self.allow_empty = allow_empty
self.empty_file_label = empty_file_label
if self.empty_file_label not in self.recent_files \
and (self.allow_empty or not self.recent_files):
self.recent_files.append(self.empty_file_label)
self.check_existence()
self.on_open.connect(on_open)
layout = QHBoxLayout(self)
layout.setContentsMargins(0, 0, 0, 0)
if recent_files is not None:
self.file_combo = QComboBox()
self.file_combo.setMinimumWidth(minimal_width)
self.file_combo.activated[int].connect(self.select)
self.update_combo()
layout.addWidget(self.file_combo)
self.browse_button = QPushButton(browse_label)
self.browse_button.setFocusPolicy(Qt.NoFocus)
self.browse_button.clicked.connect(self.browse)
self.browse_button.setIcon(self.style().standardIcon(
QStyle.SP_DirOpenIcon))
self.browse_button.setIconSize(QSize(*icon_size))
self.browse_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
layout.addWidget(self.browse_button)
if reload_button:
self.reload_button = QPushButton(reload_label)
self.reload_button.setFocusPolicy(Qt.NoFocus)
self.reload_button.clicked.connect(self.reload)
self.reload_button.setIcon(self.style().standardIcon(
QStyle.SP_BrowserReload))
self.reload_button.setSizePolicy(QSizePolicy.Fixed,
QSizePolicy.Fixed)
self.reload_button.setIconSize(QSize(*icon_size))
layout.addWidget(self.reload_button)
def browse(self, start_dir=None):
start_dir = start_dir or self.start_dir
path, _ = QFileDialog().getOpenFileName(self, self.dialog_title,
start_dir, self.dialog_format)
if path and self.recent_files is not None:
if path in self.recent_files:
self.recent_files.remove(path)
self.recent_files.insert(0, path)
self.update_combo()
if path:
self.open_file(path)
def select(self, n):
name = self.file_combo.currentText()
if name == self.empty_file_label:
del self.recent_files[n]
self.recent_files.insert(0, self.empty_file_label)
self.update_combo()
self.open_file(self.empty_file_label)
elif name in self.directory_aliases:
self.browse(self.directory_aliases[name])
elif n < len(self.recent_files):
name = self.recent_files[n]
del self.recent_files[n]
self.recent_files.insert(0, name)
self.update_combo()
self.open_file(self.recent_files[0])
def update_combo(self):
""" Sync combo values to the changes in self.recent_files. """
if self.recent_files is not None:
self.file_combo.clear()
for i, file in enumerate(self.recent_files):
# remove (none) when we have some files and allow_empty=False
if file == self.empty_file_label and \
not self.allow_empty and len(self.recent_files) > 1:
del self.recent_files[i]
else:
self.file_combo.addItem(os.path.split(file)[1])
for alias in self.directory_aliases.keys():
self.file_combo.addItem(alias)
def reload(self):
if self.recent_files:
self.select(0)
def check_existence(self):
if self.recent_files:
to_remove = []
for file in self.recent_files:
doc_path = os.path.join(get_sample_corpora_dir(), file)
exists = any(os.path.exists(f) for f in [file, doc_path])
if file != self.empty_file_label and not exists:
to_remove.append(file)
for file in to_remove:
self.recent_files.remove(file)
def open_file(self, path):
self.on_open.emit(path if path != self.empty_file_label else '')
def get_selected_filename(self):
if self.recent_files:
return self.recent_files[0]
else:
return self.empty_file_label
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)
class FileWidget(QWidget):
on_open = pyqtSignal(str)
def __init__(self, dialog_title='', dialog_format='',
start_dir=os.path.expanduser('~/'),
icon_size=(12, 20), minimal_width=200,
browse_label='Browse', on_open=None,
reload_button=True, reload_label='Reload',
recent_files=None, directory_aliases=None,
allow_empty=True, empty_file_label='(none)'):
""" Creates a widget with a button for file loading and
an optional combo box for recent files and reload buttons.
Args:
dialog_title (str): The title of the dialog.
dialog_format (str): Formats for the dialog.
start_dir (str): A directory to start from.
icon_size (int, int): The size of buttons' icons.
on_open (callable): A callback function that accepts filepath as the only argument.
reload_button (bool): Whether to show reload button.
reload_label (str): The text displayed on the reload button.
recent_files (List[str]): List of recent files.
directory_aliases (dict): An {alias: dir} dictionary for fast directories' access.
allow_empty (bool): Whether empty path is allowed.
"""
super().__init__()
self.dialog_title = dialog_title
self.dialog_format = dialog_format
self.start_dir = start_dir
self.recent_files = recent_files
self.directory_aliases = directory_aliases or {}
self.check_existence()
self.on_open.connect(on_open)
self.allow_empty = allow_empty
self.empty_file_label = empty_file_label
layout = QHBoxLayout(self)
layout.setContentsMargins(0, 0, 0, 0)
if recent_files is not None:
self.file_combo = QComboBox()
self.file_combo.setMinimumWidth(minimal_width)
self.file_combo.activated[int].connect(self.select)
self.update_combo()
layout.addWidget(self.file_combo)
self.browse_button = QPushButton(browse_label)
self.browse_button.setFocusPolicy(Qt.NoFocus)
self.browse_button.clicked.connect(self.browse)
self.browse_button.setIcon(self.style()
.standardIcon(QStyle.SP_DirOpenIcon))
self.browse_button.setIconSize(QSize(*icon_size))
self.browse_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
layout.addWidget(self.browse_button)
if reload_button:
self.reload_button = QPushButton(reload_label)
self.reload_button.setFocusPolicy(Qt.NoFocus)
self.reload_button.clicked.connect(self.reload)
self.reload_button.setIcon(self.style()
.standardIcon(QStyle.SP_BrowserReload))
self.reload_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
self.reload_button.setIconSize(QSize(*icon_size))
layout.addWidget(self.reload_button)
def browse(self, start_dir=None):
start_dir = start_dir or self.start_dir
path, _ = QFileDialog().getOpenFileName(self, self.dialog_title,
start_dir, self.dialog_format)
if path and self.recent_files is not None:
if path in self.recent_files:
self.recent_files.remove(path)
self.recent_files.insert(0, path)
self.update_combo()
self.open_file(path)
def select(self, n):
name = self.file_combo.currentText()
if n < len(self.recent_files):
name = self.recent_files[n]
del self.recent_files[n]
self.recent_files.insert(0, name)
self.open_file(self.recent_files[0])
self.update_combo()
elif name == self.empty_file_label:
self.open_file(self.empty_file_label)
elif name in self.directory_aliases:
self.browse(self.directory_aliases[name])
def update_combo(self):
if self.recent_files is not None:
self.file_combo.clear()
for file in self.recent_files:
self.file_combo.addItem(os.path.split(file)[1])
if self.allow_empty or not self.recent_files:
self.file_combo.addItem(self.empty_file_label)
for alias in self.directory_aliases.keys():
self.file_combo.addItem(alias)
def reload(self):
if self.recent_files:
self.select(0)
def check_existence(self):
if self.recent_files:
to_remove = [
file for file in self.recent_files if not os.path.exists(file)
]
for file in to_remove:
self.recent_files.remove(file)
def open_file(self, path):
try:
self.on_open.emit(path if path != self.empty_file_label else '')
except (OSError, IOError):
self.loading_error_signal.emit('Could not open "{}".'
.format(path))
class FileWidget(QWidget):
on_open = pyqtSignal(str)
def __init__(self,
dialog_title='',
dialog_format='',
start_dir=os.path.expanduser('~/'),
icon_size=(12, 20),
minimal_width=200,
browse_label='Browse',
on_open=None,
reload_button=True,
reload_label='Reload',
recent_files=None,
directory_aliases=None,
allow_empty=True,
empty_file_label='(none)'):
""" Creates a widget with a button for file loading and
an optional combo box for recent files and reload buttons.
Args:
dialog_title (str): The title of the dialog.
dialog_format (str): Formats for the dialog.
start_dir (str): A directory to start from.
icon_size (int, int): The size of buttons' icons.
on_open (callable): A callback function that accepts filepath as the only argument.
reload_button (bool): Whether to show reload button.
reload_label (str): The text displayed on the reload button.
recent_files (List[str]): List of recent files.
directory_aliases (dict): An {alias: dir} dictionary for fast directories' access.
allow_empty (bool): Whether empty path is allowed.
"""
super().__init__()
self.dialog_title = dialog_title
self.dialog_format = dialog_format
self.start_dir = start_dir
self.recent_files = recent_files
self.directory_aliases = directory_aliases or {}
self.check_existence()
self.on_open.connect(on_open)
self.allow_empty = allow_empty
self.empty_file_label = empty_file_label
layout = QHBoxLayout(self)
layout.setContentsMargins(0, 0, 0, 0)
if recent_files is not None:
self.file_combo = QComboBox()
self.file_combo.setMinimumWidth(minimal_width)
self.file_combo.activated[int].connect(self.select)
self.update_combo()
layout.addWidget(self.file_combo)
self.browse_button = QPushButton(browse_label)
self.browse_button.setFocusPolicy(Qt.NoFocus)
self.browse_button.clicked.connect(self.browse)
self.browse_button.setIcon(self.style().standardIcon(
QStyle.SP_DirOpenIcon))
self.browse_button.setIconSize(QSize(*icon_size))
self.browse_button.setSizePolicy(QSizePolicy.Fixed, QSizePolicy.Fixed)
layout.addWidget(self.browse_button)
if reload_button:
self.reload_button = QPushButton(reload_label)
self.reload_button.setFocusPolicy(Qt.NoFocus)
self.reload_button.clicked.connect(self.reload)
self.reload_button.setIcon(self.style().standardIcon(
QStyle.SP_BrowserReload))
self.reload_button.setSizePolicy(QSizePolicy.Fixed,
QSizePolicy.Fixed)
self.reload_button.setIconSize(QSize(*icon_size))
layout.addWidget(self.reload_button)
def browse(self, start_dir=None):
start_dir = start_dir or self.start_dir
path, _ = QFileDialog().getOpenFileName(self, self.dialog_title,
start_dir, self.dialog_format)
if path and self.recent_files is not None:
if path in self.recent_files:
self.recent_files.remove(path)
self.recent_files.insert(0, path)
self.update_combo()
self.open_file(path)
def select(self, n):
name = self.file_combo.currentText()
if n < len(self.recent_files):
name = self.recent_files[n]
del self.recent_files[n]
self.recent_files.insert(0, name)
self.open_file(self.recent_files[0])
self.update_combo()
elif name == self.empty_file_label:
self.open_file(self.empty_file_label)
elif name in self.directory_aliases:
self.browse(self.directory_aliases[name])
def update_combo(self):
if self.recent_files is not None:
self.file_combo.clear()
for file in self.recent_files:
self.file_combo.addItem(os.path.split(file)[1])
if self.allow_empty or not self.recent_files:
self.file_combo.addItem(self.empty_file_label)
for alias in self.directory_aliases.keys():
self.file_combo.addItem(alias)
def reload(self):
if self.recent_files:
self.select(0)
def check_existence(self):
if self.recent_files:
to_remove = [
file for file in self.recent_files if not os.path.exists(file)
]
for file in to_remove:
self.recent_files.remove(file)
def open_file(self, path):
try:
self.on_open.emit(path if path != self.empty_file_label else '')
except (OSError, IOError):
self.loading_error_signal.emit('Could not open "{}".'.format(path))
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()
class OWFile(widget.OWWidget, RecentPathsWComboMixin):
name = "File"
id = "orange.widgets.data.file"
description = "Read data from an input file or network " \
"and send a data table to the output."
icon = "icons/File.svg"
priority = 10
category = "Data"
keywords = ["file", "load", "read", "open"]
class Outputs:
data = Output("Data",
Table,
doc="Attribute-valued dataset read from the input file.")
want_main_area = False
buttons_area_orientation = None
SEARCH_PATHS = [("sample-datasets", get_sample_datasets_dir())]
SIZE_LIMIT = 1e7
LOCAL_FILE, URL = range(2)
settingsHandler = PerfectDomainContextHandler(
match_values=PerfectDomainContextHandler.MATCH_VALUES_ALL)
# pylint seems to want declarations separated from definitions
recent_paths: List[RecentPath]
recent_urls: List[str]
variables: list
# Overload RecentPathsWidgetMixin.recent_paths to set defaults
recent_paths = Setting([
RecentPath("", "sample-datasets", "iris.tab"),
RecentPath("", "sample-datasets", "titanic.tab"),
RecentPath("", "sample-datasets", "housing.tab"),
RecentPath("", "sample-datasets", "heart_disease.tab"),
RecentPath("", "sample-datasets", "brown-selected.tab"),
RecentPath("", "sample-datasets", "zoo.tab"),
])
recent_urls = Setting([])
source = Setting(LOCAL_FILE)
sheet_names = Setting({})
url = Setting("")
variables = ContextSetting([])
domain_editor = SettingProvider(DomainEditor)
class Information(widget.OWWidget.Information):
no_file_selected = Msg("No file selected.")
class Warning(widget.OWWidget.Warning):
file_too_big = Msg("The file is too large to load automatically."
" Press Reload to load.")
load_warning = Msg("Read warning:\n{}")
performance_warning = Msg(
"Categorical variables with >100 values may decrease performance.")
renamed_vars = Msg("Some variables have been renamed "
"to avoid duplicates.\n{}")
multiple_targets = Msg("Most widgets do not support multiple targets")
class Error(widget.OWWidget.Error):
file_not_found = Msg("File not found.")
missing_reader = Msg("Missing reader.")
sheet_error = Msg("Error listing available sheets.")
unknown = Msg("Read error:\n{}")
UserAdviceMessages = [
widget.Message(
"Use CSV File Import widget for advanced options "
"for comma-separated files", "use-csv-file-import"),
widget.Message(
"This widget loads only tabular data. Use other widgets to load "
"other data types like models, distance matrices and networks.",
"other-data-types")
]
def __init__(self):
super().__init__()
RecentPathsWComboMixin.__init__(self)
self.domain = None
self.data = None
self.loaded_file = ""
self.reader = None
readers = [
f for f in FileFormat.formats
if getattr(f, 'read', None) and getattr(f, "EXTENSIONS", None)
]
def group_readers_per_addon_key(w):
# readers from Orange.data.io should go first
def package(w):
package = w.qualified_name().split(".")[:-1]
package = package[:2]
if ".".join(package) == "Orange.data":
return ["0"] # force "Orange" to come first
return package
return package(w), w.DESCRIPTION
self.available_readers = sorted(set(readers),
key=group_readers_per_addon_key)
layout = QGridLayout()
layout.setSpacing(4)
gui.widgetBox(self.controlArea, orientation=layout, box='Source')
vbox = gui.radioButtons(None,
self,
"source",
box=True,
callback=self.load_data,
addToLayout=False)
rb_button = gui.appendRadioButton(vbox, "File:", addToLayout=False)
layout.addWidget(rb_button, 0, 0, Qt.AlignVCenter)
box = gui.hBox(None, addToLayout=False, margin=0)
box.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.file_combo.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.file_combo.activated[int].connect(self.select_file)
box.layout().addWidget(self.file_combo)
layout.addWidget(box, 0, 1)
file_button = gui.button(None,
self,
'...',
callback=self.browse_file,
autoDefault=False)
file_button.setIcon(self.style().standardIcon(QStyle.SP_DirOpenIcon))
file_button.setSizePolicy(Policy.Maximum, Policy.Fixed)
layout.addWidget(file_button, 0, 2)
reload_button = gui.button(None,
self,
"Reload",
callback=self.load_data,
autoDefault=False)
reload_button.setIcon(self.style().standardIcon(
QStyle.SP_BrowserReload))
reload_button.setSizePolicy(Policy.Fixed, Policy.Fixed)
layout.addWidget(reload_button, 0, 3)
self.sheet_box = gui.hBox(None, addToLayout=False, margin=0)
self.sheet_combo = QComboBox()
self.sheet_combo.activated[str].connect(self.select_sheet)
self.sheet_combo.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.sheet_label = QLabel()
self.sheet_label.setText('Sheet')
self.sheet_label.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.sheet_box.layout().addWidget(self.sheet_label, Qt.AlignLeft)
self.sheet_box.layout().addWidget(self.sheet_combo, Qt.AlignVCenter)
layout.addWidget(self.sheet_box, 2, 1)
self.sheet_box.hide()
rb_button = gui.appendRadioButton(vbox, "URL:", addToLayout=False)
layout.addWidget(rb_button, 3, 0, Qt.AlignVCenter)
self.url_combo = url_combo = QComboBox()
url_model = NamedURLModel(self.sheet_names)
url_model.wrap(self.recent_urls)
url_combo.setLineEdit(LineEditSelectOnFocus())
url_combo.setModel(url_model)
url_combo.setSizePolicy(Policy.Ignored, Policy.Fixed)
url_combo.setEditable(True)
url_combo.setInsertPolicy(url_combo.InsertAtTop)
url_edit = url_combo.lineEdit()
l, t, r, b = url_edit.getTextMargins()
url_edit.setTextMargins(l + 5, t, r, b)
layout.addWidget(url_combo, 3, 1, 1, 3)
url_combo.activated.connect(self._url_set)
# whit completer we set that combo box is case sensitive when
# matching the history
completer = QCompleter()
completer.setCaseSensitivity(Qt.CaseSensitive)
url_combo.setCompleter(completer)
layout = QGridLayout()
layout.setSpacing(4)
gui.widgetBox(self.controlArea, orientation=layout, box='File Type')
box = gui.hBox(None, addToLayout=False, margin=0)
box.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.reader_combo = QComboBox(self)
self.reader_combo.setSizePolicy(Policy.MinimumExpanding, Policy.Fixed)
self.reader_combo.activated[int].connect(self.select_reader)
box.layout().addWidget(self.reader_combo)
layout.addWidget(box, 0, 1)
box = gui.vBox(self.controlArea, "Info")
self.infolabel = gui.widgetLabel(box, 'No data loaded.')
box = gui.widgetBox(self.controlArea, "Columns (Double click to edit)")
self.domain_editor = DomainEditor(self)
self.editor_model = self.domain_editor.model()
box.layout().addWidget(self.domain_editor)
box = gui.hBox(box)
gui.button(box,
self,
"Reset",
callback=self.reset_domain_edit,
autoDefault=False)
gui.rubber(box)
self.apply_button = gui.button(box,
self,
"Apply",
callback=self.apply_domain_edit)
self.apply_button.setEnabled(False)
self.apply_button.setFixedWidth(170)
self.editor_model.dataChanged.connect(
lambda: self.apply_button.setEnabled(True))
hBox = gui.hBox(self.controlArea)
gui.rubber(hBox)
gui.button(hBox,
self,
"Browse documentation datasets",
callback=lambda: self.browse_file(True),
autoDefault=False)
gui.rubber(hBox)
self.set_file_list()
# Must not call open_file from within __init__. open_file
# explicitly re-enters the event loop (by a progress bar)
self.setAcceptDrops(True)
if self.source == self.LOCAL_FILE:
last_path = self.last_path()
if last_path and os.path.exists(last_path) and \
os.path.getsize(last_path) > self.SIZE_LIMIT:
self.Warning.file_too_big()
return
QTimer.singleShot(0, self.load_data)
@staticmethod
def sizeHint():
return QSize(600, 550)
def select_file(self, n):
assert n < len(self.recent_paths)
super().select_file(n)
if self.recent_paths:
self.source = self.LOCAL_FILE
self.load_data()
self.set_file_list()
def select_sheet(self):
self.recent_paths[0].sheet = self.sheet_combo.currentText()
self.load_data()
def select_reader(self, n):
if self.source != self.LOCAL_FILE:
return # ignore for URL's
if self.recent_paths:
path = self.recent_paths[0]
if n == 0: # default
path.file_format = None
self.load_data()
elif n <= len(self.available_readers):
reader = self.available_readers[n - 1]
path.file_format = reader.qualified_name()
self.load_data()
else: # the rest include just qualified names
path.file_format = self.reader_combo.itemText(n)
self.load_data()
def _url_set(self):
url = self.url_combo.currentText()
pos = self.recent_urls.index(url)
url = url.strip()
if not urlparse(url).scheme:
url = 'http://' + url
self.url_combo.setItemText(pos, url)
self.recent_urls[pos] = url
self.source = self.URL
self.load_data()
def browse_file(self, in_demos=False):
if in_demos:
start_file = get_sample_datasets_dir()
if not os.path.exists(start_file):
QMessageBox.information(
None, "File",
"Cannot find the directory with documentation datasets")
return
else:
start_file = self.last_path() or os.path.expanduser("~/")
filename, reader, _ = open_filename_dialog(start_file, None,
self.available_readers)
if not filename:
return
self.add_path(filename)
if reader is not None:
self.recent_paths[0].file_format = reader.qualified_name()
self.source = self.LOCAL_FILE
self.load_data()
# Open a file, create data from it and send it over the data channel
def load_data(self):
# We need to catch any exception type since anything can happen in
# file readers
self.closeContext()
self.domain_editor.set_domain(None)
self.apply_button.setEnabled(False)
self.clear_messages()
self.set_file_list()
error = self._try_load()
if error:
error()
self.data = None
self.sheet_box.hide()
self.Outputs.data.send(None)
self.infolabel.setText("No data.")
def _try_load(self):
self._initialize_reader_combo()
# pylint: disable=broad-except
if self.source == self.LOCAL_FILE:
if self.last_path() is None:
return self.Information.no_file_selected
elif not os.path.exists(self.last_path()):
return self.Error.file_not_found
else:
url = self.url_combo.currentText().strip()
if not url:
return self.Information.no_file_selected
def mark_problematic_reader():
self.reader_combo.setItemData(self.reader_combo.currentIndex(),
QBrush(Qt.red), Qt.ForegroundRole)
try:
self.reader = self._get_reader() # also sets current reader index
assert self.reader is not None
except MissingReaderException:
mark_problematic_reader()
return self.Error.missing_reader
except Exception as ex:
mark_problematic_reader()
log.exception(ex)
return lambda x=ex: self.Error.unknown(str(x))
try:
self._update_sheet_combo()
except Exception:
return self.Error.sheet_error
with log_warnings() as warnings:
try:
data = self.reader.read()
except Exception as ex:
mark_problematic_reader()
log.exception(ex)
return lambda x=ex: self.Error.unknown(str(x))
if warnings:
self.Warning.load_warning(warnings[-1].message.args[0])
self.infolabel.setText(self._describe(data))
self.loaded_file = self.last_path()
add_origin(data, self.loaded_file)
self.data = data
self.openContext(data.domain)
self.apply_domain_edit() # sends data
return None
def _get_reader(self) -> FileFormat:
if self.source == self.LOCAL_FILE:
path = self.last_path()
self.reader_combo.setEnabled(True)
if self.recent_paths and self.recent_paths[0].file_format:
qname = self.recent_paths[0].file_format
qname_index = {
r.qualified_name(): i
for i, r in enumerate(self.available_readers)
}
if qname in qname_index:
self.reader_combo.setCurrentIndex(qname_index[qname] + 1)
else:
# reader may be accessible, but not in self.available_readers
# (perhaps its code was moved)
self.reader_combo.addItem(qname)
self.reader_combo.setCurrentIndex(
len(self.reader_combo) - 1)
try:
reader_class = class_from_qualified_name(qname)
except Exception as ex:
raise MissingReaderException(
f'Can not find reader "{qname}"') from ex
reader = reader_class(path)
else:
self.reader_combo.setCurrentIndex(0)
reader = FileFormat.get_reader(path)
if self.recent_paths and self.recent_paths[0].sheet:
reader.select_sheet(self.recent_paths[0].sheet)
return reader
else:
url = self.url_combo.currentText().strip()
return UrlReader(url)
def _update_sheet_combo(self):
if len(self.reader.sheets) < 2:
self.sheet_box.hide()
self.reader.select_sheet(None)
return
self.sheet_combo.clear()
self.sheet_combo.addItems(self.reader.sheets)
self._select_active_sheet()
self.sheet_box.show()
def _select_active_sheet(self):
try:
idx = self.reader.sheets.index(self.reader.sheet)
self.sheet_combo.setCurrentIndex(idx)
except ValueError:
# Requested sheet does not exist in this file
self.reader.select_sheet(None)
self.sheet_combo.setCurrentIndex(0)
def _initialize_reader_combo(self):
self.reader_combo.clear()
filters = [format_filter(f) for f in self.available_readers]
self.reader_combo.addItems([DEFAULT_READER_TEXT] + filters)
self.reader_combo.setCurrentIndex(0)
self.reader_combo.setDisabled(True)
# additional readers may be added in self._get_reader()
@staticmethod
def _describe(table):
def missing_prop(prop):
if prop:
return f"({prop * 100:.1f}% missing values)"
else:
return "(no missing values)"
domain = table.domain
text = ""
attrs = getattr(table, "attributes", {})
descs = [
attrs[desc] for desc in ("Name", "Description") if desc in attrs
]
if len(descs) == 2:
descs[0] = f"<b>{descs[0]}</b>"
if descs:
text += f"<p>{'<br/>'.join(descs)}</p>"
text += f"<p>{len(table)} instance(s)"
missing_in_attr = missing_prop(table.has_missing_attribute()
and table.get_nan_frequency_attribute())
missing_in_class = missing_prop(table.has_missing_class()
and table.get_nan_frequency_class())
text += f"<br/>{len(domain.attributes)} feature(s) {missing_in_attr}"
if domain.has_continuous_class:
text += f"<br/>Regression; numerical class {missing_in_class}"
elif domain.has_discrete_class:
text += "<br/>Classification; categorical class " \
f"with {len(domain.class_var.values)} values {missing_in_class}"
elif table.domain.class_vars:
text += "<br/>Multi-target; " \
f"{len(table.domain.class_vars)} target variables " \
f"{missing_in_class}"
else:
text += "<br/>Data has no target variable."
text += f"<br/>{len(domain.metas)} meta attribute(s)"
text += "</p>"
if 'Timestamp' in table.domain:
# Google Forms uses this header to timestamp responses
text += f"<p>First entry: {table[0, 'Timestamp']}<br/>" \
f"Last entry: {table[-1, 'Timestamp']}</p>"
return text
def storeSpecificSettings(self):
self.current_context.modified_variables = self.variables[:]
def retrieveSpecificSettings(self):
if hasattr(self.current_context, "modified_variables"):
self.variables[:] = self.current_context.modified_variables
def reset_domain_edit(self):
self.domain_editor.reset_domain()
self.apply_domain_edit()
def _inspect_discrete_variables(self, domain):
for var in chain(domain.variables, domain.metas):
if var.is_discrete and len(var.values) > 100:
self.Warning.performance_warning()
def apply_domain_edit(self):
self.Warning.performance_warning.clear()
self.Warning.renamed_vars.clear()
if self.data is None:
table = None
else:
domain, cols, renamed = \
self.domain_editor.get_domain(self.data.domain, self.data,
deduplicate=True)
if not (domain.variables or domain.metas):
table = None
elif domain is self.data.domain:
table = self.data
else:
X, y, m = cols
table = Table.from_numpy(domain, X, y, m, self.data.W)
table.name = self.data.name
table.ids = np.array(self.data.ids)
table.attributes = getattr(self.data, 'attributes', {})
self._inspect_discrete_variables(domain)
if renamed:
self.Warning.renamed_vars(f"Renamed: {', '.join(renamed)}")
self.Warning.multiple_targets(
shown=table is not None and len(table.domain.class_vars) > 1)
self.Outputs.data.send(table)
self.apply_button.setEnabled(False)
def get_widget_name_extension(self):
_, name = os.path.split(self.loaded_file)
return os.path.splitext(name)[0]
def send_report(self):
def get_ext_name(filename):
try:
return FileFormat.names[os.path.splitext(filename)[1]]
except KeyError:
return "unknown"
if self.data is None:
self.report_paragraph("File", "No file.")
return
if self.source == self.LOCAL_FILE:
home = os.path.expanduser("~")
if self.loaded_file.startswith(home):
# os.path.join does not like ~
name = "~" + os.path.sep + \
self.loaded_file[len(home):].lstrip("/").lstrip("\\")
else:
name = self.loaded_file
if self.sheet_combo.isVisible():
name += f" ({self.sheet_combo.currentText()})"
self.report_items("File", [("File name", name),
("Format", get_ext_name(name))])
else:
self.report_items("Data", [("Resource", self.url),
("Format", get_ext_name(self.url))])
self.report_data("Data", self.data)
@staticmethod
def dragEnterEvent(event):
"""Accept drops of valid file urls"""
urls = event.mimeData().urls()
if urls:
try:
FileFormat.get_reader(urls[0].toLocalFile())
event.acceptProposedAction()
except MissingReaderException:
pass
def dropEvent(self, event):
"""Handle file drops"""
urls = event.mimeData().urls()
if urls:
self.add_path(urls[0].toLocalFile()) # add first file
self.source = self.LOCAL_FILE
self.load_data()
def workflowEnvChanged(self, key, value, oldvalue):
"""
Function called when environment changes (e.g. while saving the scheme)
It make sure that all environment connected values are modified
(e.g. relative file paths are changed)
"""
self.update_file_list(key, value, oldvalue)
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 OWSql(OWBaseSql):
name = "SQL Table"
id = "orange.widgets.data.sql"
description = "Load dataset from SQL."
icon = "icons/SQLTable.svg"
priority = 30
category = "Data"
keywords = ["load"]
class Outputs:
data = Output("Data",
Table,
doc="Attribute-valued dataset read from the input file.")
settings_version = 2
table = Setting(None)
sql = Setting("")
guess_values = Setting(True)
download = Setting(False)
materialize = Setting(False)
materialize_table_name = Setting("")
class Information(OWBaseSql.Information):
data_sampled = Msg("Data description was generated from a sample.")
class Warning(OWBaseSql.Warning):
missing_extension = Msg("Database is missing extensions: {}")
class Error(OWBaseSql.Error):
no_backends = Msg("Please install a backend to use this widget.")
def __init__(self):
# Lint
self.backends = None
self.backendcombo = None
self.tables = None
self.tablecombo = None
self.sqltext = None
self.custom_sql = None
self.downloadcb = None
super().__init__()
def _setup_gui(self):
super()._setup_gui()
self._add_backend_controls()
self._add_tables_controls()
def _add_backend_controls(self):
box = self.serverbox
self.backends = BackendModel(Backend.available_backends())
self.backendcombo = QComboBox(box)
if self.backends:
self.backendcombo.setModel(self.backends)
else:
self.Error.no_backends()
box.setEnabled(False)
box.layout().insertWidget(0, self.backendcombo)
def _add_tables_controls(self):
vbox = gui.vBox(self.controlArea, "Tables", addSpace=True)
box = gui.vBox(vbox)
self.tables = TableModel()
self.tablecombo = QComboBox(
minimumContentsLength=35,
sizeAdjustPolicy=QComboBox.AdjustToMinimumContentsLength)
self.tablecombo.setModel(self.tables)
self.tablecombo.setToolTip('table')
self.tablecombo.activated[int].connect(self.select_table)
box.layout().addWidget(self.tablecombo)
self.custom_sql = gui.vBox(box)
self.custom_sql.setVisible(False)
self.sqltext = QTextEdit(self.custom_sql)
self.sqltext.setPlainText(self.sql)
self.custom_sql.layout().addWidget(self.sqltext)
mt = gui.hBox(self.custom_sql)
cb = gui.checkBox(mt, self, 'materialize', 'Materialize to table ')
cb.setToolTip('Save results of the query in a table')
le = gui.lineEdit(mt, self, 'materialize_table_name')
le.setToolTip('Save results of the query in a table')
gui.button(self.custom_sql, self, 'Execute', callback=self.open_table)
box.layout().addWidget(self.custom_sql)
gui.checkBox(box,
self,
"guess_values",
"Auto-discover categorical variables",
callback=self.open_table)
self.downloadcb = gui.checkBox(box,
self,
"download",
"Download data to local memory",
callback=self.open_table)
def highlight_error(self, text=""):
err = ['', 'QLineEdit {border: 2px solid red;}']
self.servertext.setStyleSheet(err['server' in text or 'host' in text])
self.usernametext.setStyleSheet(err['role' in text])
self.databasetext.setStyleSheet(err['database' in text])
def get_backend(self):
if self.backendcombo.currentIndex() < 0:
return None
return self.backends[self.backendcombo.currentIndex()]
def on_connection_success(self):
if getattr(self.backend, 'missing_extension', False):
self.Warning.missing_extension(", ".join(
self.backend.missing_extension))
self.download = True
self.downloadcb.setEnabled(False)
if not is_postgres(self.backend):
self.download = True
self.downloadcb.setEnabled(False)
super().on_connection_success()
self.refresh_tables()
self.select_table()
def on_connection_error(self, err):
super().on_connection_error(err)
self.highlight_error(str(err).split("\n")[0])
def clear(self):
super().clear()
self.Warning.missing_extension.clear()
self.downloadcb.setEnabled(True)
self.highlight_error()
self.tablecombo.clear()
self.tablecombo.repaint()
def refresh_tables(self):
self.tables.clear()
if self.backend is None:
self.data_desc_table = None
return
self.tables.append("Select a table")
self.tables.append("Custom SQL")
self.tables.extend(self.backend.list_tables(self.schema))
index = self.tablecombo.findText(str(self.table))
self.tablecombo.setCurrentIndex(index if index != -1 else 0)
self.tablecombo.repaint()
# Called on tablecombo selection change:
def select_table(self):
curIdx = self.tablecombo.currentIndex()
if self.tablecombo.itemText(curIdx) != "Custom SQL":
self.custom_sql.setVisible(False)
return self.open_table()
else:
self.custom_sql.setVisible(True)
self.data_desc_table = None
self.database_desc["Table"] = "(None)"
self.table = None
if len(str(self.sql)) > 14:
return self.open_table()
return None
def get_table(self):
curIdx = self.tablecombo.currentIndex()
if curIdx <= 0:
if self.database_desc:
self.database_desc["Table"] = "(None)"
self.data_desc_table = None
return None
if self.tablecombo.itemText(curIdx) != "Custom SQL":
self.table = self.tables[self.tablecombo.currentIndex()]
self.database_desc["Table"] = self.table
if "Query" in self.database_desc:
del self.database_desc["Query"]
what = self.table
else:
what = self.sql = self.sqltext.toPlainText()
self.table = "Custom SQL"
if self.materialize:
if not self.materialize_table_name:
self.Error.connection(
"Specify a table name to materialize the query")
return None
try:
with self.backend.execute_sql_query(
"DROP TABLE IF EXISTS " +
self.materialize_table_name):
pass
with self.backend.execute_sql_query(
"CREATE TABLE " + self.materialize_table_name +
" AS " + self.sql):
pass
with self.backend.execute_sql_query(
"ANALYZE " + self.materialize_table_name):
pass
except BackendError as ex:
self.Error.connection(str(ex))
return None
try:
table = SqlTable(dict(host=self.host,
port=self.port,
database=self.database,
user=self.username,
password=self.password),
what,
backend=type(self.backend),
inspect_values=False)
except BackendError as ex:
self.Error.connection(str(ex))
return None
self.Error.connection.clear()
sample = False
if table.approx_len() > LARGE_TABLE and self.guess_values:
confirm = QMessageBox(self)
confirm.setIcon(QMessageBox.Warning)
confirm.setText("Attribute discovery might take "
"a long time on large tables.\n"
"Do you want to auto discover attributes?")
confirm.addButton("Yes", QMessageBox.YesRole)
no_button = confirm.addButton("No", QMessageBox.NoRole)
if is_postgres(self.backend):
sample_button = confirm.addButton("Yes, on a sample",
QMessageBox.YesRole)
confirm.exec()
if confirm.clickedButton() == no_button:
self.guess_values = False
elif is_postgres(self.backend) and \
confirm.clickedButton() == sample_button:
sample = True
self.Information.clear()
if self.guess_values:
QApplication.setOverrideCursor(QCursor(Qt.WaitCursor))
if sample:
s = table.sample_time(1)
domain = s.get_domain(inspect_values=True)
self.Information.data_sampled()
else:
domain = table.get_domain(inspect_values=True)
QApplication.restoreOverrideCursor()
table.domain = domain
if self.download:
if table.approx_len() > AUTO_DL_LIMIT:
if is_postgres(self.backend):
confirm = QMessageBox(self)
confirm.setIcon(QMessageBox.Warning)
confirm.setText("Data appears to be big. Do you really "
"want to download it to local memory?")
if table.approx_len() <= MAX_DL_LIMIT:
confirm.addButton("Yes", QMessageBox.YesRole)
no_button = confirm.addButton("No", QMessageBox.NoRole)
sample_button = confirm.addButton("Yes, a sample",
QMessageBox.YesRole)
confirm.exec()
if confirm.clickedButton() == no_button:
return None
elif confirm.clickedButton() == sample_button:
table = table.sample_percentage(
AUTO_DL_LIMIT / table.approx_len() * 100)
else:
if table.approx_len() > MAX_DL_LIMIT:
QMessageBox.warning(self, 'Warning',
"Data is too big to download.\n")
return None
else:
confirm = QMessageBox.question(
self, 'Question',
"Data appears to be big. Do you really "
"want to download it to local memory?",
QMessageBox.Yes | QMessageBox.No, QMessageBox.No)
if confirm == QMessageBox.No:
return None
table.download_data(MAX_DL_LIMIT)
table = Table(table)
return table
@classmethod
def migrate_settings(cls, settings, version):
if version < 2:
# Until Orange version 3.4.4 username and password had been stored
# in Settings.
cm = cls._credential_manager(settings["host"], settings["port"])
cm.username = settings["username"]
cm.password = settings["password"]
class FileLoader(QWidget):
activated = pyqtSignal()
file_loaded = pyqtSignal()
def __init__(self):
super().__init__()
self.recent_paths = []
self.file_combo = QComboBox()
self.file_combo.setMinimumWidth(80)
self.file_combo.activated.connect(self._activate)
self.browse_btn = QPushButton("...")
icon = self.style().standardIcon(QStyle.SP_DirOpenIcon)
self.browse_btn.setIcon(icon)
self.browse_btn.setSizePolicy(QSizePolicy.Maximum, QSizePolicy.Fixed)
self.browse_btn.clicked.connect(self.browse)
self.load_btn = QPushButton("")
icon = self.style().standardIcon(QStyle.SP_BrowserReload)
self.load_btn.setIcon(icon)
self.load_btn.setSizePolicy(QSizePolicy.Maximum, QSizePolicy.Fixed)
self.load_btn.setAutoDefault(True)
self.load_btn.clicked.connect(self.file_loaded)
def browse(self):
start_file = self.last_path() or os.path.expanduser("~/")
formats = ["Text files (*.txt)", "All files (*)"]
file_name, _ = QFileDialog.getOpenFileName(None, "Open...", start_file,
";;".join(formats),
formats[0])
if not file_name:
return
self.add_path(file_name)
self._activate()
def _activate(self):
self.activated.emit()
self.file_loaded.emit()
def set_current_file(self, path: str):
if path:
self.add_path(path)
self.file_combo.setCurrentText(path)
else:
self.file_combo.setCurrentText("(none)")
def get_current_file(self) -> Optional[RecentPath]:
index = self.file_combo.currentIndex()
if index >= len(self.recent_paths) or index < 0:
return None
path = self.recent_paths[index]
return path if isinstance(path, RecentPath) else None
def add_path(self, filename: str):
recent = RecentPath.create(filename, [])
if recent in self.recent_paths:
self.recent_paths.remove(recent)
self.recent_paths.insert(0, recent)
self.set_file_list()
def set_file_list(self):
self.file_combo.clear()
for i, recent in enumerate(self.recent_paths):
self.file_combo.addItem(recent.basename)
self.file_combo.model().item(i).setToolTip(recent.abspath)
if not os.path.exists(recent.abspath):
self.file_combo.setItemData(i, QBrush(Qt.red),
Qt.TextColorRole)
self.file_combo.addItem(_DEFAULT_NONE)
def last_path(self) -> Optional[str]:
return self.recent_paths[0].abspath if self.recent_paths else None
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)
