class OWFeatureConstructor(OWWidget):
name = "Feature Constructor"
description = "Construct new features (data columns) from a set of " \
"existing features in the input data set."
icon = "icons/FeatureConstructor.svg"
inputs = [("Data", Orange.data.Table, "setData")]
outputs = [("Data", Orange.data.Table)]
want_main_area = False
settingsHandler = FeatureConstructorSettingsHandler()
descriptors = ContextSetting([])
currentIndex = ContextSetting(-1)
EDITORS = [
(ContinuousDescriptor, ContinuousFeatureEditor),
(DiscreteDescriptor, DiscreteFeatureEditor),
(StringDescriptor, StringFeatureEditor)
]
class Error(OWWidget.Error):
more_values_needed = Msg("Discrete feature {} needs more values.")
invalid_expressions = Msg("Invalid expressions: {}.")
def __init__(self):
super().__init__()
self.data = None
self.editors = {}
box = gui.vBox(self.controlArea, "Variable Definitions")
toplayout = QHBoxLayout()
toplayout.setContentsMargins(0, 0, 0, 0)
box.layout().addLayout(toplayout)
self.editorstack = QStackedWidget(
sizePolicy=QSizePolicy(QSizePolicy.MinimumExpanding,
QSizePolicy.MinimumExpanding)
)
for descclass, editorclass in self.EDITORS:
editor = editorclass()
editor.featureChanged.connect(self._on_modified)
self.editors[descclass] = editor
self.editorstack.addWidget(editor)
self.editorstack.setEnabled(False)
buttonlayout = QVBoxLayout(spacing=10)
buttonlayout.setContentsMargins(0, 0, 0, 0)
self.addbutton = QPushButton(
"New", toolTip="Create a new variable",
minimumWidth=120,
shortcut=QKeySequence.New
)
def unique_name(fmt, reserved):
candidates = (fmt.format(i) for i in count(1))
return next(c for c in candidates if c not in reserved)
def reserved_names():
varnames = []
if self.data is not None:
varnames = [var.name for var in
self.data.domain.variables + self.data.domain.metas]
varnames += [desc.name for desc in self.featuremodel]
return set(varnames)
def generate_newname(fmt):
return unique_name(fmt, reserved_names())
menu = QMenu(self.addbutton)
cont = menu.addAction("Continuous")
cont.triggered.connect(
lambda: self.addFeature(
ContinuousDescriptor(generate_newname("X{}"), "", 3))
)
disc = menu.addAction("Discrete")
disc.triggered.connect(
lambda: self.addFeature(
DiscreteDescriptor(generate_newname("D{}"), "",
("A", "B"), -1, False))
)
string = menu.addAction("String")
string.triggered.connect(
lambda: self.addFeature(
StringDescriptor(generate_newname("S{}"), ""))
)
menu.addSeparator()
self.duplicateaction = menu.addAction("Duplicate Selected Variable")
self.duplicateaction.triggered.connect(self.duplicateFeature)
self.duplicateaction.setEnabled(False)
self.addbutton.setMenu(menu)
self.removebutton = QPushButton(
"Remove", toolTip="Remove selected variable",
minimumWidth=120,
shortcut=QKeySequence.Delete
)
self.removebutton.clicked.connect(self.removeSelectedFeature)
buttonlayout.addWidget(self.addbutton)
buttonlayout.addWidget(self.removebutton)
buttonlayout.addStretch(10)
toplayout.addLayout(buttonlayout, 0)
toplayout.addWidget(self.editorstack, 10)
# Layout for the list view
layout = QVBoxLayout(spacing=1, margin=0)
self.featuremodel = DescriptorModel(parent=self)
self.featureview = QListView(
minimumWidth=200,
sizePolicy=QSizePolicy(QSizePolicy.Minimum,
QSizePolicy.MinimumExpanding)
)
self.featureview.setItemDelegate(FeatureItemDelegate(self))
self.featureview.setModel(self.featuremodel)
self.featureview.selectionModel().selectionChanged.connect(
self._on_selectedVariableChanged
)
layout.addWidget(self.featureview)
box.layout().addLayout(layout, 1)
box = gui.hBox(self.controlArea)
box.layout().addWidget(self.report_button)
self.report_button.setMinimumWidth(180)
gui.rubber(box)
commit = gui.button(box, self, "Send", callback=self.apply,
default=True)
commit.setMinimumWidth(180)
def setCurrentIndex(self, index):
index = min(index, len(self.featuremodel) - 1)
self.currentIndex = index
if index >= 0:
itemmodels.select_row(self.featureview, index)
desc = self.featuremodel[min(index, len(self.featuremodel) - 1)]
editor = self.editors[type(desc)]
self.editorstack.setCurrentWidget(editor)
editor.setEditorData(desc, self.data.domain if self.data else None)
self.editorstack.setEnabled(index >= 0)
self.duplicateaction.setEnabled(index >= 0)
self.removebutton.setEnabled(index >= 0)
def _on_selectedVariableChanged(self, selected, *_):
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
else:
self.setCurrentIndex(-1)
def _on_modified(self):
if self.currentIndex >= 0:
editor = self.editorstack.currentWidget()
self.featuremodel[self.currentIndex] = editor.editorData()
self.descriptors = list(self.featuremodel)
def setDescriptors(self, descriptors):
"""
Set a list of variable descriptors to edit.
"""
self.descriptors = descriptors
self.featuremodel[:] = list(self.descriptors)
@check_sql_input
def setData(self, data=None):
"""Set the input dataset."""
self.closeContext()
self.data = data
if self.data is not None:
descriptors = list(self.descriptors)
currindex = self.currentIndex
self.descriptors = []
self.currentIndex = -1
self.openContext(data)
if descriptors != self.descriptors or \
self.currentIndex != currindex:
# disconnect from the selection model while reseting the model
selmodel = self.featureview.selectionModel()
selmodel.selectionChanged.disconnect(
self._on_selectedVariableChanged)
self.featuremodel[:] = list(self.descriptors)
self.setCurrentIndex(self.currentIndex)
selmodel.selectionChanged.connect(
self._on_selectedVariableChanged)
self.editorstack.setEnabled(self.currentIndex >= 0)
def handleNewSignals(self):
if self.data is not None:
self.apply()
else:
self.send("Data", None)
def addFeature(self, descriptor):
self.featuremodel.append(descriptor)
self.setCurrentIndex(len(self.featuremodel) - 1)
editor = self.editorstack.currentWidget()
editor.nameedit.setFocus()
editor.nameedit.selectAll()
def removeFeature(self, index):
del self.featuremodel[index]
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
elif index is None and len(self.featuremodel) > 0:
# Deleting the last item clears selection
self.setCurrentIndex(len(self.featuremodel) - 1)
def removeSelectedFeature(self):
if self.currentIndex >= 0:
self.removeFeature(self.currentIndex)
def duplicateFeature(self):
desc = self.featuremodel[self.currentIndex]
self.addFeature(copy.deepcopy(desc))
def check_attrs_values(self, attr, data):
for i in range(len(data)):
for var in attr:
if not math.isnan(data[i, var]) \
and int(data[i, var]) >= len(var.values):
return var.name
return None
def _validate_descriptors(self, desc):
def validate(source):
try:
return validate_exp(ast.parse(source, mode="eval"))
except Exception:
return False
final = []
invalid = []
for d in desc:
if validate(d.expression):
final.append(d)
else:
final.append(d._replace(expression=""))
invalid.append(d)
if invalid:
self.Error.invalid_expressions(", ".join(s.name for s in invalid))
return final
def apply(self):
self.Error.clear()
if self.data is None:
return
desc = list(self.featuremodel)
desc = self._validate_descriptors(desc)
source_vars = tuple(self.data.domain) + self.data.domain.metas
new_variables = construct_variables(desc, source_vars)
attrs = [var for var in new_variables if var.is_primitive()]
metas = [var for var in new_variables if not var.is_primitive()]
new_domain = Orange.data.Domain(
self.data.domain.attributes + tuple(attrs),
self.data.domain.class_vars,
metas=self.data.domain.metas + tuple(metas)
)
try:
data = Orange.data.Table(new_domain, self.data)
except Exception as err:
log = logging.getLogger(__name__)
log.error("", exc_info=True)
self.error("".join(format_exception_only(type(err), err)).rstrip())
return
disc_attrs_not_ok = self.check_attrs_values(
[var for var in attrs if var.is_discrete], data)
if disc_attrs_not_ok:
self.Error.more_values_needed(disc_attrs_not_ok)
return
self.send("Data", data)
def send_report(self):
items = OrderedDict()
for feature in self.featuremodel:
if isinstance(feature, DiscreteDescriptor):
items[feature.name] = "{} (discrete with values {}{})".format(
feature.expression, feature.values,
"; ordered" * feature.ordered)
elif isinstance(feature, ContinuousDescriptor):
items[feature.name] = "{} (numeric)".format(feature.expression)
else:
items[feature.name] = "{} (text)".format(feature.expression)
self.report_items(
report.plural("Constructed feature{s}", len(items)), items)
class OWFeatureConstructor(OWWidget):
name = "特征构造器(Feature Constructor)"
description = "用输入数据集中的现有特征构造新特征。"
icon = "icons/FeatureConstructor.svg"
keywords = ['tezheng', 'gouzao', 'tezhenggouzao']
category = "数据(Data)"
icon = "icons/FeatureConstructor.svg"
class Inputs:
data = Input("数据(Data)", Orange.data.Table, replaces=['Data'])
class Outputs:
data = Output("数据(Data)", Orange.data.Table, replaces=['Data'])
want_main_area = False
settingsHandler = FeatureConstructorHandler()
descriptors = ContextSetting([])
currentIndex = ContextSetting(-1)
expressions_with_values = ContextSetting(False)
settings_version = 2
EDITORS = [(ContinuousDescriptor, ContinuousFeatureEditor),
(DateTimeDescriptor, DateTimeFeatureEditor),
(DiscreteDescriptor, DiscreteFeatureEditor),
(StringDescriptor, StringFeatureEditor)]
class Error(OWWidget.Error):
more_values_needed = Msg("Categorical feature {} needs more values.")
invalid_expressions = Msg("Invalid expressions: {}.")
class Warning(OWWidget.Warning):
renamed_var = Msg("Recently added variable has been renamed, "
"to avoid duplicates.\n")
def __init__(self):
super().__init__()
self.data = None
self.editors = {}
box = gui.vBox(self.controlArea, "变量定义")
toplayout = QHBoxLayout()
toplayout.setContentsMargins(0, 0, 0, 0)
box.layout().addLayout(toplayout)
self.editorstack = QStackedWidget(sizePolicy=QSizePolicy(
QSizePolicy.MinimumExpanding, QSizePolicy.MinimumExpanding))
for descclass, editorclass in self.EDITORS:
editor = editorclass()
editor.featureChanged.connect(self._on_modified)
self.editors[descclass] = editor
self.editorstack.addWidget(editor)
self.editorstack.setEnabled(False)
buttonlayout = QVBoxLayout(spacing=10)
buttonlayout.setContentsMargins(0, 0, 0, 0)
self.addbutton = QPushButton("新建",
toolTip="Create a new variable",
minimumWidth=120,
shortcut=QKeySequence.New)
def unique_name(fmt, reserved):
candidates = (fmt.format(i) for i in count(1))
return next(c for c in candidates if c not in reserved)
def generate_newname(fmt):
return unique_name(fmt, self.reserved_names())
menu = QMenu(self.addbutton)
cont = menu.addAction("数值数据")
cont.triggered.connect(lambda: self.addFeature(
ContinuousDescriptor(generate_newname("X{}"), "", 3)))
disc = menu.addAction("分类数据")
disc.triggered.connect(lambda: self.addFeature(
DiscreteDescriptor(generate_newname("D{}"), "", (), False)))
string = menu.addAction("文本")
string.triggered.connect(lambda: self.addFeature(
StringDescriptor(generate_newname("S{}"), "")))
datetime = menu.addAction("日期/时间")
datetime.triggered.connect(lambda: self.addFeature(
DateTimeDescriptor(generate_newname("T{}"), "")))
menu.addSeparator()
self.duplicateaction = menu.addAction("复制选中变量")
self.duplicateaction.triggered.connect(self.duplicateFeature)
self.duplicateaction.setEnabled(False)
self.addbutton.setMenu(menu)
self.removebutton = QPushButton("删除",
toolTip="删除选中变量",
minimumWidth=120,
shortcut=QKeySequence.Delete)
self.removebutton.clicked.connect(self.removeSelectedFeature)
buttonlayout.addWidget(self.addbutton)
buttonlayout.addWidget(self.removebutton)
buttonlayout.addStretch(10)
toplayout.addLayout(buttonlayout, 0)
toplayout.addWidget(self.editorstack, 10)
# Layout for the list view
layout = QVBoxLayout(spacing=1, margin=0)
self.featuremodel = DescriptorModel(parent=self)
self.featureview = QListView(minimumWidth=200,
minimumHeight=50,
sizePolicy=QSizePolicy(
QSizePolicy.Minimum,
QSizePolicy.MinimumExpanding))
self.featureview.setItemDelegate(FeatureItemDelegate(self))
self.featureview.setModel(self.featuremodel)
self.featureview.selectionModel().selectionChanged.connect(
self._on_selectedVariableChanged)
layout.addWidget(self.featureview)
box.layout().addLayout(layout, 1)
self.fix_button = gui.button(self.buttonsArea,
self,
"Upgrade Expressions",
callback=self.fix_expressions)
self.fix_button.setHidden(True)
gui.button(self.buttonsArea,
self,
"Send",
callback=self.apply,
default=True)
def setCurrentIndex(self, index):
index = min(index, len(self.featuremodel) - 1)
self.currentIndex = index
if index >= 0:
itemmodels.select_row(self.featureview, index)
desc = self.featuremodel[min(index, len(self.featuremodel) - 1)]
editor = self.editors[type(desc)]
self.editorstack.setCurrentWidget(editor)
editor.setEditorData(desc, self.data.domain if self.data else None)
self.editorstack.setEnabled(index >= 0)
self.duplicateaction.setEnabled(index >= 0)
self.removebutton.setEnabled(index >= 0)
def _on_selectedVariableChanged(self, selected, *_):
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
else:
self.setCurrentIndex(-1)
def _on_modified(self):
if self.currentIndex >= 0:
self.Warning.clear()
editor = self.editorstack.currentWidget()
proposed = editor.editorData().name
uniq = get_unique_names(self.reserved_names(self.currentIndex),
proposed)
feature = editor.editorData()
if editor.editorData().name != uniq:
self.Warning.renamed_var()
feature = feature.__class__(uniq, *feature[1:])
self.featuremodel[self.currentIndex] = feature
self.descriptors = list(self.featuremodel)
def setDescriptors(self, descriptors):
"""
Set a list of variable descriptors to edit.
"""
self.descriptors = descriptors
self.featuremodel[:] = list(self.descriptors)
def reserved_names(self, idx_=None):
varnames = []
if self.data is not None:
varnames = [
var.name
for var in self.data.domain.variables + self.data.domain.metas
]
varnames += [
desc.name for idx, desc in enumerate(self.featuremodel)
if idx != idx_
]
return set(varnames)
@Inputs.data
@check_sql_input
def setData(self, data=None):
"""Set the input dataset."""
self.closeContext()
self.data = data
self.expressions_with_values = False
self.descriptors = []
self.currentIndex = -1
if self.data is not None:
self.openContext(data)
# disconnect from the selection model while reseting the model
selmodel = self.featureview.selectionModel()
selmodel.selectionChanged.disconnect(self._on_selectedVariableChanged)
self.featuremodel[:] = list(self.descriptors)
self.setCurrentIndex(self.currentIndex)
selmodel.selectionChanged.connect(self._on_selectedVariableChanged)
self.fix_button.setHidden(not self.expressions_with_values)
self.editorstack.setEnabled(self.currentIndex >= 0)
def handleNewSignals(self):
if self.data is not None:
self.apply()
else:
self.Outputs.data.send(None)
self.fix_button.setHidden(True)
def addFeature(self, descriptor):
self.featuremodel.append(descriptor)
self.setCurrentIndex(len(self.featuremodel) - 1)
editor = self.editorstack.currentWidget()
editor.nameedit.setFocus()
editor.nameedit.selectAll()
def removeFeature(self, index):
del self.featuremodel[index]
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
elif index is None and self.featuremodel.rowCount():
# Deleting the last item clears selection
self.setCurrentIndex(self.featuremodel.rowCount() - 1)
def removeSelectedFeature(self):
if self.currentIndex >= 0:
self.removeFeature(self.currentIndex)
def duplicateFeature(self):
desc = self.featuremodel[self.currentIndex]
self.addFeature(copy.deepcopy(desc))
@staticmethod
def check_attrs_values(attr, data):
for i in range(len(data)):
for var in attr:
if not math.isnan(data[i, var]) \
and int(data[i, var]) >= len(var.values):
return var.name
return None
def _validate_descriptors(self, desc):
def validate(source):
try:
return validate_exp(ast.parse(source, mode="eval"))
# ast.parse can return arbitrary errors, not only SyntaxError
# pylint: disable=broad-except
except Exception:
return False
final = []
invalid = []
for d in desc:
if validate(d.expression):
final.append(d)
else:
final.append(d._replace(expression=""))
invalid.append(d)
if invalid:
self.Error.invalid_expressions(", ".join(s.name for s in invalid))
return final
def apply(self):
def report_error(err):
log = logging.getLogger(__name__)
log.error("", exc_info=True)
self.error("".join(format_exception_only(type(err), err)).rstrip())
self.Error.clear()
if self.data is None:
return
desc = list(self.featuremodel)
desc = self._validate_descriptors(desc)
try:
new_variables = construct_variables(desc, self.data,
self.expressions_with_values)
# user's expression can contain arbitrary errors
except Exception as err: # pylint: disable=broad-except
report_error(err)
return
attrs = [var for var in new_variables if var.is_primitive()]
metas = [var for var in new_variables if not var.is_primitive()]
new_domain = Orange.data.Domain(
self.data.domain.attributes + tuple(attrs),
self.data.domain.class_vars,
metas=self.data.domain.metas + tuple(metas))
try:
for variable in new_variables:
variable.compute_value.mask_exceptions = False
data = self.data.transform(new_domain)
# user's expression can contain arbitrary errors
# pylint: disable=broad-except
except Exception as err:
report_error(err)
return
finally:
for variable in new_variables:
variable.compute_value.mask_exceptions = True
disc_attrs_not_ok = self.check_attrs_values(
[var for var in attrs if var.is_discrete], data)
if disc_attrs_not_ok:
self.Error.more_values_needed(disc_attrs_not_ok)
return
self.Outputs.data.send(data)
def send_report(self):
items = OrderedDict()
for feature in self.featuremodel:
if isinstance(feature, DiscreteDescriptor):
desc = "categorical"
if feature.values:
desc += " with values " \
+ ", ".join(f"'{val}'" for val in feature.values)
if feature.ordered:
desc += "; ordered"
elif isinstance(feature, ContinuousDescriptor):
desc = "numeric"
elif isinstance(feature, DateTimeDescriptor):
desc = "date/time"
else:
desc = "text"
items[feature.name] = f"{feature.expression} ({desc})"
self.report_items(report.plural("Constructed feature{s}", len(items)),
items)
def fix_expressions(self):
dlg = QMessageBox(
QMessageBox.Question, "Fix Expressions",
"This widget's behaviour has changed. Values of categorical "
"variables are now inserted as their textual representations "
"(strings); previously they appeared as integer numbers, with an "
"attribute '.value' that contained the text.\n\n"
"The widget currently runs in compatibility mode. After "
"expressions are updated, manually check for their correctness.")
dlg.addButton("Update", QMessageBox.ApplyRole)
dlg.addButton("Cancel", QMessageBox.RejectRole)
if dlg.exec() == QMessageBox.RejectRole:
return
def fixer(mo):
var = domain[mo.group(2)]
if mo.group(3) == ".value": # uses string; remove `.value`
return "".join(mo.group(1, 2, 4))
# Uses ints: get them by indexing
return mo.group(1) + "{" + \
", ".join(f"'{val}': {i}"
for i, val in enumerate(var.values)) + \
f"}}[{var.name}]" + mo.group(4)
domain = self.data.domain
disc_vars = "|".join(f"{var.name}"
for var in chain(domain.variables, domain.metas)
if var.is_discrete)
expr = re.compile(r"(^|\W)(" + disc_vars + r")(\.value)?(\W|$)")
self.descriptors[:] = [
descriptor._replace(
expression=expr.sub(fixer, descriptor.expression))
for descriptor in self.descriptors
]
self.expressions_with_values = False
self.fix_button.hide()
index = self.currentIndex
self.featuremodel[:] = list(self.descriptors)
self.setCurrentIndex(index)
self.apply()
@classmethod
def migrate_context(cls, context, version):
if version is None or version < 2:
used_vars = set(
chain(*(
freevars(ast.parse(descriptor.expression, mode="eval"), [])
for descriptor in context.values["descriptors"]
if descriptor.expression)))
disc_vars = {
name
for (name, vtype) in chain(context.attributes.items(),
context.metas.items()) if vtype == 1
}
if used_vars & disc_vars:
context.values["expressions_with_values"] = True
class OWFeatureConstructor(OWWidget):
name = "Feature Constructor"
description = "Construct new features (data columns) from a set of " \
"existing features in the input data set."
icon = "icons/FeatureConstructor.svg"
class Inputs:
data = Input("Data", Orange.data.Table)
class Outputs:
data = Output("Data", Orange.data.Table)
want_main_area = False
settingsHandler = FeatureConstructorHandler()
descriptors = ContextSetting([])
currentIndex = ContextSetting(-1)
EDITORS = [
(ContinuousDescriptor, ContinuousFeatureEditor),
(DiscreteDescriptor, DiscreteFeatureEditor),
(StringDescriptor, StringFeatureEditor)
]
class Error(OWWidget.Error):
more_values_needed = Msg("Categorical feature {} needs more values.")
invalid_expressions = Msg("Invalid expressions: {}.")
def __init__(self):
super().__init__()
self.data = None
self.editors = {}
box = gui.vBox(self.controlArea, "Variable Definitions")
toplayout = QHBoxLayout()
toplayout.setContentsMargins(0, 0, 0, 0)
box.layout().addLayout(toplayout)
self.editorstack = QStackedWidget(
sizePolicy=QSizePolicy(QSizePolicy.MinimumExpanding,
QSizePolicy.MinimumExpanding)
)
for descclass, editorclass in self.EDITORS:
editor = editorclass()
editor.featureChanged.connect(self._on_modified)
self.editors[descclass] = editor
self.editorstack.addWidget(editor)
self.editorstack.setEnabled(False)
buttonlayout = QVBoxLayout(spacing=10)
buttonlayout.setContentsMargins(0, 0, 0, 0)
self.addbutton = QPushButton(
"New", toolTip="Create a new variable",
minimumWidth=120,
shortcut=QKeySequence.New
)
def unique_name(fmt, reserved):
candidates = (fmt.format(i) for i in count(1))
return next(c for c in candidates if c not in reserved)
def reserved_names():
varnames = []
if self.data is not None:
varnames = [var.name for var in
self.data.domain.variables + self.data.domain.metas]
varnames += [desc.name for desc in self.featuremodel]
return set(varnames)
def generate_newname(fmt):
return unique_name(fmt, reserved_names())
menu = QMenu(self.addbutton)
cont = menu.addAction("Numeric")
cont.triggered.connect(
lambda: self.addFeature(
ContinuousDescriptor(generate_newname("X{}"), "", 3))
)
disc = menu.addAction("Categorical")
disc.triggered.connect(
lambda: self.addFeature(
DiscreteDescriptor(generate_newname("D{}"), "",
("A", "B"), -1, False))
)
string = menu.addAction("Text")
string.triggered.connect(
lambda: self.addFeature(
StringDescriptor(generate_newname("S{}"), ""))
)
menu.addSeparator()
self.duplicateaction = menu.addAction("Duplicate Selected Variable")
self.duplicateaction.triggered.connect(self.duplicateFeature)
self.duplicateaction.setEnabled(False)
self.addbutton.setMenu(menu)
self.removebutton = QPushButton(
"Remove", toolTip="Remove selected variable",
minimumWidth=120,
shortcut=QKeySequence.Delete
)
self.removebutton.clicked.connect(self.removeSelectedFeature)
buttonlayout.addWidget(self.addbutton)
buttonlayout.addWidget(self.removebutton)
buttonlayout.addStretch(10)
toplayout.addLayout(buttonlayout, 0)
toplayout.addWidget(self.editorstack, 10)
# Layout for the list view
layout = QVBoxLayout(spacing=1, margin=0)
self.featuremodel = DescriptorModel(parent=self)
self.featureview = QListView(
minimumWidth=200,
sizePolicy=QSizePolicy(QSizePolicy.Minimum,
QSizePolicy.MinimumExpanding)
)
self.featureview.setItemDelegate(FeatureItemDelegate(self))
self.featureview.setModel(self.featuremodel)
self.featureview.selectionModel().selectionChanged.connect(
self._on_selectedVariableChanged
)
layout.addWidget(self.featureview)
box.layout().addLayout(layout, 1)
box = gui.hBox(self.controlArea)
gui.rubber(box)
commit = gui.button(box, self, "Send", callback=self.apply,
default=True)
commit.setMinimumWidth(180)
def setCurrentIndex(self, index):
index = min(index, len(self.featuremodel) - 1)
self.currentIndex = index
if index >= 0:
itemmodels.select_row(self.featureview, index)
desc = self.featuremodel[min(index, len(self.featuremodel) - 1)]
editor = self.editors[type(desc)]
self.editorstack.setCurrentWidget(editor)
editor.setEditorData(desc, self.data.domain if self.data else None)
self.editorstack.setEnabled(index >= 0)
self.duplicateaction.setEnabled(index >= 0)
self.removebutton.setEnabled(index >= 0)
def _on_selectedVariableChanged(self, selected, *_):
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
else:
self.setCurrentIndex(-1)
def _on_modified(self):
if self.currentIndex >= 0:
editor = self.editorstack.currentWidget()
self.featuremodel[self.currentIndex] = editor.editorData()
self.descriptors = list(self.featuremodel)
def setDescriptors(self, descriptors):
"""
Set a list of variable descriptors to edit.
"""
self.descriptors = descriptors
self.featuremodel[:] = list(self.descriptors)
@Inputs.data
@check_sql_input
def setData(self, data=None):
"""Set the input dataset."""
self.closeContext()
self.data = data
if self.data is not None:
descriptors = list(self.descriptors)
currindex = self.currentIndex
self.descriptors = []
self.currentIndex = -1
self.openContext(data)
if descriptors != self.descriptors or \
self.currentIndex != currindex:
# disconnect from the selection model while reseting the model
selmodel = self.featureview.selectionModel()
selmodel.selectionChanged.disconnect(
self._on_selectedVariableChanged)
self.featuremodel[:] = list(self.descriptors)
self.setCurrentIndex(self.currentIndex)
selmodel.selectionChanged.connect(
self._on_selectedVariableChanged)
self.editorstack.setEnabled(self.currentIndex >= 0)
def handleNewSignals(self):
if self.data is not None:
self.apply()
else:
self.Outputs.data.send(None)
def addFeature(self, descriptor):
self.featuremodel.append(descriptor)
self.setCurrentIndex(len(self.featuremodel) - 1)
editor = self.editorstack.currentWidget()
editor.nameedit.setFocus()
editor.nameedit.selectAll()
def removeFeature(self, index):
del self.featuremodel[index]
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
elif index is None and len(self.featuremodel) > 0:
# Deleting the last item clears selection
self.setCurrentIndex(len(self.featuremodel) - 1)
def removeSelectedFeature(self):
if self.currentIndex >= 0:
self.removeFeature(self.currentIndex)
def duplicateFeature(self):
desc = self.featuremodel[self.currentIndex]
self.addFeature(copy.deepcopy(desc))
def check_attrs_values(self, attr, data):
for i in range(len(data)):
for var in attr:
if not math.isnan(data[i, var]) \
and int(data[i, var]) >= len(var.values):
return var.name
return None
def _validate_descriptors(self, desc):
def validate(source):
try:
return validate_exp(ast.parse(source, mode="eval"))
except Exception:
return False
final = []
invalid = []
for d in desc:
if validate(d.expression):
final.append(d)
else:
final.append(d._replace(expression=""))
invalid.append(d)
if invalid:
self.Error.invalid_expressions(", ".join(s.name for s in invalid))
return final
def apply(self):
self.Error.clear()
if self.data is None:
return
desc = list(self.featuremodel)
desc = self._validate_descriptors(desc)
source_vars = self.data.domain.variables + self.data.domain.metas
new_variables = construct_variables(desc, source_vars)
attrs = [var for var in new_variables if var.is_primitive()]
metas = [var for var in new_variables if not var.is_primitive()]
new_domain = Orange.data.Domain(
self.data.domain.attributes + tuple(attrs),
self.data.domain.class_vars,
metas=self.data.domain.metas + tuple(metas)
)
try:
data = self.data.transform(new_domain)
except Exception as err:
log = logging.getLogger(__name__)
log.error("", exc_info=True)
self.error("".join(format_exception_only(type(err), err)).rstrip())
return
disc_attrs_not_ok = self.check_attrs_values(
[var for var in attrs if var.is_discrete], data)
if disc_attrs_not_ok:
self.Error.more_values_needed(disc_attrs_not_ok)
return
self.Outputs.data.send(data)
def send_report(self):
items = OrderedDict()
for feature in self.featuremodel:
if isinstance(feature, DiscreteDescriptor):
items[feature.name] = "{} (categorical with values {}{})".format(
feature.expression, feature.values,
"; ordered" * feature.ordered)
elif isinstance(feature, ContinuousDescriptor):
items[feature.name] = "{} (numeric)".format(feature.expression)
else:
items[feature.name] = "{} (text)".format(feature.expression)
self.report_items(
report.plural("Constructed feature{s}", len(items)), items)
class OWFeatureConstructor(OWWidget):
name = "Feature Constructor"
description = "Construct new features (data columns) from a set of " \
"existing features in the input dataset."
icon = "icons/FeatureConstructor.svg"
keywords = ['function', 'lambda']
class Inputs:
data = Input("Data", Orange.data.Table)
class Outputs:
data = Output("Data", Orange.data.Table)
want_main_area = False
settingsHandler = FeatureConstructorHandler()
descriptors = ContextSetting([])
currentIndex = ContextSetting(-1)
EDITORS = [(ContinuousDescriptor, ContinuousFeatureEditor),
(DateTimeDescriptor, DateTimeFeatureEditor),
(DiscreteDescriptor, DiscreteFeatureEditor),
(StringDescriptor, StringFeatureEditor)]
class Error(OWWidget.Error):
more_values_needed = Msg("Categorical feature {} needs more values.")
invalid_expressions = Msg("Invalid expressions: {}.")
class Warning(OWWidget.Warning):
renamed_var = Msg("Recently added variable has been renamed, "
"to avoid duplicates.\n")
def __init__(self):
super().__init__()
self.data = None
self.editors = {}
box = gui.vBox(self.controlArea, "Variable Definitions")
toplayout = QHBoxLayout()
toplayout.setContentsMargins(0, 0, 0, 0)
box.layout().addLayout(toplayout)
self.editorstack = QStackedWidget(sizePolicy=QSizePolicy(
QSizePolicy.MinimumExpanding, QSizePolicy.MinimumExpanding))
for descclass, editorclass in self.EDITORS:
editor = editorclass()
editor.featureChanged.connect(self._on_modified)
self.editors[descclass] = editor
self.editorstack.addWidget(editor)
self.editorstack.setEnabled(False)
buttonlayout = QVBoxLayout(spacing=10)
buttonlayout.setContentsMargins(0, 0, 0, 0)
self.addbutton = QPushButton("New",
toolTip="Create a new variable",
minimumWidth=120,
shortcut=QKeySequence.New)
def unique_name(fmt, reserved):
candidates = (fmt.format(i) for i in count(1))
return next(c for c in candidates if c not in reserved)
def generate_newname(fmt):
return unique_name(fmt, self.reserved_names())
menu = QMenu(self.addbutton)
cont = menu.addAction("Numeric")
cont.triggered.connect(lambda: self.addFeature(
ContinuousDescriptor(generate_newname("X{}"), "", 3)))
disc = menu.addAction("Categorical")
disc.triggered.connect(lambda: self.addFeature(
DiscreteDescriptor(generate_newname("D{}"), "", (), False)))
string = menu.addAction("Text")
string.triggered.connect(lambda: self.addFeature(
StringDescriptor(generate_newname("S{}"), "")))
datetime = menu.addAction("Date/Time")
datetime.triggered.connect(lambda: self.addFeature(
DateTimeDescriptor(generate_newname("T{}"), "")))
menu.addSeparator()
self.duplicateaction = menu.addAction("Duplicate Selected Variable")
self.duplicateaction.triggered.connect(self.duplicateFeature)
self.duplicateaction.setEnabled(False)
self.addbutton.setMenu(menu)
self.removebutton = QPushButton("Remove",
toolTip="Remove selected variable",
minimumWidth=120,
shortcut=QKeySequence.Delete)
self.removebutton.clicked.connect(self.removeSelectedFeature)
buttonlayout.addWidget(self.addbutton)
buttonlayout.addWidget(self.removebutton)
buttonlayout.addStretch(10)
toplayout.addLayout(buttonlayout, 0)
toplayout.addWidget(self.editorstack, 10)
# Layout for the list view
layout = QVBoxLayout(spacing=1, margin=0)
self.featuremodel = DescriptorModel(parent=self)
self.featureview = QListView(minimumWidth=200,
minimumHeight=50,
sizePolicy=QSizePolicy(
QSizePolicy.Minimum,
QSizePolicy.MinimumExpanding))
self.featureview.setItemDelegate(FeatureItemDelegate(self))
self.featureview.setModel(self.featuremodel)
self.featureview.selectionModel().selectionChanged.connect(
self._on_selectedVariableChanged)
self.info.set_input_summary(self.info.NoInput)
self.info.set_output_summary(self.info.NoOutput)
layout.addWidget(self.featureview)
box.layout().addLayout(layout, 1)
gui.button(self.buttonsArea,
self,
"Send",
callback=self.apply,
default=True)
def setCurrentIndex(self, index):
index = min(index, len(self.featuremodel) - 1)
self.currentIndex = index
if index >= 0:
itemmodels.select_row(self.featureview, index)
desc = self.featuremodel[min(index, len(self.featuremodel) - 1)]
editor = self.editors[type(desc)]
self.editorstack.setCurrentWidget(editor)
editor.setEditorData(desc, self.data.domain if self.data else None)
self.editorstack.setEnabled(index >= 0)
self.duplicateaction.setEnabled(index >= 0)
self.removebutton.setEnabled(index >= 0)
def _on_selectedVariableChanged(self, selected, *_):
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
else:
self.setCurrentIndex(-1)
def _on_modified(self):
if self.currentIndex >= 0:
self.Warning.clear()
editor = self.editorstack.currentWidget()
proposed = editor.editorData().name
unique = get_unique_names(self.reserved_names(self.currentIndex),
proposed)
feature = editor.editorData()
if editor.editorData().name != unique:
self.Warning.renamed_var()
feature = feature.__class__(unique, *feature[1:])
self.featuremodel[self.currentIndex] = feature
self.descriptors = list(self.featuremodel)
def setDescriptors(self, descriptors):
"""
Set a list of variable descriptors to edit.
"""
self.descriptors = descriptors
self.featuremodel[:] = list(self.descriptors)
def reserved_names(self, idx_=None):
varnames = []
if self.data is not None:
varnames = [
var.name
for var in self.data.domain.variables + self.data.domain.metas
]
varnames += [
desc.name for idx, desc in enumerate(self.featuremodel)
if idx != idx_
]
return set(varnames)
@Inputs.data
@check_sql_input
def setData(self, data=None):
"""Set the input dataset."""
self.closeContext()
self.data = data
self.info.set_input_summary(self.info.NoInput)
if self.data is not None:
descriptors = list(self.descriptors)
currindex = self.currentIndex
self.descriptors = []
self.currentIndex = -1
self.openContext(data)
self.info.set_input_summary(len(data),
format_summary_details(data))
if descriptors != self.descriptors or \
self.currentIndex != currindex:
# disconnect from the selection model while reseting the model
selmodel = self.featureview.selectionModel()
selmodel.selectionChanged.disconnect(
self._on_selectedVariableChanged)
self.featuremodel[:] = list(self.descriptors)
self.setCurrentIndex(self.currentIndex)
selmodel.selectionChanged.connect(
self._on_selectedVariableChanged)
self.editorstack.setEnabled(self.currentIndex >= 0)
def handleNewSignals(self):
if self.data is not None:
self.apply()
else:
self.info.set_output_summary(self.info.NoOutput)
self.Outputs.data.send(None)
def addFeature(self, descriptor):
self.featuremodel.append(descriptor)
self.setCurrentIndex(len(self.featuremodel) - 1)
editor = self.editorstack.currentWidget()
editor.nameedit.setFocus()
editor.nameedit.selectAll()
def removeFeature(self, index):
del self.featuremodel[index]
index = selected_row(self.featureview)
if index is not None:
self.setCurrentIndex(index)
elif index is None and self.featuremodel.rowCount():
# Deleting the last item clears selection
self.setCurrentIndex(self.featuremodel.rowCount() - 1)
def removeSelectedFeature(self):
if self.currentIndex >= 0:
self.removeFeature(self.currentIndex)
def duplicateFeature(self):
desc = self.featuremodel[self.currentIndex]
self.addFeature(copy.deepcopy(desc))
@staticmethod
def check_attrs_values(attr, data):
for i in range(len(data)):
for var in attr:
if not math.isnan(data[i, var]) \
and int(data[i, var]) >= len(var.values):
return var.name
return None
def _validate_descriptors(self, desc):
def validate(source):
try:
return validate_exp(ast.parse(source, mode="eval"))
# ast.parse can return arbitrary errors, not only SyntaxError
# pylint: disable=broad-except
except Exception:
return False
final = []
invalid = []
for d in desc:
if validate(d.expression):
final.append(d)
else:
final.append(d._replace(expression=""))
invalid.append(d)
if invalid:
self.Error.invalid_expressions(", ".join(s.name for s in invalid))
return final
def apply(self):
def report_error(err):
log = logging.getLogger(__name__)
log.error("", exc_info=True)
self.error("".join(format_exception_only(type(err), err)).rstrip())
self.Error.clear()
if self.data is None:
return
desc = list(self.featuremodel)
desc = self._validate_descriptors(desc)
try:
new_variables = construct_variables(desc, self.data)
# user's expression can contain arbitrary errors
except Exception as err: # pylint: disable=broad-except
report_error(err)
return
attrs = [var for var in new_variables if var.is_primitive()]
metas = [var for var in new_variables if not var.is_primitive()]
new_domain = Orange.data.Domain(
self.data.domain.attributes + tuple(attrs),
self.data.domain.class_vars,
metas=self.data.domain.metas + tuple(metas))
try:
for variable in new_variables:
variable.compute_value.mask_exceptions = False
data = self.data.transform(new_domain)
# user's expression can contain arbitrary errors
# pylint: disable=broad-except
except Exception as err:
report_error(err)
return
finally:
for variable in new_variables:
variable.compute_value.mask_exceptions = True
disc_attrs_not_ok = self.check_attrs_values(
[var for var in attrs if var.is_discrete], data)
if disc_attrs_not_ok:
self.Error.more_values_needed(disc_attrs_not_ok)
return
self.info.set_output_summary(len(data), format_summary_details(data))
self.Outputs.data.send(data)
def send_report(self):
items = OrderedDict()
for feature in self.featuremodel:
if isinstance(feature, DiscreteDescriptor):
items[
feature.name] = "{} (categorical with values {}{})".format(
feature.expression, feature.values,
"; ordered" * feature.ordered)
elif isinstance(feature, ContinuousDescriptor):
items[feature.name] = "{} (numeric)".format(feature.expression)
elif isinstance(feature, DateTimeDescriptor):
items[feature.name] = "{} (date/time)".format(
feature.expression)
else:
items[feature.name] = "{} (text)".format(feature.expression)
self.report_items(report.plural("Constructed feature{s}", len(items)),
items)