class OWHierarchicalClustering(widget.OWWidget):
name = "Hierarchical Clustering"
description = ("Hierarchical clustering based on distance matrix, and "
"a dendrogram viewer.")
icon = "icons/HierarchicalClustering.svg"
priority = 2100
inputs = [("Distances", Orange.misc.DistMatrix, "set_distances")]
outputs = [("Selected Data", Orange.data.Table),
("Other Data", Orange.data.Table)]
#: Selected linkage
linkage = settings.Setting(1)
#: Index of the selected annotation item (variable, ...)
annotation_idx = settings.Setting(0)
#: Selected tree pruning (none/max depth)
pruning = settings.Setting(0)
#: Maximum depth when max depth pruning is selected
max_depth = settings.Setting(10)
#: Selected cluster selection method (none, cut distance, top n)
selection_method = settings.Setting(0)
#: Cut height ratio wrt root height
cut_ratio = settings.Setting(75.0)
#: Number of top clusters to select
top_n = settings.Setting(3)
append_clusters = settings.Setting(True)
cluster_role = settings.Setting(2)
cluster_name = settings.Setting("Cluster")
autocommit = settings.Setting(False)
#: Cluster variable domain role
AttributeRole, ClassRole, MetaRole = 0, 1, 2
def __init__(self, parent=None):
super().__init__(parent)
self.matrix = None
self.items = None
self.linkmatrix = None
self.root = None
self._displayed_root = None
self.cutoff_height = 0.0
self._invalidated = False
gui.comboBox(gui.widgetBox(self.controlArea, "Linkage"),
self,
"linkage",
items=LINKAGE,
callback=self._invalidate_clustering)
box = gui.widgetBox(self.controlArea, "Annotation")
self.label_cb = gui.comboBox(box,
self,
"annotation_idx",
callback=self._update_labels)
self.label_cb.setModel(itemmodels.VariableListModel())
self.label_cb.model()[:] = ["None", "Enumeration"]
box = gui.radioButtons(self.controlArea,
self,
"pruning",
box="Pruning",
callback=self._invalidate_pruning)
grid = QGridLayout()
box.layout().addLayout(grid)
grid.addWidget(gui.appendRadioButton(box, "None", addToLayout=False),
0, 0)
self.max_depth_spin = gui.spin(box,
self,
"max_depth",
minv=1,
maxv=100,
callback=self._invalidate_pruning,
keyboardTracking=False)
grid.addWidget(
gui.appendRadioButton(box, "Max depth", addToLayout=False), 1, 0)
grid.addWidget(self.max_depth_spin, 1, 1)
box = gui.radioButtons(self.controlArea,
self,
"selection_method",
box="Selection",
callback=self._selection_method_changed)
grid = QGridLayout()
box.layout().addLayout(grid)
grid.addWidget(gui.appendRadioButton(box, "Manual", addToLayout=False),
0, 0)
grid.addWidget(
gui.appendRadioButton(box, "Height ratio", addToLayout=False), 1,
0)
self.cut_ratio_spin = gui.spin(box,
self,
"cut_ratio",
0,
100,
step=1e-1,
spinType=float,
callback=self._selection_method_changed)
self.cut_ratio_spin.setSuffix("%")
grid.addWidget(self.cut_ratio_spin, 1, 1)
grid.addWidget(gui.appendRadioButton(box, "Top N", addToLayout=False),
2, 0)
self.top_n_spin = gui.spin(box,
self,
"top_n",
1,
20,
callback=self._selection_method_changed)
grid.addWidget(self.top_n_spin, 2, 1)
box.layout().addLayout(grid)
self.controlArea.layout().addStretch()
box = gui.widgetBox(self.controlArea, "Output")
gui.checkBox(box,
self,
"append_clusters",
"Append cluster IDs",
callback=self._invalidate_output)
ibox = gui.indentedBox(box)
name_edit = gui.lineEdit(ibox, self, "cluster_name")
name_edit.editingFinished.connect(self._invalidate_output)
cb = gui.comboBox(
ibox,
self,
"cluster_role",
callback=self._invalidate_output,
items=["Attribute", "Class variable", "Meta variable"])
form = QFormLayout(fieldGrowthPolicy=QFormLayout.AllNonFixedFieldsGrow,
labelAlignment=Qt.AlignLeft,
spacing=8)
form.addRow("Name", name_edit)
form.addRow("Place", cb)
ibox.layout().addSpacing(5)
ibox.layout().addLayout(form)
ibox.layout().addSpacing(5)
cb = gui.checkBox(box, self, "autocommit", "Commit automatically")
b = gui.button(box, self, "Commit", callback=self.commit, default=True)
gui.setStopper(self, b, cb, "_invalidated", callback=self.commit)
self.scene = QGraphicsScene()
self.view = QGraphicsView(
self.scene,
horizontalScrollBarPolicy=Qt.ScrollBarAlwaysOff,
verticalScrollBarPolicy=Qt.ScrollBarAlwaysOn,
alignment=Qt.AlignLeft | Qt.AlignVCenter)
def axis_view(orientation):
ax = pg.AxisItem(orientation=orientation, maxTickLength=7)
scene = QGraphicsScene()
scene.addItem(ax)
view = QGraphicsView(
scene,
horizontalScrollBarPolicy=Qt.ScrollBarAlwaysOff,
verticalScrollBarPolicy=Qt.ScrollBarAlwaysOn,
alignment=Qt.AlignLeft | Qt.AlignVCenter)
view.setFixedHeight(ax.size().height())
ax.line = SliderLine(orientation=Qt.Horizontal,
length=ax.size().height())
scene.addItem(ax.line)
return view, ax
self.top_axis_view, self.top_axis = axis_view("top")
self.mainArea.layout().setSpacing(1)
self.mainArea.layout().addWidget(self.top_axis_view)
self.mainArea.layout().addWidget(self.view)
self.bottom_axis_view, self.bottom_axis = axis_view("bottom")
self.mainArea.layout().addWidget(self.bottom_axis_view)
self._main_graphics = QGraphicsWidget()
self._main_layout = QGraphicsLinearLayout(Qt.Horizontal)
self._main_layout.setSpacing(1)
self._main_graphics.setLayout(self._main_layout)
self.scene.addItem(self._main_graphics)
self.dendrogram = DendrogramWidget()
self.dendrogram.setSizePolicy(QSizePolicy.MinimumExpanding,
QSizePolicy.MinimumExpanding)
self.dendrogram.selectionChanged.connect(self._invalidate_output)
self.dendrogram.selectionEdited.connect(self._selection_edited)
fm = self.fontMetrics()
self.dendrogram.setContentsMargins(5,
fm.lineSpacing() / 2, 5,
fm.lineSpacing() / 2)
self.labels = GraphicsSimpleTextList()
self.labels.setSizePolicy(QSizePolicy.Preferred, QSizePolicy.Preferred)
self.labels.setAlignment(Qt.AlignLeft)
self.labels.setMaximumWidth(200)
self.labels.layout().setSpacing(0)
self._main_layout.addItem(self.dendrogram)
self._main_layout.addItem(self.labels)
self._main_layout.setAlignment(self.dendrogram,
Qt.AlignLeft | Qt.AlignVCenter)
self._main_layout.setAlignment(self.labels,
Qt.AlignLeft | Qt.AlignVCenter)
self.view.viewport().installEventFilter(self)
self.top_axis_view.viewport().installEventFilter(self)
self.bottom_axis_view.viewport().installEventFilter(self)
self._main_graphics.installEventFilter(self)
self.cut_line = SliderLine(self.dendrogram, orientation=Qt.Horizontal)
self.cut_line.valueChanged.connect(self._dendrogram_slider_changed)
self.cut_line.hide()
self.bottom_axis.line.valueChanged.connect(self._axis_slider_changed)
self.top_axis.line.valueChanged.connect(self._axis_slider_changed)
self.dendrogram.geometryChanged.connect(self._dendrogram_geom_changed)
self._set_cut_line_visible(self.selection_method == 1)
def set_distances(self, matrix):
self.matrix = matrix
self._invalidate_clustering()
self._set_items(matrix.row_items if matrix is not None else None)
def _set_items(self, items):
self.items = items
if items is None:
self.label_cb.model()[:] = ["None", "Enumeration"]
elif isinstance(items, Orange.data.Table):
vars = list(items.domain)
self.label_cb.model()[:] = ["None", "Enumeration"] + vars
elif isinstance(items, list) and \
all(isinstance(var, Orange.data.Variable) for var in items):
self.label_cb.model()[:] = ["None", "Enumeration", "Name"]
else:
self.label_cb.model()[:] = ["None", "Enumeration"]
self.annotation_idx = min(self.annotation_idx,
len(self.label_cb.model()) - 1)
def handleNewSignals(self):
self._update_labels()
def _clear_plot(self):
self.labels.set_labels([])
self.dendrogram.set_root(None)
def _set_displayed_root(self, root):
self._clear_plot()
self._displayed_root = root
self.dendrogram.set_root(root)
self._update_labels()
self._main_graphics.resize(
self._main_graphics.size().width(),
self._main_graphics.sizeHint(Qt.PreferredSize).height())
self._main_graphics.layout().activate()
def _update(self):
self._clear_plot()
distances = self.matrix
if distances is not None:
# Convert to flat upper triangular distances
i, j = numpy.triu_indices(distances.X.shape[0], k=1)
distances = distances.X[i, j]
method = LINKAGE[self.linkage].lower()
Z = scipy.cluster.hierarchy.linkage(distances, method=method)
tree = tree_from_linkage(Z)
self.linkmatrix = Z
self.root = tree
self.top_axis.setRange(tree.value.height, 0.0)
self.bottom_axis.setRange(tree.value.height, 0.0)
if self.pruning:
self._set_displayed_root(prune(tree, level=self.max_depth))
else:
self._set_displayed_root(tree)
else:
self.linkmatrix = None
self.root = None
self._set_displayed_root(None)
self._apply_selection()
def _update_labels(self):
labels = []
if self.root and self._displayed_root:
indices = [leaf.value.index for leaf in leaves(self.root)]
if self.annotation_idx == 0:
labels = []
elif self.annotation_idx == 1:
labels = [str(i) for i in indices]
elif isinstance(self.items, Orange.data.Table):
var = self.label_cb.model()[self.annotation_idx]
col = self.items[:, var]
labels = [var.repr_val(next(iter(row))) for row in col]
labels = [labels[idx] for idx in indices]
else:
labels = []
if labels and self._displayed_root is not self.root:
joined = leaves(self._displayed_root)
labels = [
", ".join(labels[leaf.value.first:leaf.value.last])
for leaf in joined
]
self.labels.set_labels(labels)
self.labels.setMinimumWidth(1 if labels else -1)
def _invalidate_clustering(self):
self._update()
self._update_labels()
def _invalidate_output(self):
self._invalidated = True
if self.autocommit:
self.commit()
def _invalidate_pruning(self):
if self.root:
selection = self.dendrogram.selected_nodes()
ranges = [node.value.range for node in selection]
if self.pruning:
self._set_displayed_root(prune(self.root,
level=self.max_depth))
else:
self._set_displayed_root(self.root)
selected = [
node for node in preorder(self._displayed_root)
if node.value.range in ranges
]
self.dendrogram.set_selected_clusters(selected)
self._apply_selection()
def commit(self):
self._invalidated = False
items = getattr(self.matrix, "items", self.items)
if not items:
# nothing to commit
return
selection = self.dendrogram.selected_nodes()
selection = sorted(selection, key=lambda c: c.value.first)
indices = [leaf.value.index for leaf in leaves(self.root)]
maps = [
indices[node.value.first:node.value.last] for node in selection
]
selected_indices = list(chain(*maps))
unselected_indices = sorted(
set(range(self.root.value.last)) - set(selected_indices))
selected = [items[k] for k in selected_indices]
unselected = [items[k] for k in unselected_indices]
if not selected:
self.send("Selected Data", None)
self.send("Other Data", None)
return
selected_data = unselected_data = None
if isinstance(items, Orange.data.Table):
c = numpy.zeros(len(items))
for i, indices in enumerate(maps):
c[indices] = i
c[unselected_indices] = len(maps)
mask = c != len(maps)
if self.append_clusters:
clust_var = Orange.data.DiscreteVariable(
str(self.cluster_name),
values=[
"Cluster {}".format(i + 1) for i in range(len(maps))
] + ["Other"])
data, domain = items, items.domain
attrs = domain.attributes
class_ = domain.class_vars
metas = domain.metas
X, Y, M = data.X, data.Y, data.metas
if self.cluster_role == self.AttributeRole:
attrs = attrs + (clust_var, )
X = numpy.c_[X, c]
elif self.cluster_role == self.ClassRole:
class_ = class_ + (clust_var, )
Y = numpy.c_[Y, c]
elif self.cluster_role == self.MetaRole:
metas = metas + (clust_var, )
M = numpy.c_[M, c]
domain = Orange.data.Domain(attrs, class_, metas)
data = Orange.data.Table(domain, X, Y, M)
else:
data = items
if selected:
selected_data = data[mask]
if unselected:
unselected_data = data[~mask]
self.send("Selected Data", selected_data)
self.send("Other Data", unselected_data)
def sizeHint(self):
return QSize(800, 500)
def eventFilter(self, obj, event):
if obj is self.view.viewport() and event.type() == QEvent.Resize:
width = self.view.viewport().width() - 2
self._main_graphics.setMaximumWidth(width)
self._main_graphics.setMinimumWidth(width)
self._main_graphics.layout().activate()
elif event.type() == QEvent.MouseButtonPress and \
(obj is self.top_axis_view.viewport() or
obj is self.bottom_axis_view.viewport()):
self.selection_method = 1
# Map click point to cut line local coordinates
pos = self.top_axis_view.mapToScene(event.pos())
cut = self.top_axis.line.mapFromScene(pos)
self.top_axis.line.setValue(cut.x())
# update the line visibility, output, ...
self._selection_method_changed()
return super().eventFilter(obj, event)
def _dendrogram_geom_changed(self):
pos = self.dendrogram.pos_at_height(self.cutoff_height)
geom = self.dendrogram.geometry()
crect = self.dendrogram.contentsRect()
self._set_slider_value(pos.x(), geom.width())
self.cut_line.setLength(geom.height())
self.top_axis.resize(crect.width(), self.top_axis.height())
self.top_axis.setPos(geom.left() + crect.left(), 0)
self.top_axis.line.setPos(self.cut_line.scenePos().x(), 0)
self.bottom_axis.resize(crect.width(), self.bottom_axis.height())
self.bottom_axis.setPos(geom.left() + crect.left(), 0)
self.bottom_axis.line.setPos(self.cut_line.scenePos().x(), 0)
geom = self._main_graphics.geometry()
assert geom.topLeft() == QPointF(0, 0)
self.scene.setSceneRect(geom)
geom.setHeight(self.top_axis.size().height())
self.top_axis.scene().setSceneRect(geom)
self.bottom_axis.scene().setSceneRect(geom)
def _axis_slider_changed(self, value):
self.cut_line.setValue(value)
def _dendrogram_slider_changed(self, value):
p = QPointF(value, 0)
cl_height = self.dendrogram.height_at(p)
self.set_cutoff_height(cl_height)
# Sync the cut positions between the dendrogram and the axis.
self._set_slider_value(value, self.dendrogram.size().width())
def _set_slider_value(self, value, span):
with blocked(self.cut_line):
self.cut_line.setValue(value)
self.cut_line.setRange(0, span)
with blocked(self.top_axis.line):
self.top_axis.line.setValue(value)
self.top_axis.line.setRange(0, span)
with blocked(self.bottom_axis.line):
self.bottom_axis.line.setValue(value)
self.bottom_axis.line.setRange(0, span)
def set_cutoff_height(self, height):
self.cutoff_height = height
if self.root:
self.cut_ratio = 100 * height / self.root.value.height
self.select_max_height(height)
def _set_cut_line_visible(self, visible):
self.cut_line.setVisible(visible)
self.top_axis.line.setVisible(visible)
self.bottom_axis.line.setVisible(visible)
def select_top_n(self, n):
root = self._displayed_root
if root:
clusters = top_clusters(root, n)
self.dendrogram.set_selected_clusters(clusters)
def select_max_height(self, height):
root = self._displayed_root
if root:
clusters = clusters_at_height(root, height)
self.dendrogram.set_selected_clusters(clusters)
def _selection_method_changed(self):
self._set_cut_line_visible(self.selection_method == 1)
if self.root:
self._apply_selection()
def _apply_selection(self):
if not self.root:
return
if self.selection_method == 0:
pass
elif self.selection_method == 1:
height = self.cut_ratio * self.root.value.height / 100
self.set_cutoff_height(height)
pos = self.dendrogram.pos_at_height(height)
self._set_slider_value(pos.x(), self.dendrogram.size().width())
elif self.selection_method == 2:
self.select_top_n(self.top_n)
def _selection_edited(self):
# Selection was edited by clicking on a cluster in the
# dendrogram view.
self.selection_method = 0
self._selection_method_changed()
class OWHierarchicalClustering(widget.OWWidget):
name = "Hierarchical Clustering"
description = ("Hierarchical clustering based on distance matrix, and "
"a dendrogram viewer.")
icon = "icons/HierarchicalClustering.svg"
priority = 2100
inputs = [("Distances", Orange.misc.DistMatrix, "set_distances")]
outputs = [("Selected Data", Orange.data.Table),
("Other Data", Orange.data.Table)]
#: Selected linkage
linkage = settings.Setting(1)
#: Index of the selected annotation item (variable, ...)
annotation_idx = settings.Setting(0)
#: Selected tree pruning (none/max depth)
pruning = settings.Setting(0)
#: Maximum depth when max depth pruning is selected
max_depth = settings.Setting(10)
#: Selected cluster selection method (none, cut distance, top n)
selection_method = settings.Setting(0)
#: Cut height ratio wrt root height
cut_ratio = settings.Setting(75.0)
#: Number of top clusters to select
top_n = settings.Setting(3)
append_clusters = settings.Setting(True)
cluster_role = settings.Setting(2)
cluster_name = settings.Setting("Cluster")
autocommit = settings.Setting(False)
#: Cluster variable domain role
AttributeRole, ClassRole, MetaRole = 0, 1, 2
def __init__(self, parent=None):
super().__init__(parent)
self.matrix = None
self.items = None
self.linkmatrix = None
self.root = None
self._displayed_root = None
self.cutoff_height = 0.0
gui.comboBox(gui.widgetBox(self.controlArea, "Linkage"),
self, "linkage", items=LINKAGE,
callback=self._invalidate_clustering)
box = gui.widgetBox(self.controlArea, "Annotation")
self.label_cb = gui.comboBox(
box, self, "annotation_idx", callback=self._update_labels)
self.label_cb.setModel(itemmodels.VariableListModel())
self.label_cb.model()[:] = ["None", "Enumeration"]
box = gui.radioButtons(
self.controlArea, self, "pruning", box="Pruning",
callback=self._invalidate_pruning
)
grid = QGridLayout()
box.layout().addLayout(grid)
grid.addWidget(
gui.appendRadioButton(box, "None", addToLayout=False),
0, 0
)
self.max_depth_spin = gui.spin(
box, self, "max_depth", minv=1, maxv=100,
callback=self._invalidate_pruning,
keyboardTracking=False
)
grid.addWidget(
gui.appendRadioButton(box, "Max depth", addToLayout=False),
1, 0)
grid.addWidget(self.max_depth_spin, 1, 1)
box = gui.radioButtons(
self.controlArea, self, "selection_method",
box="Selection",
callback=self._selection_method_changed)
grid = QGridLayout()
box.layout().addLayout(grid)
grid.addWidget(
gui.appendRadioButton(box, "Manual", addToLayout=False),
0, 0
)
grid.addWidget(
gui.appendRadioButton(box, "Height ratio", addToLayout=False),
1, 0
)
self.cut_ratio_spin = gui.spin(
box, self, "cut_ratio", 0, 100, step=1e-1, spinType=float,
callback=self._selection_method_changed
)
self.cut_ratio_spin.setSuffix("%")
grid.addWidget(self.cut_ratio_spin, 1, 1)
grid.addWidget(
gui.appendRadioButton(box, "Top N", addToLayout=False),
2, 0
)
self.top_n_spin = gui.spin(box, self, "top_n", 1, 20,
callback=self._selection_method_changed)
grid.addWidget(self.top_n_spin, 2, 1)
box.layout().addLayout(grid)
self.controlArea.layout().addStretch()
box = gui.widgetBox(self.controlArea, "Output")
gui.checkBox(box, self, "append_clusters", "Append cluster IDs",
callback=self._invalidate_output)
ibox = gui.indentedBox(box)
name_edit = gui.lineEdit(ibox, self, "cluster_name")
name_edit.editingFinished.connect(self._invalidate_output)
cb = gui.comboBox(
ibox, self, "cluster_role", callback=self._invalidate_output,
items=["Attribute",
"Class variable",
"Meta variable"]
)
form = QFormLayout(
fieldGrowthPolicy=QFormLayout.AllNonFixedFieldsGrow,
labelAlignment=Qt.AlignLeft,
spacing=8
)
form.addRow("Name", name_edit)
form.addRow("Place", cb)
ibox.layout().addSpacing(5)
ibox.layout().addLayout(form)
ibox.layout().addSpacing(5)
gui.auto_commit(box, self, "autocommit", "Send data", "Auto send is on",
box=False)
self.scene = QGraphicsScene()
self.view = QGraphicsView(
self.scene,
horizontalScrollBarPolicy=Qt.ScrollBarAlwaysOff,
verticalScrollBarPolicy=Qt.ScrollBarAlwaysOn,
alignment=Qt.AlignLeft | Qt.AlignVCenter
)
def axis_view(orientation):
ax = pg.AxisItem(orientation=orientation, maxTickLength=7)
scene = QGraphicsScene()
scene.addItem(ax)
view = QGraphicsView(
scene,
horizontalScrollBarPolicy=Qt.ScrollBarAlwaysOff,
verticalScrollBarPolicy=Qt.ScrollBarAlwaysOn,
alignment=Qt.AlignLeft | Qt.AlignVCenter
)
view.setFixedHeight(ax.size().height())
ax.line = SliderLine(orientation=Qt.Horizontal,
length=ax.size().height())
scene.addItem(ax.line)
return view, ax
self.top_axis_view, self.top_axis = axis_view("top")
self.mainArea.layout().setSpacing(1)
self.mainArea.layout().addWidget(self.top_axis_view)
self.mainArea.layout().addWidget(self.view)
self.bottom_axis_view, self.bottom_axis = axis_view("bottom")
self.mainArea.layout().addWidget(self.bottom_axis_view)
self._main_graphics = QGraphicsWidget()
self._main_layout = QGraphicsLinearLayout(Qt.Horizontal)
self._main_layout.setSpacing(1)
self._main_graphics.setLayout(self._main_layout)
self.scene.addItem(self._main_graphics)
self.dendrogram = DendrogramWidget()
self.dendrogram.setSizePolicy(QSizePolicy.MinimumExpanding,
QSizePolicy.MinimumExpanding)
self.dendrogram.selectionChanged.connect(self._invalidate_output)
self.dendrogram.selectionEdited.connect(self._selection_edited)
fm = self.fontMetrics()
self.dendrogram.setContentsMargins(
5, fm.lineSpacing() / 2,
5, fm.lineSpacing() / 2
)
self.labels = GraphicsSimpleTextList()
self.labels.setSizePolicy(QSizePolicy.Preferred, QSizePolicy.Preferred)
self.labels.setAlignment(Qt.AlignLeft)
self.labels.setMaximumWidth(200)
self.labels.layout().setSpacing(0)
self._main_layout.addItem(self.dendrogram)
self._main_layout.addItem(self.labels)
self._main_layout.setAlignment(
self.dendrogram, Qt.AlignLeft | Qt.AlignVCenter)
self._main_layout.setAlignment(
self.labels, Qt.AlignLeft | Qt.AlignVCenter)
self.view.viewport().installEventFilter(self)
self.top_axis_view.viewport().installEventFilter(self)
self.bottom_axis_view.viewport().installEventFilter(self)
self._main_graphics.installEventFilter(self)
self.cut_line = SliderLine(self.dendrogram,
orientation=Qt.Horizontal)
self.cut_line.valueChanged.connect(self._dendrogram_slider_changed)
self.cut_line.hide()
self.bottom_axis.line.valueChanged.connect(self._axis_slider_changed)
self.top_axis.line.valueChanged.connect(self._axis_slider_changed)
self.dendrogram.geometryChanged.connect(self._dendrogram_geom_changed)
self._set_cut_line_visible(self.selection_method == 1)
def set_distances(self, matrix):
self.matrix = matrix
self._invalidate_clustering()
self._set_items(matrix.row_items if matrix is not None else None)
def _set_items(self, items):
self.items = items
if items is None:
self.label_cb.model()[:] = ["None", "Enumeration"]
elif isinstance(items, Orange.data.Table):
vars = list(items.domain)
self.label_cb.model()[:] = ["None", "Enumeration"] + vars
elif isinstance(items, list) and \
all(isinstance(var, Orange.data.Variable) for var in items):
self.label_cb.model()[:] = ["None", "Enumeration", "Name"]
else:
self.label_cb.model()[:] = ["None", "Enumeration"]
self.annotation_idx = min(self.annotation_idx,
len(self.label_cb.model()) - 1)
def handleNewSignals(self):
self._update_labels()
def _clear_plot(self):
self.labels.set_labels([])
self.dendrogram.set_root(None)
def _set_displayed_root(self, root):
self._clear_plot()
self._displayed_root = root
self.dendrogram.set_root(root)
self._update_labels()
self._main_graphics.resize(
self._main_graphics.size().width(),
self._main_graphics.sizeHint(Qt.PreferredSize).height()
)
self._main_graphics.layout().activate()
def _update(self):
self._clear_plot()
distances = self.matrix
if distances is not None:
# Convert to flat upper triangular distances
i, j = numpy.triu_indices(distances.X.shape[0], k=1)
distances = distances.X[i, j]
method = LINKAGE[self.linkage].lower()
Z = scipy.cluster.hierarchy.linkage(
distances, method=method
)
tree = tree_from_linkage(Z)
self.linkmatrix = Z
self.root = tree
self.top_axis.setRange(tree.value.height, 0.0)
self.bottom_axis.setRange(tree.value.height, 0.0)
if self.pruning:
self._set_displayed_root(prune(tree, level=self.max_depth))
else:
self._set_displayed_root(tree)
else:
self.linkmatrix = None
self.root = None
self._set_displayed_root(None)
self._apply_selection()
def _update_labels(self):
labels = []
if self.root and self._displayed_root:
indices = [leaf.value.index for leaf in leaves(self.root)]
if self.annotation_idx == 0:
labels = []
elif self.annotation_idx == 1:
labels = [str(i) for i in indices]
elif isinstance(self.items, Orange.data.Table):
var = self.label_cb.model()[self.annotation_idx]
col = self.items[:, var]
labels = [var.repr_val(next(iter(row))) for row in col]
labels = [labels[idx] for idx in indices]
else:
labels = []
if labels and self._displayed_root is not self.root:
joined = leaves(self._displayed_root)
labels = [", ".join(labels[leaf.value.first: leaf.value.last])
for leaf in joined]
self.labels.set_labels(labels)
self.labels.setMinimumWidth(1 if labels else -1)
def _invalidate_clustering(self):
self._update()
self._update_labels()
def _invalidate_output(self):
self.commit()
def _invalidate_pruning(self):
if self.root:
selection = self.dendrogram.selected_nodes()
ranges = [node.value.range for node in selection]
if self.pruning:
self._set_displayed_root(
prune(self.root, level=self.max_depth))
else:
self._set_displayed_root(self.root)
selected = [node for node in preorder(self._displayed_root)
if node.value.range in ranges]
self.dendrogram.set_selected_clusters(selected)
self._apply_selection()
def commit(self):
items = getattr(self.matrix, "items", self.items)
if not items:
# nothing to commit
return
selection = self.dendrogram.selected_nodes()
selection = sorted(selection, key=lambda c: c.value.first)
indices = [leaf.value.index for leaf in leaves(self.root)]
maps = [indices[node.value.first:node.value.last]
for node in selection]
selected_indices = list(chain(*maps))
unselected_indices = sorted(set(range(self.root.value.last)) -
set(selected_indices))
selected = [items[k] for k in selected_indices]
unselected = [items[k] for k in unselected_indices]
if not selected:
self.send("Selected Data", None)
self.send("Other Data", None)
return
selected_data = unselected_data = None
if isinstance(items, Orange.data.Table):
c = numpy.zeros(len(items))
for i, indices in enumerate(maps):
c[indices] = i
c[unselected_indices] = len(maps)
mask = c != len(maps)
if self.append_clusters:
clust_var = Orange.data.DiscreteVariable(
str(self.cluster_name),
values=["Cluster {}".format(i + 1)
for i in range(len(maps))] +
["Other"], ordered=True
)
data, domain = items, items.domain
attrs = domain.attributes
class_ = domain.class_vars
metas = domain.metas
if self.cluster_role == self.AttributeRole:
attrs = attrs + (clust_var,)
elif self.cluster_role == self.ClassRole:
class_ = class_ + (clust_var,)
elif self.cluster_role == self.MetaRole:
metas = metas + (clust_var,)
domain = Orange.data.Domain(attrs, class_, metas)
data = Orange.data.Table(domain, data)
data.get_column_view(clust_var)[0][:] = c
else:
data = items
if selected:
selected_data = data[mask]
if unselected:
unselected_data = data[~mask]
self.send("Selected Data", selected_data)
self.send("Other Data", unselected_data)
def sizeHint(self):
return QSize(800, 500)
def eventFilter(self, obj, event):
if obj is self.view.viewport() and event.type() == QEvent.Resize:
width = self.view.viewport().width() - 2
self._main_graphics.setMaximumWidth(width)
self._main_graphics.setMinimumWidth(width)
self._main_graphics.layout().activate()
elif event.type() == QEvent.MouseButtonPress and \
(obj is self.top_axis_view.viewport() or
obj is self.bottom_axis_view.viewport()):
self.selection_method = 1
# Map click point to cut line local coordinates
pos = self.top_axis_view.mapToScene(event.pos())
cut = self.top_axis.line.mapFromScene(pos)
self.top_axis.line.setValue(cut.x())
# update the line visibility, output, ...
self._selection_method_changed()
return super().eventFilter(obj, event)
def _dendrogram_geom_changed(self):
pos = self.dendrogram.pos_at_height(self.cutoff_height)
geom = self.dendrogram.geometry()
crect = self.dendrogram.contentsRect()
self._set_slider_value(pos.x(), geom.width())
self.cut_line.setLength(geom.height())
self.top_axis.resize(crect.width(), self.top_axis.height())
self.top_axis.setPos(geom.left() + crect.left(), 0)
self.top_axis.line.setPos(self.cut_line.scenePos().x(), 0)
self.bottom_axis.resize(crect.width(), self.bottom_axis.height())
self.bottom_axis.setPos(geom.left() + crect.left(), 0)
self.bottom_axis.line.setPos(self.cut_line.scenePos().x(), 0)
geom = self._main_graphics.geometry()
assert geom.topLeft() == QPointF(0, 0)
self.scene.setSceneRect(geom)
geom.setHeight(self.top_axis.size().height())
self.top_axis.scene().setSceneRect(geom)
self.bottom_axis.scene().setSceneRect(geom)
def _axis_slider_changed(self, value):
self.cut_line.setValue(value)
def _dendrogram_slider_changed(self, value):
p = QPointF(value, 0)
cl_height = self.dendrogram.height_at(p)
self.set_cutoff_height(cl_height)
# Sync the cut positions between the dendrogram and the axis.
self._set_slider_value(value, self.dendrogram.size().width())
def _set_slider_value(self, value, span):
with blocked(self.cut_line):
self.cut_line.setValue(value)
self.cut_line.setRange(0, span)
with blocked(self.top_axis.line):
self.top_axis.line.setValue(value)
self.top_axis.line.setRange(0, span)
with blocked(self.bottom_axis.line):
self.bottom_axis.line.setValue(value)
self.bottom_axis.line.setRange(0, span)
def set_cutoff_height(self, height):
self.cutoff_height = height
if self.root:
self.cut_ratio = 100 * height / self.root.value.height
self.select_max_height(height)
def _set_cut_line_visible(self, visible):
self.cut_line.setVisible(visible)
self.top_axis.line.setVisible(visible)
self.bottom_axis.line.setVisible(visible)
def select_top_n(self, n):
root = self._displayed_root
if root:
clusters = top_clusters(root, n)
self.dendrogram.set_selected_clusters(clusters)
def select_max_height(self, height):
root = self._displayed_root
if root:
clusters = clusters_at_height(root, height)
self.dendrogram.set_selected_clusters(clusters)
def _selection_method_changed(self):
self._set_cut_line_visible(self.selection_method == 1)
if self.root:
self._apply_selection()
def _apply_selection(self):
if not self.root:
return
if self.selection_method == 0:
pass
elif self.selection_method == 1:
height = self.cut_ratio * self.root.value.height / 100
self.set_cutoff_height(height)
pos = self.dendrogram.pos_at_height(height)
self._set_slider_value(pos.x(), self.dendrogram.size().width())
elif self.selection_method == 2:
self.select_top_n(self.top_n)
def _selection_edited(self):
# Selection was edited by clicking on a cluster in the
# dendrogram view.
self.selection_method = 0
self._selection_method_changed()
