[ENH] OWSelectByDataIndex: New widget (move from Prototypes)

Orange/widgets/data/owselectbydataindex.py

@@ -0,0 +1,106 @@
+import numpy as np
+
+from AnyQt.QtWidgets import QApplication
+
+from Orange.data import Table
+from Orange.widgets import widget, gui
+from Orange.widgets.utils import itemmodels
+from Orange.widgets.utils.sql import check_sql_input
+from Orange.widgets.widget import Input, Output
+
+
+class OWSelectByDataIndex(widget.OWWidget):
+    name = "Select by Data Index"
+    description = "Match instances by index from data subset."
+    icon = "icons/SelectByDataIndex.svg"
+    priority = 1112
+
+    class Inputs:
+        data = Input("Data", Table)
+        data_subset = Input("Data Subset", Table)
+
+    class Outputs:
+        data = Output("Data", Table)
+
+    want_main_area = False
+    resizing_enabled = False
+
+    class Warning(widget.OWWidget.Warning):
+        instances_not_matching = widget.Msg("Input tables do not share any instances.")
+
+    def __init__(self):
+        super().__init__()
+
+        self.data = None
+        self.data_subset = None
+
+        self.model = itemmodels.VariableListModel()
+        self.model_unique_with_id = itemmodels.VariableListModel()
+        self.extra_model_unique = itemmodels.VariableListModel()
+        self.extra_model_unique_with_id = itemmodels.VariableListModel()
+
+        box = gui.hBox(self.controlArea, box=None)
+        self.infoBoxData = gui.label(
+            box, self, self.data_info_text(None), box="Data")
+        self.infoBoxExtraData = gui.label(
+            box, self, self.data_info_text(None), box="Data Subset")
+
+    @Inputs.data
+    @check_sql_input
+    def set_data(self, data):
+        self.data = data
+        self.infoBoxData.setText(self.data_info_text(data))
+
+    @Inputs.data_subset
+    @check_sql_input
+    def set_data_subset(self, data):
+        self.data_subset = data
+        self.infoBoxExtraData.setText(self.data_info_text(data))
+
+    def handleNewSignals(self):
+        self._invalidate()
+
+    def data_info_text(self, data):
+        if data is None:
+            return "No data."
+        else:
+            return "{}\n{} instances\n{} variables".format(
+                data.name, len(data), len(data.domain) + len(data.domain.metas))
+
+    def commit(self):
+        self.Warning.instances_not_matching.clear()
+        subset_ids = []
+        if self.data_subset:
+            subset_ids = self.data_subset.ids
+        if not self.data:
+            output_data = None
+        else:
+            if self.data_subset and len(np.intersect1d(subset_ids, self.data.ids)) == 0:
+                self.Warning.instances_not_matching()
+            subset_indices = np.in1d(self.data.ids, subset_ids)
+            output_data = self.data[subset_indices]
+        self.Outputs.data.send(output_data)
+
+    def _invalidate(self):
+        self.commit()
+
+    def send_report(self):
+        d_text = self.data_info_text(self.data).replace("\n", ", ")
+        ds_text = self.data_info_text(self.data_subset).replace("\n", ", ")
+        self.report_items("", [("Data", d_text), ("Data Subset", ds_text)])
+
+
+def main():
+    app = QApplication([])
+    w = OWSelectByDataIndex()
+    data = Table("iris.tab")
+    data_subset = data[:20]
+    w.set_data(data)
+    w.set_data_subset(data_subset)
+    w.handleNewSignals()
+    w.show()
+    app.exec_()
+
+
+if __name__ == "__main__":
+    main()

Orange/widgets/data/tests/test_owselectbydataindex.py

@@ -0,0 +1,26 @@
+from Orange.data import Table, Domain
+from Orange.widgets.data.owselectbydataindex import OWSelectByDataIndex
+from Orange.widgets.tests.base import WidgetTest
+
+
+class TestOWSelectSubset(WidgetTest):
+
+    def setUp(self):
+        self.widget = self.create_widget(OWSelectByDataIndex)
+
+    def test_subset(self):
+        data = Table("iris")
+        data_subset = data[20:40].transform(Domain([]))  # destroy domain
+        self.send_signal(self.widget.Inputs.data, data)
+        self.send_signal(self.widget.Inputs.data_subset, data_subset)
+        out = self.get_output(self.widget.Outputs.data)
+        self.assertEqual(list(data[20:40]), list(out))
+
+    def test_subset_nosubset(self):
+        data = Table("iris")
+        data_subset = Table("titanic")
+        self.send_signal(self.widget.Inputs.data, data)
+        self.send_signal(self.widget.Inputs.data_subset, data_subset)
+        out = self.get_output(self.widget.Outputs.data)
+        self.assertTrue(self.widget.Warning.instances_not_matching.is_shown())
+        self.assertEqual([], list(out))

doc/visual-programming/source/index.rst

@@ -37,6 +37,7 @@ Data
    widgets/data/createclass
    widgets/data/randomize
    widgets/data/concatenate
+   widgets/data/select-by-data-index
    widgets/data/paintdata
    widgets/data/pythonscript
    widgets/data/featureconstructor

doc/visual-programming/source/widgets/data/select-by-data-index.rst

@@ -0,0 +1,35 @@
+Select by Data Index
+====================
+
+Match instances by index from data subset.
+
+Inputs
+    Data
+        reference data set
+    Data Subset
+        subset to match
+
+Outputs
+    Data
+        subset from reference data set that matches indices from subset data
+
+
+**Select by Data Index** enables matching the data by indices. Each row in a data set has an index and given a subset, this widget can match these indices to indices from the reference data. Most often it is used to retrieve the original data from the transformed data (say, from PCA space).
+
+.. figure:: images/Select-by-data-index-stamped.png
+   :scale: 50%
+
+1. Information on the reference data set. This data is used as index reference.
+2. Information on the data subset. The indices of this data set are used to find matching data in the reference data set. Matching data are on the output by default.
+
+Example
+-------
+
+A typical use of **Select by Data Index** is to retrieve the original data after a transformation. We will load *iris.tab* data in the **File** widget. Then we will transform this data with **PCA**. We can project the transformed data in a **Scatter Plot**, where we can only see PCA components and not the original features.
+
+Now we will select an interesting subset (we could also select the entire data set). If we observe it in a **Data Table**, we can see that the data is transformed. If we would like to see this data with the original features, we will have to retrieve them with **Select by Data Index**.
+
+Connect the original data and the subset from **Scatter Plot** to **Select by Data Index**. The widget will match the indices of the subset with the indices of the reference (original) data and output the matching reference data. A final inspection in another **Data Table** confirms the data on the output is from the original data space.
+
+.. figure:: images/Select-by-Data-Index-Example1.png
+