Merge tag '3.31.1' into stable

CHANGELOG.md

@@ -4,6 +4,15 @@ Change Log
 [next] - TBA
 ------------
 
+
+[3.31.1] - 2022-01-07
+--------------------
+##### Bugfixes
+* Group by: compute mode when all values in group nan ([#5763](../../pull/5763))
+* Support numpy 1.22 ([#5760](../../pull/5760))
+* Unpickling pre-3.28.0 Transformation ([#5759](../../pull/5759))
+
+
 [3.31.0] - 2021-12-17
 --------------------
 ##### Enhancements
@@ -1630,7 +1639,8 @@ Change Log
 * Initial version based on Python 1.5.2 and Qt 2.3
 
 
-[next]: https://github.com/biolab/orange3/compare/3.31.0...HEAD
+[next]: https://github.com/biolab/orange3/compare/3.31.1...HEAD
+[3.31.1]: https://github.com/biolab/orange3/compare/3.31.0...3.31.1
 [3.31.0]: https://github.com/biolab/orange3/compare/3.30.2...3.31.0
 [3.30.2]: https://github.com/biolab/orange3/compare/3.30.1...3.30.2
 [3.30.1]: https://github.com/biolab/orange3/compare/3.30.0...3.30.1

Orange/data/io_util.py

@@ -109,7 +109,7 @@ def isnastr(arr, out=None):
     arr = np.asarray(arr)
     if out is None and arr.shape != ():
         out = np.empty_like(arr, dtype=bool)
-    return __isnastr(arr, out=out)
+    return __isnastr(arr, out=out, casting="unsafe")
 
 
 def guess_data_type(orig_values, namask=None):
@@ -182,7 +182,7 @@ def get_number_of_decimals(values):
 
         def mapvalues(arr):
             arr = np.asarray(arr, dtype=object)
-            return mapvalues_(arr, out=np.empty_like(arr, dtype=float))
+            return mapvalues_(arr, out=np.empty_like(arr, dtype=float), casting="unsafe")
 
         values = mapvalues(orig_values)
 

Orange/data/tests/test_table.py

@@ -350,11 +350,7 @@ def test_locking_flag(self):
     def test_unpickled_empty_weights(self):
         # ensure that unpickled empty arrays could be unlocked
         self.assertEqual(0, self.table.W.size)
-        # flags.owndata asserts touch numpy internals
-        # feel free to remove when they do not pass
-        self.assertTrue(self.table.W.flags.owndata)
         unpickled = pickle.loads(pickle.dumps(self.table))
-        self.assertFalse(unpickled.W.flags.owndata)
         with unpickled.unlocked():
             pass
 

Orange/preprocess/transformation.py

@@ -16,12 +16,28 @@ def __init__(self, variable):
         :type variable: int or str or :obj:`~Orange.data.Variable`
         """
         self.variable = variable
+        self._create_cached_target_domain()
 
+    def _create_cached_target_domain(self):
+        """ If the same domain is used everytime this allows better caching of
+        domain transformations in from_table"""
         if self.variable is not None:
             if self.variable.is_primitive():
-                self.need_domain = Domain([self.variable])
+                self._target_domain = Domain([self.variable])
             else:
-                self.need_domain = Domain([], metas=[self.variable])
+                self._target_domain = Domain([], metas=[self.variable])
+
+    def __getstate__(self):
+        # Do not pickle the cached domain; rather recreate it after unpickling
+        state = self.__dict__.copy()
+        state.pop("_target_domain")
+        return state
+
+    def __setstate__(self, state):
+        # Ensure that cached target domain is created after unpickling.
+        # This solves the problem of unpickling old pickled models.
+        self.__dict__.update(state)
+        self._create_cached_target_domain()
 
     def __call__(self, data):
         """
@@ -31,7 +47,7 @@ def __call__(self, data):
         inst = isinstance(data, Instance)
         if inst:
             data = Table.from_list(data.domain, [data])
-        data = data.transform(self.need_domain)
+        data = data.transform(self._target_domain)
         if self.variable.is_primitive():
             col = data.X
         else:

Orange/tests/test_transformation.py

@@ -1,3 +1,4 @@
+import pickle
 import unittest
 
 import numpy as np
@@ -43,6 +44,17 @@ def test_transform_fails(self):
         trans = Transformation(self.data.domain[2])
         self.assertRaises(NotImplementedError, trans, self.data)
 
+    def test_pickling_target_domain(self):
+        data = self.data
+        trans = self.TransformationMock(data.domain[2])
+        self.assertIn("_target_domain", trans.__dict__)
+        # _target_domain should not be pickled
+        state = trans.__getstate__()
+        self.assertNotIn("_target_domain", state)
+        # _target_domain should be recreated when unpickled
+        unpickled = pickle.loads(pickle.dumps(trans))
+        self.assertIn("_target_domain", unpickled.__dict__)
+
 
 class IdentityTest(unittest.TestCase):
     def test_identity(self):

Orange/widgets/data/oweditdomain.py

@@ -2645,7 +2645,7 @@ def mapper(arr, out=None, dtype=dtype, **kwargs):
         arr = np.asanyarray(arr)
         if out is None and dtype is not None and arr.shape != ():
             out = np.empty_like(arr, dtype)
-        return _vmapper(arr, out, dtype=dtype, **kwargs)
+        return _vmapper(arr, out, dtype=dtype, casting="unsafe", **kwargs)
     return mapper
 
 
@@ -2684,7 +2684,7 @@ def as_float_or_nan(
             np.issubdtype(arr.dtype, np.integer):
         np.copyto(out, arr, casting="unsafe", where=where)
         return out
-    return _parse_float(arr, out, where=where, **kwargs)
+    return _parse_float(arr, out, where=where, casting="unsafe", **kwargs)
 
 
 def copy_attributes(dst: V, src: Orange.data.Variable) -> V:

Orange/widgets/data/owgroupby.py

@@ -5,6 +5,7 @@
 from typing import Any, Dict, List, Optional, Set
 
 import pandas as pd
+from numpy import nan
 from AnyQt.QtCore import (
     QAbstractTableModel,
     QEvent,
@@ -58,7 +59,7 @@ def concatenate(x):
     "Mean": Aggregation("mean", {ContinuousVariable, TimeVariable}),
     "Median": Aggregation("median", {ContinuousVariable, TimeVariable}),
     "Mode": Aggregation(
-        lambda x: pd.Series.mode(x)[0], {ContinuousVariable, TimeVariable}
+        lambda x: pd.Series.mode(x).get(0, nan), {ContinuousVariable, TimeVariable}
     ),
     "Standard deviation": Aggregation("std", {ContinuousVariable, TimeVariable}),
     "Variance": Aggregation("var", {ContinuousVariable, TimeVariable}),
@@ -404,7 +405,7 @@ def __gb_changed(self) -> None:
         self.gb_attrs = [values[row.row()] for row in sorted(rows)]
         # everything cached in result should be recomputed on gb change
         self.result = Result()
-        self.commit()
+        self.commit.deferred()
 
     def __aggregation_changed(self, agg: str) -> None:
         """
@@ -420,7 +421,7 @@ def __aggregation_changed(self, agg: str) -> None:
             else:
                 self.aggregations[attr].discard(agg)
             self.agg_table_model.update_aggregation(attr)
-        self.commit()
+        self.commit.deferred()
 
     @Inputs.data
     def set_data(self, data: Table) -> None:
@@ -448,11 +449,12 @@ def set_data(self, data: Table) -> None:
         self.agg_table_model.set_domain(data.domain if data else None)
         self._set_gb_selection()
 
-        self.commit()
+        self.commit.now()
 
     #########################
     # Task connected methods
 
+    @gui.deferred
     def commit(self) -> None:
         self.Error.clear()
         self.Warning.clear()

Orange/widgets/data/tests/test_owgroupby.py

@@ -12,6 +12,8 @@
 from Orange.data import (
     Table,
     table_to_frame,
+    Domain,
+    ContinuousVariable,
 )
 from Orange.data.tests.test_aggregate import create_sample_data
 from Orange.widgets.data.owgroupby import OWGroupBy
@@ -689,6 +691,65 @@ def test_time_variable(self):
         output = self.get_output(self.widget.Outputs.data)
         self.assertEqual(2, len(output))
 
+    def test_only_nan_in_group(self):
+        data = Table(
+            Domain([ContinuousVariable("A"), ContinuousVariable("B")]),
+            np.array([[1, np.nan], [2, 1], [1, np.nan], [2, 1]]),
+        )
+        self.send_signal(self.widget.Inputs.data, data)
+
+        # select feature A as group-by
+        self._set_selection(self.widget.gb_attrs_view, [0])
+        # select all aggregations for feature B
+        self.select_table_rows(self.widget.agg_table_view, [1])
+        for cb in self.widget.agg_checkboxes.values():
+            while not cb.isChecked():
+                cb.click()
+
+        # unselect all aggregations for attr A
+        self.select_table_rows(self.widget.agg_table_view, [0])
+        for cb in self.widget.agg_checkboxes.values():
+            while cb.isChecked():
+                cb.click()
+
+        expected_columns = [
+            "B - Mean",
+            "B - Median",
+            "B - Mode",
+            "B - Standard deviation",
+            "B - Variance",
+            "B - Sum",
+            "B - Min. value",
+            "B - Max. value",
+            "B - Span",
+            "B - First value",
+            "B - Last value",
+            "B - Random value",
+            "B - Count defined",
+            "B - Count",
+            "B - Proportion defined",
+            "B - Concatenate",
+            "A",
+        ]
+        n = np.nan
+        expected_df = pd.DataFrame(
+            [
+                [n, n, n, n, n, 0, n, n, n, n, n, n, 0, 2, 0, "", 1],
+                [1, 1, 1, 0, 0, 2, 1, 1, 0, 1, 1, 1, 2, 2, 1, "1.0 1.0", 2],
+            ],
+            columns=expected_columns,
+        )
+        output_df = table_to_frame(
+            self.get_output(self.widget.Outputs.data), include_metas=True
+        )
+        pd.testing.assert_frame_equal(
+            output_df,
+            expected_df,
+            check_dtype=False,
+            check_column_type=False,
+            check_categorical=False,
+        )
+
 
 if __name__ == "__main__":
     unittest.main()

doc/visual-programming/source/index.rst

@@ -63,6 +63,7 @@ Data
    widgets/data/melt
    widgets/data/neighbors
    widgets/data/unique
+   widgets/data/groupby
 
 
 Visualize

doc/visual-programming/source/widgets/data/groupby.md

@@ -36,4 +36,4 @@ In the **Data Table** widget, we can see that both females and males have lower
 You can also observe differences between median **cholesterol** level and mean value of **major vessel colored** between groups.
 
 
-![](images/Group-by-Example.png)
+![](images/Group-by-example.png)

setup.py

@@ -44,7 +44,7 @@
 
 NAME = 'Orange3'
 
-VERSION = '3.31.0'
+VERSION = '3.31.1'
 ISRELEASED = True
 # full version identifier including a git revision identifier for development
 # build/releases (this is filled/updated in `write_version_py`)