From b70639c9483113c5de0139a1c157ebb30502795b Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Sun, 13 Sep 2020 16:54:13 +0100
Subject: [PATCH 01/26] Record output of transformers in ColumnTransformer

---
 sklearn/compose/_column_transformer.py        | 19 +++++++++++++++++
 .../compose/tests/test_column_transformer.py  | 21 +++++++++++++++----
 2 files changed, 36 insertions(+), 4 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 780ce46433b90..fcb61be5af792 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -136,6 +136,10 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         sparse matrix or a dense numpy array, which depends on the output
         of the individual transformers and the `sparse_threshold` keyword.
 
+    transformers_output_ : dict
+        A dictionary from transformer names to column indices, recording the
+        columns produced by each transformer (if any).
+
     Notes
     -----
     The order of the columns in the transformed feature matrix follows the
@@ -417,6 +421,20 @@ def _validate_output(self, result):
                     "The output of the '{0}' transformer should be 2D (scipy "
                     "matrix, array, or pandas DataFrame).".format(name))
 
+    def _index_output(self, Xs):
+        """
+        Record which transformer produced which column.
+        """
+        idx_ = 0
+        self.transformers_output_ = {}
+        for idx, (name, trans, _, _) in enumerate(
+            self._iter(fitted=True, replace_strings=True)
+        ):
+            n_columns = Xs[idx].shape[1]
+            self.transformers_output_[name] = list(range(idx_, idx_ + n_columns))
+            idx_ += n_columns
+        return
+
     def _validate_features(self, n_features, feature_names):
         """Ensures feature counts and names are the same during fit and
         transform.
@@ -555,6 +573,7 @@ def fit_transform(self, X, y=None):
 
         self._update_fitted_transformers(transformers)
         self._validate_output(Xs)
+        self._index_output(Xs)
 
         return self._hstack(list(Xs))
 
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 4e58769e244c7..b4ae02123913f 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -115,6 +115,7 @@ def test_column_transformer():
     assert_array_equal(ct.fit_transform(X_array), X_res_both)
     assert_array_equal(ct.fit(X_array).transform(X_array), X_res_both)
     assert len(ct.transformers_) == 2
+    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test with transformer_weights
     transformer_weights = {'trans1': .1, 'trans2': 10}
@@ -126,12 +127,14 @@ def test_column_transformer():
     assert_array_equal(both.fit_transform(X_array), res)
     assert_array_equal(both.fit(X_array).transform(X_array), res)
     assert len(both.transformers_) == 2
+    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     both = ColumnTransformer([('trans', Trans(), [0, 1])],
                              transformer_weights={'trans': .1})
     assert_array_equal(both.fit_transform(X_array), 0.1 * X_res_both)
     assert_array_equal(both.fit(X_array).transform(X_array), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
+    assert both.transformers_output_ == {'trans': [0, 1]}
 
 
 def test_column_transformer_dataframe():
@@ -190,6 +193,7 @@ def test_column_transformer_dataframe():
     assert_array_equal(ct.fit(X_df).transform(X_df), X_res_both)
     assert len(ct.transformers_) == 2
     assert ct.transformers_[-1][0] != 'remainder'
+    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
@@ -197,6 +201,7 @@ def test_column_transformer_dataframe():
     assert_array_equal(ct.fit(X_df).transform(X_df), X_res_both)
     assert len(ct.transformers_) == 2
     assert ct.transformers_[-1][0] != 'remainder'
+    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test with transformer_weights
     transformer_weights = {'trans1': .1, 'trans2': 10}
@@ -208,7 +213,8 @@ def test_column_transformer_dataframe():
     assert_array_equal(both.fit_transform(X_df), res)
     assert_array_equal(both.fit(X_df).transform(X_df), res)
     assert len(both.transformers_) == 2
-    assert ct.transformers_[-1][0] != 'remainder'
+    assert both.transformers_[-1][0] != 'remainder'
+    assert both.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test multiple columns
     both = ColumnTransformer([('trans', Trans(), ['first', 'second'])],
@@ -216,16 +222,18 @@ def test_column_transformer_dataframe():
     assert_array_equal(both.fit_transform(X_df), 0.1 * X_res_both)
     assert_array_equal(both.fit(X_df).transform(X_df), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
-    assert ct.transformers_[-1][0] != 'remainder'
+    assert both.transformers_[-1][0] != 'remainder'
+    assert both.transformers_output_ == {'trans': [0, 1]}
 
     both = ColumnTransformer([('trans', Trans(), [0, 1])],
                              transformer_weights={'trans': .1})
     assert_array_equal(both.fit_transform(X_df), 0.1 * X_res_both)
     assert_array_equal(both.fit(X_df).transform(X_df), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
-    assert ct.transformers_[-1][0] != 'remainder'
+    assert both.transformers_[-1][0] != 'remainder'
+    assert both.transformers_output_ == {'trans': [0, 1]}
 
-    # ensure pandas object is passes through
+    # ensure pandas object is passed through
 
     class TransAssert(BaseEstimator):
 
@@ -286,6 +294,7 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2
     assert isinstance(ct.transformers_[1][1], TransRaise)
+    assert ct.transformers_output_ == {'trans1': [0, 1]}
 
     ct = ColumnTransformer([('trans1', TransRaise(), column),
                             ('trans2', Trans(), [0, 1])])
@@ -293,6 +302,7 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2
     assert isinstance(ct.transformers_[0][1], TransRaise)
+    assert ct.transformers_output_ == {'trans2': [0, 1]}
 
     ct = ColumnTransformer([('trans', TransRaise(), column)],
                            remainder='passthrough')
@@ -300,6 +310,7 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2  # including remainder
     assert isinstance(ct.transformers_[0][1], TransRaise)
+    assert ct.transformers_output_ == {'remainder': [0, 1]}
 
     fixture = np.array([[], [], []])
     ct = ColumnTransformer([('trans', TransRaise(), column)],
@@ -355,6 +366,8 @@ def test_column_transformer_list():
     assert_array_equal(ct.fit_transform(X_list), expected_result)
     assert_array_equal(ct.fit(X_list).transform(X_list), expected_result)
 
+    assert ct.transformers_output_ == {'numerical': [0, 1], 'categorical': [2, 3]}
+
 
 def test_column_transformer_sparse_stacking():
     X_array = np.array([[0, 1, 2], [2, 4, 6]]).T

From 80a9b14f5036efc61932c46ad370bd1aa7d3b3fc Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Sun, 13 Sep 2020 17:06:34 +0100
Subject: [PATCH 02/26] Shorten lines

---
 sklearn/compose/_column_transformer.py           | 4 +++-
 sklearn/compose/tests/test_column_transformer.py | 4 +++-
 2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index fcb61be5af792..1417341083e4d 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -431,7 +431,9 @@ def _index_output(self, Xs):
             self._iter(fitted=True, replace_strings=True)
         ):
             n_columns = Xs[idx].shape[1]
-            self.transformers_output_[name] = list(range(idx_, idx_ + n_columns))
+            self.transformers_output_[name] = list(
+                range(idx_, idx_ + n_columns)
+            )
             idx_ += n_columns
         return
 
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index b4ae02123913f..e3b76ba7fdb8d 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -366,7 +366,9 @@ def test_column_transformer_list():
     assert_array_equal(ct.fit_transform(X_list), expected_result)
     assert_array_equal(ct.fit(X_list).transform(X_list), expected_result)
 
-    assert ct.transformers_output_ == {'numerical': [0, 1], 'categorical': [2, 3]}
+    assert ct.transformers_output_ == {
+        'numerical': [0, 1], 'categorical': [2, 3]
+    }
 
 
 def test_column_transformer_sparse_stacking():

From 5ce664694cfcb5fdf6920b2ea4017009aef13416 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Sun, 13 Sep 2020 18:50:50 +0200
Subject: [PATCH 03/26] Control variable 'trans' not used within the loop

---
 sklearn/compose/_column_transformer.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 1417341083e4d..d35240d004224 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -427,7 +427,7 @@ def _index_output(self, Xs):
         """
         idx_ = 0
         self.transformers_output_ = {}
-        for idx, (name, trans, _, _) in enumerate(
+        for idx, (name, _, _, _) in enumerate(
             self._iter(fitted=True, replace_strings=True)
         ):
             n_columns = Xs[idx].shape[1]

From fbc374f21f43ef54fc5af18c88268df741ef5d22 Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Fri, 25 Sep 2020 08:38:52 +0100
Subject: [PATCH 04/26] Use slices as values instead of lists

---
 sklearn/compose/_column_transformer.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index d35240d004224..f88cd59da6bd9 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -431,9 +431,7 @@ def _index_output(self, Xs):
             self._iter(fitted=True, replace_strings=True)
         ):
             n_columns = Xs[idx].shape[1]
-            self.transformers_output_[name] = list(
-                range(idx_, idx_ + n_columns)
-            )
+            self.transformers_output_[name] = slice(idx_, idx_ + n_columns)
             idx_ += n_columns
         return
 

From 6e86e807d1fc808f52997fc22caed8c8d8207bdd Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Fri, 25 Sep 2020 08:39:41 +0100
Subject: [PATCH 05/26] Clarify docstring (hopefully)

---
 sklearn/compose/_column_transformer.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index f88cd59da6bd9..1a1342acf2ec6 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -137,8 +137,9 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         of the individual transformers and the `sparse_threshold` keyword.
 
     transformers_output_ : dict
-        A dictionary from transformer names to column indices, recording the
-        columns produced by each transformer (if any).
+        A dictionary from transformer names to slices to track the results
+        of each transformer. The slices are the positions in the stacked
+        results of the output of each particular step in `transformers`.
 
     Notes
     -----

From 9c789de248f27d5cbdefdeaac85aef376ef82708 Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Fri, 25 Sep 2020 08:41:09 +0100
Subject: [PATCH 06/26] Update and spin off tests

---
 .../compose/tests/test_column_transformer.py  | 59 ++++++++++++++-----
 1 file changed, 44 insertions(+), 15 deletions(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index e3b76ba7fdb8d..135abcdbceeec 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -115,7 +115,6 @@ def test_column_transformer():
     assert_array_equal(ct.fit_transform(X_array), X_res_both)
     assert_array_equal(ct.fit(X_array).transform(X_array), X_res_both)
     assert len(ct.transformers_) == 2
-    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test with transformer_weights
     transformer_weights = {'trans1': .1, 'trans2': 10}
@@ -127,14 +126,12 @@ def test_column_transformer():
     assert_array_equal(both.fit_transform(X_array), res)
     assert_array_equal(both.fit(X_array).transform(X_array), res)
     assert len(both.transformers_) == 2
-    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     both = ColumnTransformer([('trans', Trans(), [0, 1])],
                              transformer_weights={'trans': .1})
     assert_array_equal(both.fit_transform(X_array), 0.1 * X_res_both)
     assert_array_equal(both.fit(X_array).transform(X_array), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
-    assert both.transformers_output_ == {'trans': [0, 1]}
 
 
 def test_column_transformer_dataframe():
@@ -193,7 +190,6 @@ def test_column_transformer_dataframe():
     assert_array_equal(ct.fit(X_df).transform(X_df), X_res_both)
     assert len(ct.transformers_) == 2
     assert ct.transformers_[-1][0] != 'remainder'
-    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
@@ -201,7 +197,6 @@ def test_column_transformer_dataframe():
     assert_array_equal(ct.fit(X_df).transform(X_df), X_res_both)
     assert len(ct.transformers_) == 2
     assert ct.transformers_[-1][0] != 'remainder'
-    assert ct.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test with transformer_weights
     transformer_weights = {'trans1': .1, 'trans2': 10}
@@ -214,7 +209,6 @@ def test_column_transformer_dataframe():
     assert_array_equal(both.fit(X_df).transform(X_df), res)
     assert len(both.transformers_) == 2
     assert both.transformers_[-1][0] != 'remainder'
-    assert both.transformers_output_ == {'trans1': [0], 'trans2': [1]}
 
     # test multiple columns
     both = ColumnTransformer([('trans', Trans(), ['first', 'second'])],
@@ -223,7 +217,6 @@ def test_column_transformer_dataframe():
     assert_array_equal(both.fit(X_df).transform(X_df), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
     assert both.transformers_[-1][0] != 'remainder'
-    assert both.transformers_output_ == {'trans': [0, 1]}
 
     both = ColumnTransformer([('trans', Trans(), [0, 1])],
                              transformer_weights={'trans': .1})
@@ -231,7 +224,6 @@ def test_column_transformer_dataframe():
     assert_array_equal(both.fit(X_df).transform(X_df), 0.1 * X_res_both)
     assert len(both.transformers_) == 1
     assert both.transformers_[-1][0] != 'remainder'
-    assert both.transformers_output_ == {'trans': [0, 1]}
 
     # ensure pandas object is passed through
 
@@ -294,7 +286,6 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2
     assert isinstance(ct.transformers_[1][1], TransRaise)
-    assert ct.transformers_output_ == {'trans1': [0, 1]}
 
     ct = ColumnTransformer([('trans1', TransRaise(), column),
                             ('trans2', Trans(), [0, 1])])
@@ -302,7 +293,6 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2
     assert isinstance(ct.transformers_[0][1], TransRaise)
-    assert ct.transformers_output_ == {'trans2': [0, 1]}
 
     ct = ColumnTransformer([('trans', TransRaise(), column)],
                            remainder='passthrough')
@@ -310,7 +300,6 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert_array_equal(ct.fit(X).transform(X), X_res_both)
     assert len(ct.transformers_) == 2  # including remainder
     assert isinstance(ct.transformers_[0][1], TransRaise)
-    assert ct.transformers_output_ == {'remainder': [0, 1]}
 
     fixture = np.array([[], [], []])
     ct = ColumnTransformer([('trans', TransRaise(), column)],
@@ -321,6 +310,50 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert isinstance(ct.transformers_[0][1], TransRaise)
 
 
+def test_column_transformer_transformers_output():
+    X_array = np.array([[0, 1, 2], [2, 4, 6]]).T
+
+    ct = ColumnTransformer([('trans1', Trans(), [0]),
+                            ('trans2', Trans(), [1])])
+    ct.fit(X_array)
+    assert ct.transformers_output_ == {'trans1': slice(0, 1),
+                                       'trans2': slice(1, 2)}
+
+    # test with transformer_weights and multiple columns
+    ct = ColumnTransformer([('trans', Trans(), [0, 1])],
+                             transformer_weights={'trans': .1})
+    ct.fit(X_array)
+    assert ct.transformers_output_ == {'trans': slice(0, 2)}
+
+    # test case that ensures that the attribute does also work when
+    # a given transformer doesn't have any columns to work on
+    ct = ColumnTransformer([('trans1', Trans(), [0, 1]),
+                            ('trans2', TransRaise(), [])])
+    ct.fit(X_array)
+    assert ct.transformers_output_ == {'trans1': slice(0, 2)}
+
+    ct = ColumnTransformer([('trans', TransRaise(), [])],
+                           remainder='passthrough')
+    ct.fit(X_array)
+    assert ct.transformers_output_ == {'remainder': slice(0, 2)}
+
+    # test with data frame
+    pd = pytest.importorskip('pandas')
+    X_df = pd.DataFrame(X_array, columns=['first', 'second'])
+
+    ct = ColumnTransformer([('trans1', Trans(), ['first']),
+                            ('trans2', Trans(), ['second'])])
+    ct.fit(X_df)
+    assert ct.transformers_output_ == {'trans1': slice(0, 1),
+                                       'trans2': slice(1, 2)}
+
+    ct = ColumnTransformer([('trans1', Trans(), [0]),
+                            ('trans2', Trans(), [1])])
+    ct.fit(X_df)
+    assert ct.transformers_output_ == {'trans1': slice(0, 1),
+                                       'trans2': slice(1, 2)}
+
+
 def test_column_transformer_sparse_array():
     X_sparse = sparse.eye(3, 2).tocsr()
 
@@ -366,10 +399,6 @@ def test_column_transformer_list():
     assert_array_equal(ct.fit_transform(X_list), expected_result)
     assert_array_equal(ct.fit(X_list).transform(X_list), expected_result)
 
-    assert ct.transformers_output_ == {
-        'numerical': [0, 1], 'categorical': [2, 3]
-    }
-
 
 def test_column_transformer_sparse_stacking():
     X_array = np.array([[0, 1, 2], [2, 4, 6]]).T

From bcc35eeacbf5ea1f96b4745e26db86b5d759d98b Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Fri, 25 Sep 2020 08:45:28 +0100
Subject: [PATCH 07/26] Fix white space

---
 sklearn/compose/tests/test_column_transformer.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 135abcdbceeec..1deda43a0a784 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -321,7 +321,7 @@ def test_column_transformer_transformers_output():
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
-                             transformer_weights={'trans': .1})
+                           transformer_weights={'trans': .1})
     ct.fit(X_array)
     assert ct.transformers_output_ == {'trans': slice(0, 2)}
 

From fc3948f3d8ed0761be227ab7ed0c934244ec2e07 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:15:28 +0100
Subject: [PATCH 08/26] Update documentation

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
Co-Authored-By: Joel Nothman <78827+jnothman@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 1a1342acf2ec6..9d163cc3ed8d7 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -137,9 +137,10 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         of the individual transformers and the `sparse_threshold` keyword.
 
     transformers_output_ : dict
-        A dictionary from transformer names to slices to track the results
-        of each transformer. The slices are the positions in the stacked
-        results of the output of each particular step in `transformers`.
+        A dictionary from transformer names to :py:class:`slice`s, where
+        each slice corresponds to indices in the transformed output.
+        This is useful to inspect which transformer is responsible for
+        which transformed feature(s).
 
     Notes
     -----

From 4a4f81d18b7d8572fa16c14934bf44a9262fc03d Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:25:03 +0100
Subject: [PATCH 09/26] Rename idx_

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 9d163cc3ed8d7..34dff9d3c54b8 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -427,14 +427,14 @@ def _index_output(self, Xs):
         """
         Record which transformer produced which column.
         """
-        idx_ = 0
+        idx = 0
         self.transformers_output_ = {}
-        for idx, (name, _, _, _) in enumerate(
+        for transformer_idx, (name, _, _, _) in enumerate(
             self._iter(fitted=True, replace_strings=True)
         ):
-            n_columns = Xs[idx].shape[1]
-            self.transformers_output_[name] = slice(idx_, idx_ + n_columns)
-            idx_ += n_columns
+            n_columns = Xs[transformer_idx].shape[1]
+            self.transformers_output_[name] = slice(idx, idx + n_columns)
+            idx += n_columns
         return
 
     def _validate_features(self, n_features, feature_names):

From e844aac31d5f959dba7d4f7a1b098b0f1aaa2618 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:26:07 +0100
Subject: [PATCH 10/26] Do not return

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py | 1 -
 1 file changed, 1 deletion(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 34dff9d3c54b8..a0a9384bbbeea 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -435,7 +435,6 @@ def _index_output(self, Xs):
             n_columns = Xs[transformer_idx].shape[1]
             self.transformers_output_[name] = slice(idx, idx + n_columns)
             idx += n_columns
-        return
 
     def _validate_features(self, n_features, feature_names):
         """Ensures feature counts and names are the same during fit and

From b8191ec7d409ed5129eba8f6fc6cf288fd1102ac Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:27:29 +0100
Subject: [PATCH 11/26] Rename _index_output

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
Co-Authored-By: Joel Nothman <78827+jnothman@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index a0a9384bbbeea..ae14ee812865d 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -423,7 +423,7 @@ def _validate_output(self, result):
                     "The output of the '{0}' transformer should be 2D (scipy "
                     "matrix, array, or pandas DataFrame).".format(name))
 
-    def _index_output(self, Xs):
+    def _record_transformer_slices(self, Xs):
         """
         Record which transformer produced which column.
         """
@@ -574,7 +574,7 @@ def fit_transform(self, X, y=None):
 
         self._update_fitted_transformers(transformers)
         self._validate_output(Xs)
-        self._index_output(Xs)
+        self._record_transformer_slices(Xs)
 
         return self._hstack(list(Xs))
 

From 6665c0341ed53242bbc10e55b89329f0f07b4ff1 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:31:35 +0100
Subject: [PATCH 12/26] Rename attribute

Co-Authored-By: Joel Nothman <78827+jnothman@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py         |  6 +++---
 .../compose/tests/test_column_transformer.py   | 18 +++++++++---------
 2 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index ae14ee812865d..31766735c80a6 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -136,7 +136,7 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         sparse matrix or a dense numpy array, which depends on the output
         of the individual transformers and the `sparse_threshold` keyword.
 
-    transformers_output_ : dict
+    transformer_slices_ : dict
         A dictionary from transformer names to :py:class:`slice`s, where
         each slice corresponds to indices in the transformed output.
         This is useful to inspect which transformer is responsible for
@@ -428,12 +428,12 @@ def _record_transformer_slices(self, Xs):
         Record which transformer produced which column.
         """
         idx = 0
-        self.transformers_output_ = {}
+        self.transformer_slices_ = {}
         for transformer_idx, (name, _, _, _) in enumerate(
             self._iter(fitted=True, replace_strings=True)
         ):
             n_columns = Xs[transformer_idx].shape[1]
-            self.transformers_output_[name] = slice(idx, idx + n_columns)
+            self.transformer_slices_[name] = slice(idx, idx + n_columns)
             idx += n_columns
 
     def _validate_features(self, n_features, feature_names):
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 1deda43a0a784..1919eb33d356c 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -316,26 +316,26 @@ def test_column_transformer_transformers_output():
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_array)
-    assert ct.transformers_output_ == {'trans1': slice(0, 1),
-                                       'trans2': slice(1, 2)}
+    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
+                                      'trans2': slice(1, 2)}
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
                            transformer_weights={'trans': .1})
     ct.fit(X_array)
-    assert ct.transformers_output_ == {'trans': slice(0, 2)}
+    assert ct.transformer_slices_ == {'trans': slice(0, 2)}
 
     # test case that ensures that the attribute does also work when
     # a given transformer doesn't have any columns to work on
     ct = ColumnTransformer([('trans1', Trans(), [0, 1]),
                             ('trans2', TransRaise(), [])])
     ct.fit(X_array)
-    assert ct.transformers_output_ == {'trans1': slice(0, 2)}
+    assert ct.transformer_slices_ == {'trans1': slice(0, 2)}
 
     ct = ColumnTransformer([('trans', TransRaise(), [])],
                            remainder='passthrough')
     ct.fit(X_array)
-    assert ct.transformers_output_ == {'remainder': slice(0, 2)}
+    assert ct.transformer_slices_ == {'remainder': slice(0, 2)}
 
     # test with data frame
     pd = pytest.importorskip('pandas')
@@ -344,14 +344,14 @@ def test_column_transformer_transformers_output():
     ct = ColumnTransformer([('trans1', Trans(), ['first']),
                             ('trans2', Trans(), ['second'])])
     ct.fit(X_df)
-    assert ct.transformers_output_ == {'trans1': slice(0, 1),
-                                       'trans2': slice(1, 2)}
+    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
+                                      'trans2': slice(1, 2)}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_df)
-    assert ct.transformers_output_ == {'trans1': slice(0, 1),
-                                       'trans2': slice(1, 2)}
+    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
+                                      'trans2': slice(1, 2)}
 
 
 def test_column_transformer_sparse_array():

From 127760bb167507ab504c7435c9bf700958a44c19 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 22:34:18 +0100
Subject: [PATCH 13/26] Add comment to tests

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/tests/test_column_transformer.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 1919eb33d356c..43ed6a2ba329c 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -310,7 +310,8 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert isinstance(ct.transformers_[0][1], TransRaise)
 
 
-def test_column_transformer_transformers_output():
+def test_column_transformer_transformer_slices():
+    # Checks for the transformer_slices_ attribute
     X_array = np.array([[0, 1, 2], [2, 4, 6]]).T
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),

From a0dd2857cab0f678bb2c59233617bd818415629d Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 23:08:14 +0100
Subject: [PATCH 14/26] Add entries for transformers without output

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py        |  8 ++++++++
 .../compose/tests/test_column_transformer.py  | 19 +++++++++++++------
 2 files changed, 21 insertions(+), 6 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 31766735c80a6..6ecf432ac5d4e 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -429,6 +429,7 @@ def _record_transformer_slices(self, Xs):
         """
         idx = 0
         self.transformer_slices_ = {}
+
         for transformer_idx, (name, _, _, _) in enumerate(
             self._iter(fitted=True, replace_strings=True)
         ):
@@ -436,6 +437,13 @@ def _record_transformer_slices(self, Xs):
             self.transformer_slices_[name] = slice(idx, idx + n_columns)
             idx += n_columns
 
+        # empty slices for transformers that generate no output (except
+        # for 'remainder' when 'drop'); these are safe for indexing
+        all_names = [t[0] for t in self.transformers] + ['remainder']
+        for name in all_names:
+            if name not in self.transformer_slices_:
+                self.transformer_slices_[name] = slice(-1, 0)
+
     def _validate_features(self, n_features, feature_names):
         """Ensures feature counts and names are the same during fit and
         transform.
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 43ed6a2ba329c..c9e7a4ad3a799 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -318,25 +318,30 @@ def test_column_transformer_transformer_slices():
                             ('trans2', Trans(), [1])])
     ct.fit(X_array)
     assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2)}
+                                      'trans2': slice(1, 2),
+                                      'remainder': slice(-1, 0)}
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
                            transformer_weights={'trans': .1})
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans': slice(0, 2)}
+    assert ct.transformer_slices_ == {'trans': slice(0, 2),
+                                      'remainder': slice(-1, 0)}
 
     # test case that ensures that the attribute does also work when
     # a given transformer doesn't have any columns to work on
     ct = ColumnTransformer([('trans1', Trans(), [0, 1]),
                             ('trans2', TransRaise(), [])])
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans1': slice(0, 2)}
+    assert ct.transformer_slices_ == {'trans1': slice(0, 2),
+                                      'trans2': slice(-1, 0),
+                                      'remainder': slice(-1, 0)}
 
     ct = ColumnTransformer([('trans', TransRaise(), [])],
                            remainder='passthrough')
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'remainder': slice(0, 2)}
+    assert ct.transformer_slices_ == {'trans': slice(-1, 0),
+                                      'remainder': slice(0, 2)}
 
     # test with data frame
     pd = pytest.importorskip('pandas')
@@ -346,13 +351,15 @@ def test_column_transformer_transformer_slices():
                             ('trans2', Trans(), ['second'])])
     ct.fit(X_df)
     assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2)}
+                                      'trans2': slice(1, 2),
+                                      'remainder': slice(-1, 0)}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_df)
     assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2)}
+                                      'trans2': slice(1, 2),
+                                      'remainder': slice(-1, 0)}
 
 
 def test_column_transformer_sparse_array():

From 29e12f3fa1dc7e72e038367df7c03b053bc255b9 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 23:37:53 +0100
Subject: [PATCH 15/26] Docstring

---
 sklearn/compose/_column_transformer.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 6ecf432ac5d4e..9eadd1276b06e 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -137,8 +137,8 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         of the individual transformers and the `sparse_threshold` keyword.
 
     transformer_slices_ : dict
-        A dictionary from transformer names to :py:class:`slice`s, where
-        each slice corresponds to indices in the transformed output.
+        A dictionary from each transformer name to a :py:object:`slice`,
+        where the slice corresponds to indices in the transformed output.
         This is useful to inspect which transformer is responsible for
         which transformed feature(s).
 

From ee44d509818bf8bc8b7e656b1d9ff9d9f3c51c1e Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 23:41:26 +0100
Subject: [PATCH 16/26] Rename attribute

---
 sklearn/compose/_column_transformer.py        | 14 ++++----
 .../compose/tests/test_column_transformer.py  | 36 +++++++++----------
 2 files changed, 25 insertions(+), 25 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 9eadd1276b06e..51e13c4d35b01 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -136,7 +136,7 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         sparse matrix or a dense numpy array, which depends on the output
         of the individual transformers and the `sparse_threshold` keyword.
 
-    transformer_slices_ : dict
+    output_indices_ : dict
         A dictionary from each transformer name to a :py:object:`slice`,
         where the slice corresponds to indices in the transformed output.
         This is useful to inspect which transformer is responsible for
@@ -423,26 +423,26 @@ def _validate_output(self, result):
                     "The output of the '{0}' transformer should be 2D (scipy "
                     "matrix, array, or pandas DataFrame).".format(name))
 
-    def _record_transformer_slices(self, Xs):
+    def _record_output_indices(self, Xs):
         """
         Record which transformer produced which column.
         """
         idx = 0
-        self.transformer_slices_ = {}
+        self.output_indices_ = {}
 
         for transformer_idx, (name, _, _, _) in enumerate(
             self._iter(fitted=True, replace_strings=True)
         ):
             n_columns = Xs[transformer_idx].shape[1]
-            self.transformer_slices_[name] = slice(idx, idx + n_columns)
+            self.output_indices_[name] = slice(idx, idx + n_columns)
             idx += n_columns
 
         # empty slices for transformers that generate no output (except
         # for 'remainder' when 'drop'); these are safe for indexing
         all_names = [t[0] for t in self.transformers] + ['remainder']
         for name in all_names:
-            if name not in self.transformer_slices_:
-                self.transformer_slices_[name] = slice(-1, 0)
+            if name not in self.output_indices_:
+                self.output_indices_[name] = slice(-1, 0)
 
     def _validate_features(self, n_features, feature_names):
         """Ensures feature counts and names are the same during fit and
@@ -582,7 +582,7 @@ def fit_transform(self, X, y=None):
 
         self._update_fitted_transformers(transformers)
         self._validate_output(Xs)
-        self._record_transformer_slices(Xs)
+        self._record_output_indices(Xs)
 
         return self._hstack(list(Xs))
 
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index c9e7a4ad3a799..63ba954935c4e 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -310,38 +310,38 @@ def test_column_transformer_empty_columns(pandas, column_selection,
     assert isinstance(ct.transformers_[0][1], TransRaise)
 
 
-def test_column_transformer_transformer_slices():
-    # Checks for the transformer_slices_ attribute
+def test_column_transformer_output_indices():
+    # Checks for the output_indices_ attribute
     X_array = np.array([[0, 1, 2], [2, 4, 6]]).T
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2),
-                                      'remainder': slice(-1, 0)}
+    assert ct.output_indices_ == {'trans1': slice(0, 1),
+                                  'trans2': slice(1, 2),
+                                  'remainder': slice(-1, 0)}
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
                            transformer_weights={'trans': .1})
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans': slice(0, 2),
-                                      'remainder': slice(-1, 0)}
+    assert ct.output_indices_ == {'trans': slice(0, 2),
+                                  'remainder': slice(-1, 0)}
 
     # test case that ensures that the attribute does also work when
     # a given transformer doesn't have any columns to work on
     ct = ColumnTransformer([('trans1', Trans(), [0, 1]),
                             ('trans2', TransRaise(), [])])
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans1': slice(0, 2),
-                                      'trans2': slice(-1, 0),
-                                      'remainder': slice(-1, 0)}
+    assert ct.output_indices_ == {'trans1': slice(0, 2),
+                                  'trans2': slice(-1, 0),
+                                  'remainder': slice(-1, 0)}
 
     ct = ColumnTransformer([('trans', TransRaise(), [])],
                            remainder='passthrough')
     ct.fit(X_array)
-    assert ct.transformer_slices_ == {'trans': slice(-1, 0),
-                                      'remainder': slice(0, 2)}
+    assert ct.output_indices_ == {'trans': slice(-1, 0),
+                                  'remainder': slice(0, 2)}
 
     # test with data frame
     pd = pytest.importorskip('pandas')
@@ -350,16 +350,16 @@ def test_column_transformer_transformer_slices():
     ct = ColumnTransformer([('trans1', Trans(), ['first']),
                             ('trans2', Trans(), ['second'])])
     ct.fit(X_df)
-    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2),
-                                      'remainder': slice(-1, 0)}
+    assert ct.output_indices_ == {'trans1': slice(0, 1),
+                                  'trans2': slice(1, 2),
+                                  'remainder': slice(-1, 0)}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_df)
-    assert ct.transformer_slices_ == {'trans1': slice(0, 1),
-                                      'trans2': slice(1, 2),
-                                      'remainder': slice(-1, 0)}
+    assert ct.output_indices_ == {'trans1': slice(0, 1),
+                                  'trans2': slice(1, 2),
+                                  'remainder': slice(-1, 0)}
 
 
 def test_column_transformer_sparse_array():

From f8013d81ce59e86be2e4c1035945bfb377f3d340 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Thu, 19 Nov 2020 23:42:44 +0100
Subject: [PATCH 17/26] Inconsistent tests

---
 sklearn/compose/tests/test_column_transformer.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 63ba954935c4e..53bd6ff1ab668 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -312,7 +312,7 @@ def test_column_transformer_empty_columns(pandas, column_selection,
 
 def test_column_transformer_output_indices():
     # Checks for the output_indices_ attribute
-    X_array = np.array([[0, 1, 2], [2, 4, 6]]).T
+    X_array = np.arange(6).reshape(3, 2)
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])

From 94cf92471dfb7d766aad4d529501434077139df6 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Fri, 20 Nov 2020 07:57:54 +0100
Subject: [PATCH 18/26] Docstrings

---
 sklearn/compose/_column_transformer.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 51e13c4d35b01..5b3081c336598 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -137,7 +137,7 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         of the individual transformers and the `sparse_threshold` keyword.
 
     output_indices_ : dict
-        A dictionary from each transformer name to a :py:object:`slice`,
+        A dictionary from each transformer name to a :object:`slice`,
         where the slice corresponds to indices in the transformed output.
         This is useful to inspect which transformer is responsible for
         which transformed feature(s).

From 14899f7bf5338b221dc066441a8d27cc3124d7a0 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Fri, 20 Nov 2020 08:43:08 +0100
Subject: [PATCH 19/26] Docstrings without reference

---
 sklearn/compose/_column_transformer.py | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 5b3081c336598..e786070258df3 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -137,10 +137,10 @@ class ColumnTransformer(TransformerMixin, _BaseComposition):
         of the individual transformers and the `sparse_threshold` keyword.
 
     output_indices_ : dict
-        A dictionary from each transformer name to a :object:`slice`,
-        where the slice corresponds to indices in the transformed output.
-        This is useful to inspect which transformer is responsible for
-        which transformed feature(s).
+        A dictionary from each transformer name to a slice, where the slice
+        corresponds to indices in the transformed output. This is useful to
+        inspect which transformer is responsible for which transformed
+        feature(s).
 
     Notes
     -----

From 7d5ed362e5de786042eb95ab9e5707ffe5a170a2 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Sat, 21 Nov 2020 22:18:55 +0100
Subject: [PATCH 20/26] slice(0, 0)

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py           |  2 +-
 sklearn/compose/tests/test_column_transformer.py | 14 +++++++-------
 2 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index e786070258df3..07fd66e9068b0 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -442,7 +442,7 @@ def _record_output_indices(self, Xs):
         all_names = [t[0] for t in self.transformers] + ['remainder']
         for name in all_names:
             if name not in self.output_indices_:
-                self.output_indices_[name] = slice(-1, 0)
+                self.output_indices_[name] = slice(0, 0)
 
     def _validate_features(self, n_features, feature_names):
         """Ensures feature counts and names are the same during fit and
diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 53bd6ff1ab668..458048c35c5a2 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -319,14 +319,14 @@ def test_column_transformer_output_indices():
     ct.fit(X_array)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
-                                  'remainder': slice(-1, 0)}
+                                  'remainder': slice(0, 0)}
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
                            transformer_weights={'trans': .1})
     ct.fit(X_array)
     assert ct.output_indices_ == {'trans': slice(0, 2),
-                                  'remainder': slice(-1, 0)}
+                                  'remainder': slice(0, 0)}
 
     # test case that ensures that the attribute does also work when
     # a given transformer doesn't have any columns to work on
@@ -334,13 +334,13 @@ def test_column_transformer_output_indices():
                             ('trans2', TransRaise(), [])])
     ct.fit(X_array)
     assert ct.output_indices_ == {'trans1': slice(0, 2),
-                                  'trans2': slice(-1, 0),
-                                  'remainder': slice(-1, 0)}
+                                  'trans2': slice(0, 0),
+                                  'remainder': slice(0, 0)}
 
     ct = ColumnTransformer([('trans', TransRaise(), [])],
                            remainder='passthrough')
     ct.fit(X_array)
-    assert ct.output_indices_ == {'trans': slice(-1, 0),
+    assert ct.output_indices_ == {'trans': slice(0, 0),
                                   'remainder': slice(0, 2)}
 
     # test with data frame
@@ -352,14 +352,14 @@ def test_column_transformer_output_indices():
     ct.fit(X_df)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
-                                  'remainder': slice(-1, 0)}
+                                  'remainder': slice(0, 0)}
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
     ct.fit(X_df)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
-                                  'remainder': slice(-1, 0)}
+                                  'remainder': slice(0, 0)}
 
 
 def test_column_transformer_sparse_array():

From 50ed71f644ed71c7c7e326e14c9dc7839ec34714 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Sat, 21 Nov 2020 22:19:26 +0100
Subject: [PATCH 21/26] Update outdated comment

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 sklearn/compose/_column_transformer.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index 07fd66e9068b0..8223898f38244 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -437,8 +437,8 @@ def _record_output_indices(self, Xs):
             self.output_indices_[name] = slice(idx, idx + n_columns)
             idx += n_columns
 
-        # empty slices for transformers that generate no output (except
-        # for 'remainder' when 'drop'); these are safe for indexing
+        # empty slices for transformers that generate no output; these
+        # are safe for indexing
         all_names = [t[0] for t in self.transformers] + ['remainder']
         for name in all_names:
             if name not in self.output_indices_:

From 06ba19a2b99365daef88823edbe924a41326d6c6 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Sat, 21 Nov 2020 22:19:51 +0100
Subject: [PATCH 22/26] More tests

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 .../compose/tests/test_column_transformer.py  | 42 ++++++++++++++++---
 1 file changed, 36 insertions(+), 6 deletions(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index 458048c35c5a2..a9beb4503243c 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -316,32 +316,50 @@ def test_column_transformer_output_indices():
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
-    ct.fit(X_array)
+    X_trans = ct.fit_transform(X_array)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
                                   'remainder': slice(0, 0)}
+    assert_array_equal(X_trans[:, [0]],
+                       X_trans[:, ct.output_indices_['trans1']])
+    assert_array_equal(X_trans[:, [1]],
+                       X_trans[:, ct.output_indices_['trans2']])
 
     # test with transformer_weights and multiple columns
     ct = ColumnTransformer([('trans', Trans(), [0, 1])],
                            transformer_weights={'trans': .1})
-    ct.fit(X_array)
+    X_trans = ct.fit_transform(X_array)
     assert ct.output_indices_ == {'trans': slice(0, 2),
                                   'remainder': slice(0, 0)}
+    assert_array_equal(X_trans[:, [0, 1]],
+                       X_trans[:, ct.output_indices_['trans']])
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['remainder']])
 
     # test case that ensures that the attribute does also work when
     # a given transformer doesn't have any columns to work on
     ct = ColumnTransformer([('trans1', Trans(), [0, 1]),
                             ('trans2', TransRaise(), [])])
-    ct.fit(X_array)
+    X_trans = ct.fit_transform(X_array)
     assert ct.output_indices_ == {'trans1': slice(0, 2),
                                   'trans2': slice(0, 0),
                                   'remainder': slice(0, 0)}
+    assert_array_equal(X_trans[:, [0, 1]],
+                       X_trans[:, ct.output_indices_['trans1']])
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['trans2']])
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['remainder']])
 
     ct = ColumnTransformer([('trans', TransRaise(), [])],
                            remainder='passthrough')
-    ct.fit(X_array)
+    X_trans = ct.fit_transform(X_array)
     assert ct.output_indices_ == {'trans': slice(0, 0),
                                   'remainder': slice(0, 2)}
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['trans']])
+    assert_array_equal(X_trans[:, [0, 1]],
+                       X_trans[:, ct.output_indices_['remainder']])
 
     # test with data frame
     pd = pytest.importorskip('pandas')
@@ -349,17 +367,29 @@ def test_column_transformer_output_indices():
 
     ct = ColumnTransformer([('trans1', Trans(), ['first']),
                             ('trans2', Trans(), ['second'])])
-    ct.fit(X_df)
+    X_trans = ct.fit_transform(X_df)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
                                   'remainder': slice(0, 0)}
+    assert_array_equal(X_trans[:, [0]],
+                       X_trans[:, ct.output_indices_['trans1']])
+    assert_array_equal(X_trans[:, [1]],
+                       X_trans[:, ct.output_indices_['trans2']])
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['remainder']])
 
     ct = ColumnTransformer([('trans1', Trans(), [0]),
                             ('trans2', Trans(), [1])])
-    ct.fit(X_df)
+    X_trans = ct.fit_transform(X_df)
     assert ct.output_indices_ == {'trans1': slice(0, 1),
                                   'trans2': slice(1, 2),
                                   'remainder': slice(0, 0)}
+    assert_array_equal(X_trans[:, [0]],
+                       X_trans[:, ct.output_indices_['trans1']])
+    assert_array_equal(X_trans[:, [1]],
+                       X_trans[:, ct.output_indices_['trans2']])
+    assert_array_equal(X_trans[:, []],
+                       X_trans[:, ct.output_indices_['remainder']])
 
 
 def test_column_transformer_sparse_array():

From 36b20e8be70d994bef999cd7dfd11a2a8af07a67 Mon Sep 17 00:00:00 2001
From: Luca Bittarello <15511539+lbittarello@users.noreply.github.com>
Date: Sat, 21 Nov 2020 22:23:17 +0100
Subject: [PATCH 23/26] Log changes

Co-Authored-By: Nicolas Hug <NicolasHug@users.noreply.github.com>
---
 doc/whats_new/v0.24.rst | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/doc/whats_new/v0.24.rst b/doc/whats_new/v0.24.rst
index e8225811194cc..ae5e974ee9a0e 100644
--- a/doc/whats_new/v0.24.rst
+++ b/doc/whats_new/v0.24.rst
@@ -89,6 +89,9 @@ Changelog
 - |FIX| :class:`compose.ColumnTransformer` now displays the remainder in the
   diagram display. :pr:`18167` by `Thomas Fan`_.
 
+- |Enhancement| :class:`compose.ColumnTransformer` now records the output
+  of each transformer. :pr:`18393` by :user:`Luca Bittarello <lbittarello>`.
+
 :mod:`sklearn.covariance`
 .........................
 

From 27cfee5615bb304778bc8a171e332e480e86529e Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Sat, 27 Mar 2021 12:49:15 +0100
Subject: [PATCH 24/26] Update change logs

---
 doc/whats_new/v0.24.rst | 3 ---
 doc/whats_new/v1.0.rst  | 7 +++++++
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/doc/whats_new/v0.24.rst b/doc/whats_new/v0.24.rst
index e34ad84a98ab3..68ea8ba0f7a72 100644
--- a/doc/whats_new/v0.24.rst
+++ b/doc/whats_new/v0.24.rst
@@ -19,9 +19,6 @@ Changelog
   :term:`get_feature_names` on transformers with an empty column selection.
   :pr:`19579` by `Thomas Fan`_.
 
-- |Enhancement| :class:`compose.ColumnTransformer` now records the output
-  of each transformer. :pr:`18393` by :user:`Luca Bittarello <lbittarello>`.
-
 :mod:`sklearn.ensemble`
 .......................
 
diff --git a/doc/whats_new/v1.0.rst b/doc/whats_new/v1.0.rst
index be894774f5a27..a54abb78730a4 100644
--- a/doc/whats_new/v1.0.rst
+++ b/doc/whats_new/v1.0.rst
@@ -102,6 +102,13 @@ Changelog
 - |Enhancement| :func:`datasets.fetch_kddcup99` raises a better message
   when the cached file is invalid. :pr:`19669` `Thomas Fan`_.
 
+:mod:`sklearn.compose`
+......................
+
+- |Enhancement| :class:`compose.ColumnTransformer` now records the output
+  of each transformer in `output_indices_`. :pr:`18393` by
+  :user:`Luca Bittarello <lbittarello>`.
+
 :mod:`sklearn.decomposition`
 ............................
 

From 2b96828eb971e1375cdfe235eda5b2f456307845 Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Sat, 27 Mar 2021 12:53:26 +0100
Subject: [PATCH 25/26] Update comment

---
 sklearn/compose/_column_transformer.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/_column_transformer.py b/sklearn/compose/_column_transformer.py
index da4eab1bf5e14..da4a2dd93507c 100644
--- a/sklearn/compose/_column_transformer.py
+++ b/sklearn/compose/_column_transformer.py
@@ -428,8 +428,9 @@ def _record_output_indices(self, Xs):
             self.output_indices_[name] = slice(idx, idx + n_columns)
             idx += n_columns
 
-        # empty slices for transformers that generate no output; these
-        # are safe for indexing
+        # `_iter` only generates transformers that have a non empty
+        # selection. Here we set empty slices for transformers that
+        # generate no output, which are safe for indexing
         all_names = [t[0] for t in self.transformers] + ['remainder']
         for name in all_names:
             if name not in self.output_indices_:

From 5ee27f9a815a2b0ff8fba3709115ad69828ac6c0 Mon Sep 17 00:00:00 2001
From: lbittarello <luca.bittarello@gmail.com>
Date: Sat, 27 Mar 2021 12:57:14 +0100
Subject: [PATCH 26/26] Split tests

---
 sklearn/compose/tests/test_column_transformer.py | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/sklearn/compose/tests/test_column_transformer.py b/sklearn/compose/tests/test_column_transformer.py
index a73f6f82d3623..f7c1874d4a1b7 100644
--- a/sklearn/compose/tests/test_column_transformer.py
+++ b/sklearn/compose/tests/test_column_transformer.py
@@ -361,9 +361,13 @@ def test_column_transformer_output_indices():
     assert_array_equal(X_trans[:, [0, 1]],
                        X_trans[:, ct.output_indices_['remainder']])
 
-    # test with data frame
+
+def test_column_transformer_output_indices_df():
+    # Checks for the output_indices_ attribute with data frames
     pd = pytest.importorskip('pandas')
-    X_df = pd.DataFrame(X_array, columns=['first', 'second'])
+
+    X_df = pd.DataFrame(np.arange(6).reshape(3, 2),
+                        columns=['first', 'second'])
 
     ct = ColumnTransformer([('trans1', Trans(), ['first']),
                             ('trans2', Trans(), ['second'])])