Fix broken properties when search has not been run

docs/source/changelog.rst

@@ -28,6 +28,7 @@ Changelog
         * Enforce requirement that builtin components save all inputted values in their parameters dict :pr:`847`
         * Don't list base classes in `all_components` output :pr:`847`
         * Standardize all components to output pandas datastructures, and accept either pandas or numpy :pr:`853`
+        * Fixed rankings and full_rankings error when search has not been run :pr:`894`
     * Changes
         * Update `all_pipelines` and `all_components` to try initializing pipelines/components, and on failure exclude them :pr:`849`
         * Refactor `handle_components` to `handle_components_class`, standardize to `ComponentBase` subclass instead of instance :pr:`850`

evalml/automl/automl_search.py

@@ -217,10 +217,10 @@ def _get_funct_name(function):
             f"Optimize Thresholds: {self.optimize_thresholds}\n"
         )
 
-        try:
+        if not self.rankings.empty:
             rankings_str = self.rankings.drop(['parameters'], axis='columns').to_string()
             rankings_desc = f"\nSearch Results: \n{'='*20}\n{rankings_str}"
-        except KeyError:
+        else:
             rankings_desc = ""
 
         return search_desc + rankings_desc
@@ -630,6 +630,11 @@ def add_to_rankings(self, pipeline, X, y):
                 return
         self._evaluate(pipeline, X, y, raise_errors=True)
 
+    @property
+    def has_searched(self):
+        searched = True if self.results['pipeline_results'] else False
+        return searched
+
     @property
     def rankings(self):
         """Returns a pandas.DataFrame with scoring results from the highest-scoring set of parameters used with each pipeline."""
@@ -642,6 +647,9 @@ def full_rankings(self):
         if self.objective.greater_is_better:
             ascending = False
 
+        if not self.has_searched:
+            return pd.DataFrame(columns=["id", "pipeline_name", "score", "high_variance_cv", "parameters"])
+
         rankings_df = pd.DataFrame(self.results['pipeline_results'].values())
         rankings_df = rankings_df[["id", "pipeline_name", "score", "high_variance_cv", "parameters"]]
         rankings_df.sort_values("score", ascending=ascending, inplace=True)
@@ -651,6 +659,10 @@ def full_rankings(self):
     @property
     def best_pipeline(self):
         """Returns the best model found"""
+
+        if not self.has_searched:
+            raise RuntimeError("automl search must be run before selecting `best_pipeline`.")
+
         best = self.rankings.iloc[0]
         return self.get_pipeline(best["id"])
 

evalml/tests/automl_tests/test_automl.py

@@ -499,6 +499,20 @@ def test_add_to_rankings(mock_fit, mock_score, dummy_binary_pipeline_class, X_y)
     assert 0.1234 in automl.rankings['score'].values
 
 
+@patch('evalml.pipelines.BinaryClassificationPipeline.score')
+@patch('evalml.pipelines.BinaryClassificationPipeline.fit')
+def test_add_to_rankings_no_search(mock_fit, mock_score, dummy_binary_pipeline_class, X_y):
+    X, y = X_y
+    automl = AutoMLSearch(problem_type='binary', max_pipelines=1, allowed_pipelines=[dummy_binary_pipeline_class])
+
+    mock_score.return_value = {'Log Loss Binary': 0.1234}
+    test_pipeline = dummy_binary_pipeline_class(parameters={})
+    automl.add_to_rankings(test_pipeline, X, y)
+
+    assert len(automl.rankings) == 1
+    assert 0.1234 in automl.rankings['score'].values
+
+
 @patch('evalml.pipelines.BinaryClassificationPipeline.score')
 @patch('evalml.pipelines.BinaryClassificationPipeline.fit')
 def test_add_to_rankings_duplicate(mock_fit, mock_score, dummy_binary_pipeline_class, X_y):
@@ -536,3 +550,36 @@ class CoolBinaryClassificationPipeline(dummy_binary_pipeline_class):
     automl.add_to_rankings(test_pipeline_trained, X, y)
 
     assert list(automl.rankings['score'].values).count(0.1234) == 2
+
+
+@patch('evalml.pipelines.BinaryClassificationPipeline.score')
+@patch('evalml.pipelines.BinaryClassificationPipeline.fit')
+def test_has_searched(mock_fit, mock_score, dummy_binary_pipeline_class, X_y):
+    X, y = X_y
+
+    automl = AutoMLSearch(problem_type='binary', max_pipelines=1)
+    assert automl.has_searched is False
+
+    automl.search(X, y)
+    assert automl.has_searched is True
+
+
+def test_no_search():
+    automl = AutoMLSearch(problem_type='binary')
+    assert isinstance(automl.rankings, pd.DataFrame)
+    assert isinstance(automl.full_rankings, pd.DataFrame)
+
+    df_columns = ["id", "pipeline_name", "score", "high_variance_cv", "parameters"]
+    assert (automl.rankings.columns == df_columns).all()
+    assert (automl.full_rankings.columns == df_columns).all()
+
+    assert automl._data_check_results is None
+
+    with pytest.raises(RuntimeError):
+        automl.best_pipeline
+
+    with pytest.raises(RuntimeError):
+        automl.get_pipeline(0)
+
+    with pytest.raises(RuntimeError):
+        automl.describe_pipeline(0)