Merge pull request #277 from kayak/data_blending_no_dimension_restric…

…tion

Data blending dimensions

fireant/queries/builder/dataset_blender_query_builder.py

@@ -5,7 +5,6 @@
     List,
 )
 
-from fireant.exceptions import DataSetException
 from fireant.queries.builder.dataset_query_builder import DataSetQueryBuilder
 from fireant.queries.finders import (
     find_dataset_metrics,
@@ -37,7 +36,7 @@ def _flatten_blend_datasets(dataset) -> List:
     return zip(*_flatten_blend_datasets(blender))
 
 
-def _replace_field(dimension, field_map=None, dataset=None, omit_umapped=False):
+def _replace_field(dimension, field_map=None, dataset=None):
     root_dimension = find_field_in_modified_field(dimension)
     if root_dimension is not dimension:
         # Handle modified dimensions
@@ -47,18 +46,12 @@ def _replace_field(dimension, field_map=None, dataset=None, omit_umapped=False):
     if field_map is not None and dimension in field_map:
         return field_map.get(dimension, None)
 
-    if dataset is not None and dimension.alias in dataset.fields:
-        return dataset.fields[dimension.alias]
+    if dataset is not None:
+        if dimension.definition is not None and dimension.definition in dataset.fields:
+            return dimension.definition
 
-    if dimension.definition is not None:
-        return dimension.definition
-
-    if not omit_umapped and dimension not in field_map:
-        raise DataSetException(
-            "Invalid Dimension {}. Dimensions must be mapped in order to be used in a blender query.".format(
-                dimension.alias
-            )
-        )
+        if dimension.alias in dataset.fields:
+            return dataset.fields[dimension.alias]
 
     return field_map.get(dimension, None)
 
@@ -80,12 +73,16 @@ def _build_dataset_query(dataset, field_map, metrics, dimensions, filters, refer
         return None
 
     blended_dimensions = [
-        _replace_field(dimension, field_map, dataset) for dimension in dimensions
+        dimension
+        for dimension in [
+            _replace_field(dimension, field_map, dataset) for dimension in dimensions
+        ]
+        if dimension is not None
     ]
 
     blended_filters = []
     for fltr in filters:
-        filter_field = _replace_field(fltr.field, field_map, dataset, omit_umapped=True)
+        filter_field = _replace_field(fltr.field, field_map, dataset)
 
         if filter_field not in dataset.fields:
             continue
@@ -114,9 +111,12 @@ def _join_criteria_for_blender_subqueries(primary, secondary, dimensions, field_
     join_criteria = []
 
     for dimension in dimensions:
-        mapped_dimension = _replace_field(dimension, field_map)
-        p_alias = alias_selector(dimension.alias)
-        s_alias = alias_selector(mapped_dimension.alias)
+        primary_dimension = find_field_in_modified_field(dimension).definition
+        if primary_dimension not in field_map:
+            continue
+        secondary_dimension = field_map[primary_dimension]
+        p_alias = alias_selector(primary_dimension.alias)
+        s_alias = alias_selector(secondary_dimension.alias)
         join_criteria.append(primary[p_alias] == secondary[s_alias])
 
     return reduce(lambda a, b: a & b, join_criteria)

fireant/tests/queries/test_build_data_blending.py

@@ -103,6 +103,45 @@ def test_using_datablender_metric_builds_query(self):
             str(queries[0]),
         )
 
+    def test_using_datablender_builds_query_with_mapped_and_unmapped_dimensions(self):
+        queries = (
+            mock_dataset_blender.query()
+            .widget(
+                f.ReactTable(mock_dataset_blender.fields["candidate-spend-per-wins"])
+            )
+            .dimension(
+                f.day(mock_dataset_blender.fields.timestamp),
+                mock_dataset_blender.fields.political_party,
+            )
+        ).sql
+
+        self.assertEqual(len(queries), 1)
+        self.assertEqual(
+            "SELECT "
+            '"sq0"."$timestamp" "$timestamp",'
+            '"sq0"."$political_party" "$political_party",'
+            '"sq1"."$candidate-spend"/"sq0"."$wins" "$candidate-spend-per-wins" '
+            "FROM ("
+            "SELECT "
+            'TRUNC("timestamp",\'DD\') "$timestamp",'
+            '"political_party" "$political_party",'
+            'SUM("is_winner") "$wins" '
+            'FROM "politics"."politician" '
+            'GROUP BY "$timestamp","$political_party" ORDER BY "$timestamp","$political_party"'
+            ') "sq0" '
+            "LEFT JOIN ("
+            "SELECT "
+            'TRUNC("timestamp",\'DD\') "$timestamp",'
+            'SUM("candidate_spend") "$candidate-spend" '
+            'FROM "politics"."politician_spend" '
+            'GROUP BY "$timestamp" ORDER BY "$timestamp"'
+            ') "sq1" '
+            "ON "
+            '"sq0"."$timestamp"="sq1"."$timestamp" '
+            'ORDER BY "$timestamp","$political_party"',
+            str(queries[0]),
+        )
+
     def test_apply_metric_filter_to_dataset_field_filters_in_nested_dataset_query(self):
         queries = (
             mock_dataset_blender.query()
@@ -494,14 +533,3 @@ def test_does_not_raise_SlicerException_when_a_dimension_is_not_mapped_for_unnec
             .widget(f.ReactTable(mock_dataset_blender.fields["votes"]))
             .dimension(mock_dataset_blender.fields["district-id"])
         ).sql
-
-    def test_raises_SlicerException_when_a_dimension_from_a_necessary_secondary_dataset_is_not_mapped(
-        self,
-    ):
-        with self.assertRaises(DataSetException):
-            queries = (
-                mock_dataset_blender.query()
-                .widget(f.ReactTable(mock_dataset_blender.fields["candidate-spend"]))
-                .dimension(f.day(mock_dataset_blender.fields.timestamp))
-                .dimension(mock_dataset_blender.fields["district-id"])
-            ).sql