Field names in big query can contain only alphanumeric and underscore

.gitignore

@@ -1,4 +1,5 @@
 *.pyc
+*.swp
 yarn-error.log
 _modules
 superset/assets/coverage/*

superset/db_engine_specs.py

@@ -426,6 +426,10 @@ def align_df_col_names_with_form_data(df, fd):
 
         return df.rename(index=str, columns=rename_cols)
 
+    @staticmethod
+    def mutate_expression_label(label):
+        return label
+
 
 class PostgresBaseEngineSpec(BaseEngineSpec):
     """ Abstract class for Postgres 'like' databases """
@@ -1414,6 +1418,18 @@ def fetch_data(cls, cursor, limit):
             data = [r.values() for r in data]
         return data
 
+    @staticmethod
+    def mutate_expression_label(label):
+        mutated_label = re.sub('[^\w]+', '_', label)
+        if not re.match('^[a-zA-Z_]+.*', mutated_label):
+            raise SupersetTemplateException('BigQuery field_name used is invalid {}, '
+                                            'should start with a letter or '
+                                            'underscore'.format(mutated_label))
+        if len(mutated_label) > 128:
+            raise SupersetTemplateException('BigQuery field_name {}, should be atmost '
+                                            '128 characters'.format(mutated_label))
+        return mutated_label
+
     @classmethod
     def _get_fields(cls, cols):
         """

superset/viz.py

@@ -108,7 +108,10 @@ def process_metrics(self):
                 if not isinstance(val, list):
                     val = [val]
                 for o in val:
-                    self.metric_dict[self.get_metric_label(o)] = o
+                    label = self.get_metric_label(o)
+                    if isinstance(o, dict):
+                        o['label'] = label
+                    self.metric_dict[label] = o
 
         # Cast to list needed to return serializable object in py3
         self.all_metrics = list(self.metric_dict.values())
@@ -118,7 +121,8 @@ def get_metric_label(self, metric):
         if isinstance(metric, string_types):
             return metric
         if isinstance(metric, dict):
-            return metric.get('label')
+            return self.datasource.database.db_engine_spec.mutate_expression_label(
+                metric.get('label'))
 
     @staticmethod
     def handle_js_int_overflow(data):

tests/viz_tests.py

@@ -122,6 +122,20 @@ def test_cache_timeout(self):
 
 
 class TableVizTestCase(unittest.TestCase):
+
+    class DBEngineSpecMock:
+        @staticmethod
+        def mutate_expression_label(label):
+            return label
+
+    class DatabaseMock:
+        def __init__(self):
+            self.db_engine_spec = TableVizTestCase.DBEngineSpecMock()
+
+    class DatasourceMock:
+        def __init__(self):
+            self.database = TableVizTestCase.DatabaseMock()
+
     def test_get_data_applies_percentage(self):
         form_data = {
             'percent_metrics': [{
@@ -137,7 +151,7 @@ def test_get_data_applies_percentage(self):
                 'column': {'column_name': 'value1', 'type': 'DOUBLE'},
             }, 'count', 'avg__C'],
         }
-        datasource = Mock()
+        datasource = TableVizTestCase.DatasourceMock()
         raw = {}
         raw['SUM(value1)'] = [15, 20, 25, 40]
         raw['avg__B'] = [10, 20, 5, 15]