Fixing bug: centroids failed for clusters with summary fields

HISTORY.rst

@@ -3,6 +3,12 @@
 History
 -------
 
+1.10.5 (2014-01-29)
+~~~~~~~~~~~~~~~~~~~
+
+- Fixing bug: centroids failed when predicted from local clusters with
+  summary fields.
+
 1.10.4 (2014-01-17)
 ~~~~~~~~~~~~~~~~~~~
 

bigml/__init__.py

@@ -1 +1 @@
-__version__ = '1.10.4'
+__version__ = '1.10.5'

bigml/cluster.py

@@ -143,6 +143,9 @@ def __init__(self, cluster, api=None):
                 self.tag_clouds = {}
                 self.term_analysis = {}
                 fields = cluster['clusters']['fields']
+                summary_fields = cluster['summary_fields']
+                for field_id in summary_fields:
+                    del fields[field_id]
                 for field_id, field in fields.items():
                     if field['optype'] == 'text':
 

tests/features/compare_predictions.feature

@@ -100,6 +100,23 @@ Feature: Compare Predictions
         | ../data/iris_sp_chars.csv | 20      | 20     | 30     | {"fields": {}} |{"pétal.length":1, "pétal&width\u0000": 2, "sépal.length":1, "sépal&width": 2, "spécies": "Iris-setosa"}       | Cluster 7    | 0.757736964835     |
 
 
+    Scenario: Successfully comparing centroids with summary fields:
+        Given I create a data source uploading a "<data>" file
+        And I wait until the source is ready less than <time_1> secs
+        And I create a dataset
+        And I wait until the dataset is ready less than <time_2> secs
+        And I create a cluster with options "<options>"
+        And I wait until the cluster is ready less than <time_3> secs
+        And I create a local cluster
+        When I create a centroid for "<data_input>"
+        Then the centroid is "<centroid>" with distance "<distance>"
+        And I create a local centroid for "<data_input>"
+        Then the local centroid is "<centroid>" with distance "<distance>"
+
+        Examples:
+        | data             | time_1  | time_2 | time_3 | options | data_input                            | centroid  | distance |
+        | ../data/iris.csv | 20      | 20     | 30     | {"summary_fields": ["sepal width"]} |{"petal length": 1, "petal width": 1, "sepal length": 1, "species": "Iris-setosa"}             | Cluster 6   | 0.713698082167   |
+
     Scenario: Successfully comparing predictions with proportional missing strategy for missing_splits models:
         Given I create a data source uploading a "<data>" file
         And I wait until the source is ready less than <time_1> secs

tests/features/create_cluster-steps.py

@@ -49,6 +49,22 @@ def i_create_a_cluster_from_dataset_list(step):
     world.cluster = resource['object']
     world.clusters.append(resource['resource'])
 
+
+@step(r'I create a cluster with options "(.*)"$')
+def i_create_a_cluster_with_options(step, options):
+    dataset = world.dataset.get('resource')
+    options = json.loads(options)
+    options.update({'seed': 'BigML',
+                    'cluster_seed': 'BigML',
+                    'k': 8})
+    resource = world.api.create_cluster(
+        dataset, options)
+    world.status = resource['code']
+    assert world.status == HTTP_CREATED
+    world.location = resource['location']
+    world.cluster = resource['object']
+    world.clusters.append(resource['resource'])
+
 @step(r'I wait until the cluster status code is either (\d) or (-\d) less than (\d+)')
 def wait_until_cluster_status_code_is(step, code1, code2, secs):
     start = datetime.utcnow()