Feature #369 (#424)

* Implementing dataset listing with more filters and adding unit tests

* Add to function documentation

Author: ArlindKadra

openml/datasets/functions.py

@@ -137,7 +137,7 @@ def _get_cached_dataset_arff(dataset_id):
                                    "cached" % dataset_id)
 
 
-def list_datasets(offset=None, size=None, tag=None, status=None):
+def list_datasets(offset=None, size=None, status=None, **kwargs):
     """Return a list of all dataset which are on OpenML.
 
     Parameters
@@ -146,12 +146,13 @@ def list_datasets(offset=None, size=None, tag=None, status=None):
         The number of datasets to skip, starting from the first.
     size : int, optional
         The maximum number of datasets to show.
-    tag : str, optional
-        Only include datasets matching this tag.
     status : str, optional
         Should be {active, in_preparation, deactivated}. By
         default active datasets are returned, but also datasets
-        from another status can be requested. 
+        from another status can be requested.
+    kwargs : dict, optional
+        Legal filter operators (keys in the dict):
+        {tag, status, limit, offset, data_name, data_version, number_instances, number_features, number_classes, number_missing_values}.
 
     Returns
     -------
@@ -175,12 +176,13 @@ def list_datasets(offset=None, size=None, tag=None, status=None):
     if size is not None:
         api_call += "/limit/%d" % int(size)
 
-    if tag is not None:
-        api_call += "/tag/%s" % tag
-
     if status is not None:
         api_call += "/status/%s" %status
 
+    if kwargs is not None:
+        for filter, value in kwargs.items():
+            api_call += "/%s/%s" % (filter, value)
+
     return _list_datasets(api_call)
 
 

tests/test_datasets/test_dataset_functions.py

@@ -115,6 +115,9 @@ def _check_dataset(self, dataset):
             self.assertIsInstance(dataset['status'], six.string_types)
             self.assertIn(dataset['status'], ['in_preparation', 'active',
                                               'deactivated'])
+    def _check_datasets(self, datasets):
+        for did in datasets:
+            self._check_dataset(datasets[did])
 
     def test_tag_untag_dataset(self):
         tag = 'test_tag_%d' %random.randint(1, 1000000)
@@ -129,23 +132,45 @@ def test_list_datasets(self):
         datasets = openml.datasets.list_datasets()
         # 1087 as the number of datasets on openml.org
         self.assertGreaterEqual(len(datasets), 100)
-        for did in datasets:
-            self._check_dataset(datasets[did])
+        self._check_datasets(datasets)
 
     def test_list_datasets_by_tag(self):
         datasets = openml.datasets.list_datasets(tag='study_14')
         self.assertGreaterEqual(len(datasets), 100)
-        for did in datasets:
-            self._check_dataset(datasets[did])
+        self._check_datasets(datasets)
+
+    def test_list_datasets_by_number_instances(self):
+        datasets = openml.datasets.list_datasets(number_instances="5..100")
+        self.assertGreaterEqual(len(datasets), 4)
+        self._check_datasets(datasets)
+
+    def test_list_datasets_by_number_features(self):
+        datasets = openml.datasets.list_datasets(number_features="50..100")
+        self.assertGreaterEqual(len(datasets), 8)
+        self._check_datasets(datasets)
+
+    def test_list_datasets_by_number_classes(self):
+        datasets = openml.datasets.list_datasets(number_classes="5")
+        self.assertGreaterEqual(len(datasets), 3)
+        self._check_datasets(datasets)
+
+    def test_list_datasets_by_number_missing_values(self):
+        datasets = openml.datasets.list_datasets(number_missing_values="5..100")
+        self.assertGreaterEqual(len(datasets), 5)
+        self._check_datasets(datasets)
+
+    def test_list_datasets_combined_filters(self):
+        datasets = openml.datasets.list_datasets(tag='study_14', number_instances="100..1000", number_missing_values="800..1000")
+        self.assertGreaterEqual(len(datasets), 1)
+        self._check_datasets(datasets)
 
     def test_list_datasets_paginate(self):
         size = 10
         max = 100
         for i in range(0, max, size):
             datasets = openml.datasets.list_datasets(offset=i, size=size)
             self.assertGreaterEqual(size, len(datasets))
-            for did in datasets:
-                self._check_dataset(datasets[did])
+            self._check_datasets(datasets)
 
     def test_list_datasets_empty(self):
         datasets = openml.datasets.list_datasets(tag='NoOneWouldUseThisTagAnyway')