ENH: Add ability to access datasets by index (Fixes Unidata#134)

When getting an item by string, work as before. For all else, treat as
an index into values. Works for catalog refs as well.

siphon/catalog.py

@@ -24,6 +24,17 @@
 log.setLevel(logging.ERROR)
 
 
+class IndexableMapping(OrderedDict):
+    """Extend ``OrderedDict`` to allow index-based access to values."""
+    def __getitem__(self, item):
+        """Return an item either by index or name."""
+        try:
+            item + ''  # Raises if item not a string
+            return super().__getitem__(item)
+        except TypeError:
+            return list(self.values())[item]
+
+
 class TDSCatalog(object):
     """
     Parse information from a THREDDS Client Catalog.
@@ -79,9 +90,9 @@ def __init__(self, catalog_url):
         root = ET.fromstring(resp.text)
         self.catalog_name = root.attrib.get('name', 'No name found')
 
-        self.datasets = OrderedDict()
+        self.datasets = IndexableMapping()
         self.services = []
-        self.catalog_refs = OrderedDict()
+        self.catalog_refs = IndexableMapping()
         self.metadata = {}
         self.ds_with_access_elements_to_process = []
         service_skip_count = 0

siphon/tests/test_catalog.py

@@ -116,6 +116,17 @@ def test_datasets_order():
                                   'Latest Collection for NAM CONUS 20km']
 
 
+@recorder.use_cassette('top_level_20km_rap_catalog')
+def test_datasets_get_by_index():
+    """Test that datasets can be accessed by index."""
+    url = ('http://thredds.ucar.edu/thredds/catalog/grib/NCEP/NAM/'
+           'CONUS_20km/noaaport/catalog.xml')
+    cat = TDSCatalog(url)
+    assert cat.datasets[0].name == 'Full Collection (Reference / Forecast Time) Dataset'
+    assert cat.datasets[1].name == 'Best NAM CONUS 20km Time Series'
+    assert cat.datasets[2].name == 'Latest Collection for NAM CONUS 20km'
+
+
 @recorder.use_cassette('top_level_cat')
 def test_catalog_ref_order():
     """Test that catalog references are properly ordered."""