diff --git a/gridtk/generator.py b/gridtk/generator.py
index b82ea6081907cc512780a170cbdc9e8f29096810..a1a744cf1105e65386527a6afc1c1ebcb5134841 100644
--- a/gridtk/generator.py
+++ b/gridtk/generator.py
@@ -11,9 +11,17 @@ import yaml
 import jinja2
 
 
+class _OrderedDict(collections.OrderedDict):
+  """An OrderedDict class that can be compared.
+  This is to avoid sort errors (in Python 3) that happen in jinja internally.
+  """
+  def __lt__(self, other):
+    return id(self) < id(other)
+
+
 def _ordered_load(stream, Loader=yaml.Loader,
-    object_pairs_hook=collections.OrderedDict):
-  '''Loads the contents of the YAML stream into :py:class:`collection.OrderedDict`'s
+    object_pairs_hook=_OrderedDict):
+  '''Loads the contents of the YAML stream into :py:class:`collections.OrderedDict`'s
 
   See: https://stackoverflow.com/questions/5121931/in-python-how-can-you-load-yaml-mappings-as-ordereddicts
 
@@ -123,8 +131,8 @@ def expand(data):
 
   # separates "unique" objects from the ones we have to iterate
   # pre-assemble return dictionary
-  iterables = collections.OrderedDict()
-  unique = collections.OrderedDict()
+  iterables = _OrderedDict()
+  unique = _OrderedDict()
   for key, value in data.items():
     if isinstance(value, list) and not key.startswith('_'):
       iterables[key] = value
@@ -133,7 +141,7 @@ def expand(data):
 
   # generates all possible combinations of iterables
   for values in itertools.product(*iterables.values()):
-    retval = collections.OrderedDict(unique)
+    retval = _OrderedDict(unique)
     keys = list(iterables.keys())
     retval.update(dict(zip(keys, values)))
     yield retval
diff --git a/gridtk/tests/test_generator.py b/gridtk/tests/test_generator.py
index 6a6d8ea9b4cde860f78a409bfa05423c1500c7e0..880afbcfb748dcbc7843e13dd890e2a4715279d8 100644
--- a/gridtk/tests/test_generator.py
+++ b/gridtk/tests/test_generator.py
@@ -275,3 +275,73 @@ def test_cmdline_unique_aggregation():
 
   finally:
     shutil.rmtree(tmpdir)
+
+
+def test_cmdline_aggregation_dict_groupby():
+
+  data = """
+model:
+  - {name: patch_1, patch_size: 28}
+
+train:
+  - {database: replaymobile, protocol: grandtest}
+  - {database: replay, protocol: grandtest}
+
+eval:
+  - {database: replaymobile, protocol: grandtest, groups: ['dev', 'eval']}
+  - {database: replay, protocol: grandtest, groups: ['dev', 'eval']}
+"""
+
+  template = '{{ model.name }}-{{ train.database }}-{{ eval.database }}'
+
+  aggtmpl = """
+{% set cfg2 = cfgset|groupby('train')|map(attribute='list') -%}
+{% for cfg3 in cfg2 %}
+{% set k = cfg3[0] -%}
+test-{{ k.model.name }}-{{ k.train.database }}-{{ k.eval.database }}
+{%- endfor %}
+"""
+
+  gen_expected = [
+      'patch_1-replay-replay',
+      'patch_1-replay-replaymobile',
+      'patch_1-replaymobile-replay',
+      'patch_1-replaymobile-replaymobile',
+  ]
+
+  agg_expected = [
+      '',
+      '',
+      'test-patch_1-replaymobile-replaymobile',
+      'test-patch_1-replay-replaymobile',
+  ]
+  tmpdir = tempfile.mkdtemp()
+
+  try:
+    variables = os.path.join(tmpdir, 'variables.yaml')
+    with open(variables, 'wt') as f: f.write(data)
+    gentmpl = os.path.join(tmpdir, 'gentmpl.txt')
+    with open(gentmpl, 'wt') as f: f.write(template)
+    genout = os.path.join(tmpdir, 'out', template + '.txt')
+
+    aggtmpl_file = os.path.join(tmpdir, 'agg.txt')
+    with open(aggtmpl_file, 'wt') as f: f.write(aggtmpl)
+    aggout = os.path.join(tmpdir, 'out', 'agg.txt')
+
+    nose.tools.eq_(jgen.main(['-vv', variables, gentmpl, genout, aggtmpl_file,
+      aggout]), 0)
+
+    # check all files are there and correspond to the expected output
+    outdir = os.path.dirname(genout)
+    for k in gen_expected:
+      ofile = os.path.join(outdir, k + '.txt')
+      assert os.path.exists(ofile)
+      with open(ofile, 'rt') as f: contents = f.read()
+      nose.tools.eq_(contents, k)
+    assert os.path.exists(aggout)
+    with open(aggout, 'rt') as f: contents = f.read()
+    for line in agg_expected:
+      assert line in contents, contents
+
+  finally:
+    shutil.rmtree(tmpdir)