All function tests are passing, yay!

jedi/evaluate/dynamic.py

@@ -24,8 +24,9 @@ from jedi import settings
 from jedi import debug
 from jedi.evaluate.cache import memoize_default
 from jedi.evaluate import imports
-from jedi.evaluate.param import TreeArguments
+from jedi.evaluate.param import TreeArguments, create_default_param
 from jedi.common import to_list, unite
+from jedi.evaluate import context
 
 
 MAX_PARAM_SEARCHES = 20
@@ -54,7 +55,7 @@ class MergedExecutedParams(object):
 
 
 @debug.increase_indent
-def search_params(evaluator, funcdef):
+def search_params(evaluator, parent_context, funcdef):
     """
     A dynamic search for param values. If you try to complete a type:
 
@@ -74,12 +75,15 @@ def search_params(evaluator, funcdef):
     try:
         debug.dbg('Dynamic param search in %s.', funcdef.name.value, color='MAGENTA')
         function_executions = _search_function_executions(evaluator, funcdef)
-        zipped_params = zip(*(
-            function_execution.get_params()
-            for function_execution in function_executions
-        ))
-        params = [MergedExecutedParams(executed_params) for executed_params in zipped_params]
-        # Evaluate the ExecutedParams to types.
+        if function_executions:
+            zipped_params = zip(*(
+                function_execution.get_params()
+                for function_execution in function_executions
+            ))
+            params = [MergedExecutedParams(executed_params) for executed_params in zipped_params]
+            # Evaluate the ExecutedParams to types.
+        else:
+            params = [create_default_param(parent_context, p) for p in funcdef.params]
         debug.dbg('Dynamic param result finished', color='MAGENTA')
         return params
     finally:

jedi/evaluate/param.py

@@ -259,11 +259,7 @@ def get_params(evaluator, parent_context, func, var_args):
         # args / kwargs will just be empty arrays / dicts, respectively.
         # Wrong value count is just ignored. If you try to test cases that are
         # not allowed in Python, Jedi will maybe not show any completions.
-        default = None
-        if param.default is not None:
-            default = context.LazyTreeContext(parent_context, param.default)
-
-        key, argument = next(var_arg_iterator, (None, default))
+        key, argument = next(var_arg_iterator, (None, None))
         while key is not None:
             keys_only = True
             try:
@@ -415,3 +411,19 @@ def _error_argument_count(func, actual_count):
         before = 'from %s to ' % (len(func.params) - default_arguments)
     return ('TypeError: %s() takes %s%s arguments (%s given).'
             % (func.name, before, len(func.params), actual_count))
+
+
+def create_default_param(parent_context, param):
+    if param.stars == 1:
+        result_arg = context.LazyKnownContext(
+            iterable.FakeSequence(parent_context._evaluator, 'tuple', [])
+        )
+    elif param.stars == 2:
+        result_arg = context.LazyKnownContext(
+            iterable.FakeDict(parent_context._evaluator, {})
+        )
+    elif param.default is None:
+        result_arg = context.LazyUnknownContext()
+    else:
+        result_arg = context.LazyTreeContext(parent_context, param.default)
+    return ExecutedParam(param, None, result_arg)

jedi/evaluate/representation.py

@@ -719,7 +719,7 @@ class AnonymousFunctionExecution(FunctionExecutionContext):
     @memoize_default(default=NO_DEFAULT)
     def get_params(self):
         # We need to do a dynamic search here.
-        return search_params(self._evaluator, self.funcdef)
+        return search_params(self._evaluator, self.parent_context, self.funcdef)
 
 
 class GlobalName(helpers.FakeName):