Remove the old star import cache, because it's not even used.

jedi/api/__init__.py

@@ -133,7 +133,6 @@ class Script(object):
 
     @cache.memoize_method
     def _get_module_node(self):
-        cache.invalidate_star_import_cache(self._path)
         parser = FastParser(self._grammar, self._source, self.path)
         save_parser(self.path, parser, pickling=False)
 

jedi/cache.py

@@ -90,31 +90,3 @@ def memoize_method(method):
             dct[key] = result
             return result
     return wrapper
-
-
-def _invalidate_star_import_cache_module(module, only_main=False):
-    """ Important if some new modules are being reparsed """
-    try:
-        t, modules = _time_caches['star_import_cache_validity'][module]
-    except KeyError:
-        pass
-    else:
-        del _time_caches['star_import_cache_validity'][module]
-
-        # This stuff was part of load_parser. However since we're most likely
-        # not going to use star import caching anymore, just ignore it.
-        #else:
-            # In case there is already a module cached and this module
-            # has to be reparsed, we also need to invalidate the import
-            # caches.
-        #    _invalidate_star_import_cache_module(parser_cache_item.parser.module)
-
-
-def invalidate_star_import_cache(path):
-    """On success returns True."""
-    try:
-        parser_cache_item = parser_cache[path]
-    except KeyError:
-        pass
-    else:
-        _invalidate_star_import_cache_module(parser_cache_item.parser.module)

jedi/parser/python/parser.py

@@ -196,8 +196,8 @@ class ParserWithRecovery(Parser):
 
     def parse(self, tokenizer):
         root_node = super(ParserWithRecovery, self).parse(self._tokenize(tokenizer))
+        root_node.path = self._module_path
         self.module = root_node
-        self.module.path = self._module_path
         return root_node
 
     def error_recovery(self, grammar, stack, arcs, typ, value, start_pos, prefix,

jedi/settings.py

@@ -68,7 +68,6 @@ definitely worse in some cases. But a completion should also be fast.
 Caching
 ~~~~~~~
 
-.. autodata:: star_import_cache_validity
 .. autodata:: call_signatures_validity
 
 
@@ -217,13 +216,6 @@ scale `max_executions` and `max_until_execution_unique`:
 # caching validity (time)
 # ----------------
 
-star_import_cache_validity = 60.0
-"""
-In huge packages like numpy, checking all star imports on every completion
-might be slow, therefore we do a star import caching, that lasts a certain
-time span (in seconds).
-"""
-
 call_signatures_validity = 3.0
 """
 Finding function calls might be slow (0.1-0.5s). This is not acceptible for