Increase Jedi version

Filter duplicate imports when completing

.github/workflows/ci.yml

@@ -1,5 +1,5 @@
 name: ci
-on: push
+on: [push, pull_request, workflow_dispatch]
 
 jobs:
   tests:
@@ -7,29 +7,28 @@ jobs:
     strategy:
       matrix:
         os: [ubuntu-20.04, windows-2019]
-        python-version: [3.9, 3.8, 3.7, 3.6]
-        environment: ['3.8', '3.9', '3.7', '3.6', 'interpreter']
+        python-version: ["3.13", "3.12", "3.11", "3.10", "3.9", "3.8", "3.7", "3.6"]
+        environment: ['3.8', '3.13', '3.12', '3.11', '3.10', '3.9', '3.7', '3.6', 'interpreter']
     steps:
       - name: Checkout code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v4
         with:
           submodules: recursive
 
-      - uses: actions/setup-python@v2
+      - uses: actions/setup-python@v5
         if: ${{ matrix.environment != 'interpreter' }}
         with:
           python-version: ${{ matrix.environment }}
+          allow-prereleases: true
 
-      - uses: actions/setup-python@v2
+      - uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
+          allow-prereleases: true
 
       - name: Install dependencies
         run: 'pip install .[testing]'
 
-      - name: Setup tmate session      
-        uses: mxschmitt/action-tmate@v3
-
       - name: Run tests
         run: python -m pytest
         env:
@@ -39,7 +38,7 @@ jobs:
     runs-on: ubuntu-20.04
     steps:
       - name: Checkout code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v4
         with:
           submodules: recursive
 
@@ -48,15 +47,15 @@ jobs:
 
       - name: Run tests
         run: |
-          python -m flake8 jedi setup.py
-          python -m mypy jedi sith.py
+          python -m flake8 jedi test setup.py
+          python -m mypy jedi sith.py setup.py
 
   coverage:
     runs-on: ubuntu-20.04
 
     steps:
       - name: Checkout code
-        uses: actions/checkout@v2
+        uses: actions/checkout@v4
         with:
           submodules: recursive
 

.gitignore

@@ -14,3 +14,4 @@ record.json
 /.pytest_cache
 /.mypy_cache
 /venv/
+.nvimrc

.readthedocs.yml

@@ -1,2 +1,16 @@
+version: 2
+
 python:
-   pip_install: true
+  install:
+    - method: pip
+      path: .
+      extra_requirements:
+        - docs
+
+submodules:
+  include: all
+
+build:
+  os: ubuntu-22.04
+  tools:
+    python: "3.11"

AUTHORS.txt

@@ -1,4 +1,4 @@
-Main Authors
+Main Authors
 ------------
 
 - David Halter (@davidhalter) <davidhalter88@gmail.com>
@@ -61,6 +61,11 @@ Code Contributors
 - Vladislav Serebrennikov (@endilll)
 - Andrii Kolomoiets (@muffinmad)
 - Leo Ryu (@Leo-Ryu)
+- Joseph Birkner (@josephbirkner)
+- Márcio Mazza (@marciomazza)
+- Martin Vielsmaier (@moser) <martin@vielsmaier.net>
+- TingJia Wu (@WutingjiaX) <wutingjia@bytedance.com>
+- Nguyễn Hồng Quân <ng.hong.quan@gmail.com>
 
 And a few more "anonymous" contributors.
 

CHANGELOG.rst

@@ -6,7 +6,41 @@ Changelog
 Unreleased
 ++++++++++
 
+0.19.2 (2024-11-10)
++++++++++++++++++++
+
+- Python 3.13 support
+
+0.19.1 (2023-10-02)
++++++++++++++++++++
+
+- Python 3.12 support (Thanks Peter!)
+
+0.19.0 (2023-07-29)
++++++++++++++++++++
+
+- Python 3.11 support
+- Massive improvements in performance for ``Interpreter`` (e.g. IPython) users.
+  This especially affects ``pandas`` users with large datasets.
+- Add ``jedi.settings.allow_unsafe_interpreter_executions`` to make it easier
+  for IPython users to avoid unsafe executions.
+
+0.18.2 (2022-11-21)
++++++++++++++++++++
+
+- Added dataclass-equivalent for attrs.define
+- Find fixtures from Pytest entrypoints; Examples of pytest plugins installed
+  like this are pytest-django, pytest-sugar and Faker.
+- Fixed Project.search, when a venv was involved, which is why for example
+  `:Pyimport django.db` did not work in some cases in jedi-vim.
+- And many smaller bugfixes
+
+0.18.1 (2021-11-17)
++++++++++++++++++++
+
 - Implict namespaces are now a separate types in ``Name().type``
+- Python 3.10 support
+- Mostly bugfixes
 
 0.18.0 (2020-12-25)
 +++++++++++++++++++

README.rst

@@ -42,7 +42,7 @@ Jedi can currently be used with the following editors/projects:
 - `GNOME Builder`_ (with support for GObject Introspection)
 - Gedit (gedi_)
 - wdb_ - Web Debugger
-- `Eric IDE`_ (Available as a plugin)
+- `Eric IDE`_
 - `IPython 6.0.0+ <https://ipython.readthedocs.io/en/stable/whatsnew/version6.html>`_
 - `xonsh shell <https://xon.sh/contents.html>`_ has `jedi extension <https://xon.sh/xontribs.html#jedi>`_
 
@@ -51,7 +51,8 @@ and many more!
 There are a few language servers that use Jedi:
 
 - `jedi-language-server <https://github.com/pappasam/jedi-language-server>`_
-- `python-language-server <https://github.com/palantir/python-language-server>`_
+- `python-language-server <https://github.com/palantir/python-language-server>`_ (currently unmaintained)
+- `python-lsp-server <https://github.com/python-lsp/python-lsp-server>`_ (fork from python-language-server)
 - `anakin-language-server <https://github.com/muffinmad/anakin-language-server>`_
 
 Here are some pictures taken from jedi-vim_:

SECURITY.md

@@ -0,0 +1,9 @@
+# Security Policy
+
+If security issues arise, we will try to fix those as soon as possible.
+
+Due to Jedi's nature, Security Issues will probably be extremely rare, but we will of course treat them seriously.
+
+## Reporting Security Problems
+
+If you need to report a security vulnerability, please send an email to davidhalter88@gmail.com. Typically, I will respond in the next few business days.

docs/docs/acknowledgements.rst

@@ -35,7 +35,7 @@ to write my own version of a completion engine.
 The first idea was to execute non-dangerous code. But I soon realized, that
 this would not work. So I started to build a static analysis tool.
 The biggest problem that I had at the time was that I did not know a thing
-about parsers.I did not did not even know the word static analysis. It turns
+about parsers. I did not even know the word static analysis. It turns
 out they are the foundation of a good static analysis tool. I of course did not
 know that and tried to write my own poor version of a parser that I ended up
 throwing away two years later.
@@ -53,7 +53,7 @@ quick and is pretty much feature complete.
 
 --------
 
-I will leave you with a small annectote that happend in 2012, if I remember
+I will leave you with a small anecdote that happened in 2012, if I remember
 correctly. After I explained Guido van Rossum, how some parts of my
 auto-completion work, he said:
 

docs/docs/features.rst

@@ -57,7 +57,7 @@ Supported Python Features
 Limitations
 -----------
 
-In general Jedi's limit are quite high, but for very big projects or very
+In general Jedi's limit is quite high, but for very big projects or very
 complex code, sometimes Jedi intentionally stops type inference, to avoid
 hanging for a long time.
 
@@ -77,7 +77,7 @@ Performance Issues
 
 Importing ``numpy`` can be quite slow sometimes, as well as loading the
 builtins the first time. If you want to speed things up, you could preload
-libriaries in |jedi|, with :func:`.preload_module`. However, once loaded, this
+libraries in |jedi|, with :func:`.preload_module`. However, once loaded, this
 should not be a problem anymore.  The same is true for huge modules like
 ``PySide``, ``wx``, ``tensorflow``, ``pandas``, etc.
 

docs/docs/usage.rst

@@ -4,7 +4,7 @@ Using Jedi
 ==========
 
 |jedi| is can be used with a variety of :ref:`plugins <editor-plugins>`,
-`language servers <language-servers>` and other software.
+:ref:`language servers <language-servers>` and other software.
 It is also possible to use |jedi| in the :ref:`Python shell or with IPython
 <repl-completion>`.
 
@@ -16,7 +16,8 @@ Language Servers
 --------------
 
 - `jedi-language-server <https://github.com/pappasam/jedi-language-server>`_
-- `python-language-server <https://github.com/palantir/python-language-server>`_
+- `python-language-server <https://github.com/palantir/python-language-server>`_ (currently unmaintained)
+- `python-lsp-server <https://github.com/python-lsp/python-lsp-server>`_ (fork from python-language-server)
 - `anakin-language-server <https://github.com/muffinmad/anakin-language-server>`_
 
 .. _editor-plugins:
@@ -86,7 +87,7 @@ Gedit
 Eric IDE
 ~~~~~~~~
 
-- `Eric IDE`_ (Available as a plugin)
+- `Eric IDE`_
 
 Web Debugger
 ~~~~~~~~~~~~

jedi/__init__.py

@@ -27,7 +27,7 @@ ad
 load
 """
 
-__version__ = '0.18.0'
+__version__ = '0.19.2'
 
 from jedi.api import Script, Interpreter, set_debug_function, preload_module
 from jedi import settings

jedi/_compatibility.py

@@ -5,27 +5,24 @@ different Python versions.
 import errno
 import sys
 import pickle
+from typing import Any
 
 
-def cast_path(string):
-    """
-    Take a bytes or str path and cast it to unicode.
-
-    Apparently it is perfectly fine to pass both byte and unicode objects into
-    the sys.path. This probably means that byte paths are normal at other
-    places as well.
-
-    Since this just really complicates everything and Python 2.7 will be EOL
-    soon anyway, just go with always strings.
-    """
-    if isinstance(string, bytes):
-        return str(string, encoding='UTF-8', errors='replace')
-    return str(string)
+class Unpickler(pickle.Unpickler):
+    def find_class(self, module: str, name: str) -> Any:
+        # Python 3.13 moved pathlib implementation out of __init__.py as part of
+        # generalising its implementation. Ensure that we support loading
+        # pickles from 3.13 on older version of Python. Since 3.13 maintained a
+        # compatible API, pickles from older Python work natively on the newer
+        # version.
+        if module == 'pathlib._local':
+            module = 'pathlib'
+        return super().find_class(module, name)
 
 
 def pickle_load(file):
     try:
-        return pickle.load(file)
+        return Unpickler(file).load()
     # Python on Windows don't throw EOF errors for pipes. So reraise them with
     # the correct type, which is caught upwards.
     except OSError:

jedi/api/__init__.py

@@ -13,7 +13,6 @@ from pathlib import Path
 import parso
 from parso.python import tree
 
-from jedi._compatibility import cast_path
 from jedi.parser_utils import get_executable_nodes
 from jedi import debug
 from jedi import settings
@@ -100,13 +99,15 @@ class Script:
     """
     def __init__(self, code=None, *, path=None, environment=None, project=None):
         self._orig_path = path
-        # An empty path (also empty string) should always result in no path.
         if isinstance(path, str):
             path = Path(path)
 
         self.path = path.absolute() if path else None
 
         if code is None:
+            if path is None:
+                raise ValueError("Must provide at least one of code or path")
+
             # TODO add a better warning than the traceback!
             with open(path, 'rb') as f:
                 code = f.read()
@@ -152,7 +153,7 @@ class Script:
         if self.path is None:
             file_io = None
         else:
-            file_io = KnownContentFileIO(cast_path(self.path), self._code)
+            file_io = KnownContentFileIO(self.path, self._code)
         if self.path is not None and self.path.suffix == '.pyi':
             # We are in a stub file. Try to load the stub properly.
             stub_module = load_proper_stub_module(
@@ -205,6 +206,7 @@ class Script:
             before magic methods and name mangled names that start with ``__``.
         :rtype: list of :class:`.Completion`
         """
+        self._inference_state.reset_recursion_limitations()
         with debug.increase_indent_cm('complete'):
             completion = Completion(
                 self._inference_state, self._get_module_context(), self._code_lines,
@@ -229,6 +231,7 @@ class Script:
         :param prefer_stubs: Prefer stubs to Python objects for this method.
         :rtype: list of :class:`.Name`
         """
+        self._inference_state.reset_recursion_limitations()
         pos = line, column
         leaf = self._module_node.get_name_of_position(pos)
         if leaf is None:
@@ -272,6 +275,7 @@ class Script:
         :param prefer_stubs: Prefer stubs to Python objects for this method.
         :rtype: list of :class:`.Name`
         """
+        self._inference_state.reset_recursion_limitations()
         tree_name = self._module_node.get_name_of_position((line, column))
         if tree_name is None:
             # Without a name we really just want to jump to the result e.g.
@@ -364,10 +368,17 @@ class Script:
 
         :rtype: list of :class:`.Name`
         """
+        self._inference_state.reset_recursion_limitations()
         definitions = self.goto(line, column, follow_imports=True)
         if definitions:
             return definitions
         leaf = self._module_node.get_leaf_for_position((line, column))
+
+        if leaf is not None and leaf.end_pos == (line, column) and leaf.type == 'newline':
+            next_ = leaf.get_next_leaf()
+            if next_ is not None and next_.start_pos == leaf.end_pos:
+                leaf = next_
+
         if leaf is not None and leaf.type in ('keyword', 'operator', 'error_leaf'):
             def need_pydoc():
                 if leaf.value in ('(', ')', '[', ']'):
@@ -399,6 +410,7 @@ class Script:
             the current module only.
         :rtype: list of :class:`.Name`
         """
+        self._inference_state.reset_recursion_limitations()
 
         def _references(include_builtins=True, scope='project'):
             if scope not in ('project', 'file'):
@@ -433,6 +445,7 @@ class Script:
 
         :rtype: list of :class:`.Signature`
         """
+        self._inference_state.reset_recursion_limitations()
         pos = line, column
         call_details = helpers.get_signature_details(self._module_node, pos)
         if call_details is None:
@@ -552,6 +565,7 @@ class Script:
         return parso_to_jedi_errors(self._inference_state.grammar, self._module_node)
 
     def _names(self, all_scopes=False, definitions=True, references=False):
+        self._inference_state.reset_recursion_limitations()
         # Set line/column to a random position, because they don't matter.
         module_context = self._get_module_context()
         defs = [
@@ -580,7 +594,7 @@ class Script:
     @validate_line_column
     def extract_variable(self, line, column, *, new_name, until_line=None, until_column=None):
         """
-        Moves an expression to a new statemenet.
+        Moves an expression to a new statement.
 
         For example if you have the cursor on ``foo`` and provide a
         ``new_name`` called ``bar``::
@@ -707,9 +721,8 @@ class Interpreter(Script):
     :param namespaces: A list of namespace dictionaries such as the one
         returned by :func:`globals` and :func:`locals`.
     """
-    _allow_descriptor_getattr_default = True
 
-    def __init__(self, code, namespaces, **kwds):
+    def __init__(self, code, namespaces, *, project=None, **kwds):
         try:
             namespaces = [dict(n) for n in namespaces]
         except Exception:
@@ -722,16 +735,32 @@ class Interpreter(Script):
             if not isinstance(environment, InterpreterEnvironment):
                 raise TypeError("The environment needs to be an InterpreterEnvironment subclass.")
 
-        super().__init__(code, environment=environment,
-                         project=Project(Path.cwd()), **kwds)
+        if project is None:
+            project = Project(Path.cwd())
+
+        super().__init__(code, environment=environment, project=project, **kwds)
+
         self.namespaces = namespaces
-        self._inference_state.allow_descriptor_getattr = self._allow_descriptor_getattr_default
+        self._inference_state.allow_unsafe_executions = \
+            settings.allow_unsafe_interpreter_executions
+        # Dynamic params search is important when we work on functions that are
+        # called by other pieces of code. However for interpreter completions
+        # this is not important at all, because the current code is always new
+        # and will never be called by something.
+        # Also sometimes this logic goes a bit too far like in
+        # https://github.com/ipython/ipython/issues/13866, where it takes
+        # seconds to do a simple completion.
+        self._inference_state.do_dynamic_params_search = False
 
     @cache.memoize_method
     def _get_module_context(self):
+        if self.path is None:
+            file_io = None
+        else:
+            file_io = KnownContentFileIO(self.path, self._code)
         tree_module_value = ModuleValue(
             self._inference_state, self._module_node,
-            file_io=KnownContentFileIO(str(self.path), self._code),
+            file_io=file_io,
             string_names=('__main__',),
             code_lines=self._code_lines,
         )

jedi/api/classes.py

@@ -27,7 +27,7 @@ from jedi.inference.compiled.mixed import MixedName
 from jedi.inference.names import ImportName, SubModuleName
 from jedi.inference.gradual.stub_value import StubModuleValue
 from jedi.inference.gradual.conversion import convert_names, convert_values
-from jedi.inference.base_value import ValueSet
+from jedi.inference.base_value import ValueSet, HasNoContext
 from jedi.api.keywords import KeywordName
 from jedi.api import completion_cache
 from jedi.api.helpers import filter_follow_imports
@@ -37,13 +37,17 @@ def _sort_names_by_start_pos(names):
     return sorted(names, key=lambda s: s.start_pos or (0, 0))
 
 
-def defined_names(inference_state, context):
+def defined_names(inference_state, value):
     """
     List sub-definitions (e.g., methods in class).
 
     :type scope: Scope
     :rtype: list of Name
     """
+    try:
+        context = value.as_context()
+    except HasNoContext:
+        return []
     filter = next(context.get_filters())
     names = [name for name in filter.values()]
     return [Name(inference_state, n) for n in _sort_names_by_start_pos(names)]
@@ -101,8 +105,7 @@ class BaseName:
             # Compiled modules should not return a module path even if they
             # have one.
             path: Optional[Path] = self._get_module_context().py__file__()
-            if path is not None:
-                return path
+            return path
 
         return None
 
@@ -759,7 +762,7 @@ class Name(BaseName):
         """
         defs = self._name.infer()
         return sorted(
-            unite(defined_names(self._inference_state, d.as_context()) for d in defs),
+            unite(defined_names(self._inference_state, d) for d in defs),
             key=lambda s: s._name.start_pos or (0, 0)
         )
 

jedi/api/completion.py

@@ -18,7 +18,8 @@ from jedi.inference import imports
 from jedi.inference.base_value import ValueSet
 from jedi.inference.helpers import infer_call_of_leaf, parse_dotted_names
 from jedi.inference.context import get_global_filters
-from jedi.inference.value import TreeInstance, ModuleValue
+from jedi.inference.value import TreeInstance
+from jedi.inference.docstring_utils import DocstringModule
 from jedi.inference.names import ParamNameWrapper, SubModuleName
 from jedi.inference.gradual.conversion import convert_values, convert_names
 from jedi.parser_utils import cut_value_at_position
@@ -64,12 +65,15 @@ def _must_be_kwarg(signatures, positional_count, used_kwargs):
     return must_be_kwarg
 
 
-def filter_names(inference_state, completion_names, stack, like_name, fuzzy, cached_name):
+def filter_names(inference_state, completion_names, stack, like_name, fuzzy,
+                 imported_names, cached_name):
     comp_dct = set()
     if settings.case_insensitive_completion:
         like_name = like_name.lower()
     for name in completion_names:
         string = name.string_name
+        if string in imported_names and string != like_name:
+            continue
         if settings.case_insensitive_completion:
             string = string.lower()
         if helpers.match(string, like_name, fuzzy=fuzzy):
@@ -137,6 +141,11 @@ class Completion:
 
         self._fuzzy = fuzzy
 
+    # Return list of completions in this order:
+    # - Beginning with what user is typing
+    # - Public (alphabet)
+    # - Private ("_xxx")
+    # - Dunder ("__xxx")
     def complete(self):
         leaf = self._module_node.get_leaf_for_position(
             self._original_position,
@@ -168,14 +177,19 @@ class Completion:
 
         cached_name, completion_names = self._complete_python(leaf)
 
+        imported_names = []
+        if leaf.parent is not None and leaf.parent.type in ['import_as_names', 'dotted_as_names']:
+            imported_names.extend(extract_imported_names(leaf.parent))
+
         completions = list(filter_names(self._inference_state, completion_names,
                                         self.stack, self._like_name,
-                                        self._fuzzy, cached_name=cached_name))
+                                        self._fuzzy, imported_names, cached_name=cached_name))
 
         return (
             # Removing duplicates mostly to remove False/True/None duplicates.
             _remove_duplicates(prefixed_completions, completions)
-            + sorted(completions, key=lambda x: (x.name.startswith('__'),
+            + sorted(completions, key=lambda x: (not x.name.startswith(self._like_name),
+                                                 x.name.startswith('__'),
                                                  x.name.startswith('_'),
                                                  x.name.lower()))
         )
@@ -194,7 +208,6 @@ class Completion:
         - In args: */**: no completion
         - In params (also lambda): no completion before =
         """
-
         grammar = self._inference_state.grammar
         self.stack = stack = None
         self._position = (
@@ -277,6 +290,10 @@ class Completion:
                 )
             elif nonterminals[-1] in ('trailer', 'dotted_name') and nodes[-1] == '.':
                 dot = self._module_node.get_leaf_for_position(self._position)
+                if dot.type == "endmarker":
+                    # This is a bit of a weird edge case, maybe we can somehow
+                    # generalize this.
+                    dot = leaf.get_previous_leaf()
                 cached_name, n = self._complete_trailer(dot.get_previous_leaf())
                 completion_names += n
             elif self._is_parameter_completion():
@@ -438,6 +455,7 @@ class Completion:
         - Having some doctest code that starts with `>>>`
         - Having backticks that doesn't have whitespace inside it
         """
+
         def iter_relevant_lines(lines):
             include_next_line = False
             for l in code_lines:
@@ -462,12 +480,12 @@ class Completion:
 
     def _complete_code_lines(self, code_lines):
         module_node = self._inference_state.grammar.parse(''.join(code_lines))
-        module_value = ModuleValue(
-            self._inference_state,
-            module_node,
+        module_value = DocstringModule(
+            in_module_context=self._module_context,
+            inference_state=self._inference_state,
+            module_node=module_node,
             code_lines=code_lines,
         )
-        module_value.parent_context = self._module_context
         return Completion(
             self._inference_state,
             module_value.as_context(),
@@ -660,3 +678,19 @@ def search_in_module(inference_state, module_context, names, wanted_names,
                     def_ = classes.Name(inference_state, n2)
                 if not wanted_type or wanted_type == def_.type:
                     yield def_
+
+
+def extract_imported_names(node):
+    imported_names = []
+
+    if node.type in ['import_as_names', 'dotted_as_names', 'import_as_name']:
+        for index, child in enumerate(node.children):
+            if child.type == 'name':
+                if (index > 0 and node.children[index - 1].type == "keyword"
+                        and node.children[index - 1].value == "as"):
+                    continue
+                imported_names.append(child.value)
+            elif child.type == 'import_as_name':
+                imported_names.extend(extract_imported_names(child))
+
+    return imported_names

jedi/api/environment.py

@@ -8,6 +8,7 @@ import hashlib
 import filecmp
 from collections import namedtuple
 from shutil import which
+from typing import TYPE_CHECKING
 
 from jedi.cache import memoize_method, time_cache
 from jedi.inference.compiled.subprocess import CompiledSubprocess, \
@@ -15,9 +16,13 @@ from jedi.inference.compiled.subprocess import CompiledSubprocess, \
 
 import parso
 
-_VersionInfo = namedtuple('VersionInfo', 'major minor micro')
+if TYPE_CHECKING:
+    from jedi.inference import InferenceState
 
-_SUPPORTED_PYTHONS = ['3.9', '3.8', '3.7', '3.6']
+
+_VersionInfo = namedtuple('VersionInfo', 'major minor micro')  # type: ignore[name-match]
+
+_SUPPORTED_PYTHONS = ['3.13', '3.12', '3.11', '3.10', '3.9', '3.8', '3.7', '3.6']
 _SAFE_PATHS = ['/usr/bin', '/usr/local/bin']
 _CONDA_VAR = 'CONDA_PREFIX'
 _CURRENT_VERSION = '%s.%s' % (sys.version_info.major, sys.version_info.minor)
@@ -102,7 +107,10 @@ class Environment(_BaseEnvironment):
         version = '.'.join(str(i) for i in self.version_info)
         return '<%s: %s in %s>' % (self.__class__.__name__, version, self.path)
 
-    def get_inference_state_subprocess(self, inference_state):
+    def get_inference_state_subprocess(
+        self,
+        inference_state: 'InferenceState',
+    ) -> InferenceStateSubprocess:
         return InferenceStateSubprocess(inference_state, self._get_subprocess())
 
     @memoize_method
@@ -134,7 +142,10 @@ class SameEnvironment(_SameEnvironmentMixin, Environment):
 
 
 class InterpreterEnvironment(_SameEnvironmentMixin, _BaseEnvironment):
-    def get_inference_state_subprocess(self, inference_state):
+    def get_inference_state_subprocess(
+        self,
+        inference_state: 'InferenceState',
+    ) -> InferenceStateSameProcess:
         return InferenceStateSameProcess(inference_state)
 
     def get_sys_path(self):
@@ -373,10 +384,13 @@ def _get_executable_path(path, safe=True):
     """
 
     if os.name == 'nt':
-        python = os.path.join(path, 'Scripts', 'python.exe')
+        pythons = [os.path.join(path, 'Scripts', 'python.exe'), os.path.join(path, 'python.exe')]
+    else:
+        pythons = [os.path.join(path, 'bin', 'python')]
+    for python in pythons:
+        if os.path.exists(python):
+            break
     else:
-        python = os.path.join(path, 'bin', 'python')
-    if not os.path.exists(python):
         raise InvalidPythonEnvironment("%s seems to be missing." % python)
 
     _assert_safe(python, safe)
@@ -384,8 +398,7 @@ def _get_executable_path(path, safe=True):
 
 
 def _get_executables_from_windows_registry(version):
-    # https://github.com/python/typeshed/pull/3794 adds winreg
-    import winreg  # type: ignore[import]
+    import winreg
 
     # TODO: support Python Anaconda.
     sub_keys = [

jedi/api/exceptions.py

@@ -23,7 +23,7 @@ class RefactoringError(_JediError):
     Refactorings can fail for various reasons. So if you work with refactorings
     like :meth:`.Script.rename`, :meth:`.Script.inline`,
     :meth:`.Script.extract_variable` and :meth:`.Script.extract_function`, make
-    sure to catch these. The descriptions in the errors are ususally valuable
+    sure to catch these. The descriptions in the errors are usually valuable
     for end users.
 
     A typical ``RefactoringError`` would tell the user that inlining is not

jedi/api/helpers.py

@@ -205,7 +205,6 @@ def filter_follow_imports(names, follow_builtin_imports=False):
 
 class CallDetails:
     def __init__(self, bracket_leaf, children, position):
-        ['bracket_leaf', 'call_index', 'keyword_name_str']
         self.bracket_leaf = bracket_leaf
         self._children = children
         self._position = position
@@ -281,7 +280,7 @@ class CallDetails:
     def count_positional_arguments(self):
         count = 0
         for star_count, key_start, had_equal in self._list_arguments()[:-1]:
-            if star_count:
+            if star_count or key_start:
                 break
             count += 1
         return count
@@ -307,7 +306,7 @@ def _iter_arguments(nodes, position):
             first = node.children[0]
             second = node.children[1]
             if second == '=':
-                if second.start_pos < position:
+                if second.start_pos < position and first.type == 'name':
                     yield 0, first.value, True
                 else:
                     yield 0, remove_after_pos(first), False

jedi/api/keywords.py

@@ -5,8 +5,7 @@ from typing import Dict, Optional
 from jedi.inference.names import AbstractArbitraryName
 
 try:
-    # https://github.com/python/typeshed/pull/4351 adds pydoc_data
-    from pydoc_data import topics  # type: ignore[import]
+    from pydoc_data import topics
     pydoc_topics: Optional[Dict[str, str]] = topics.topics
 except ImportError:
     # Python 3.6.8 embeddable does not have pydoc_data.

jedi/api/project.py

@@ -106,7 +106,16 @@ class Project:
         with open(self._get_json_path(self._path), 'w') as f:
             return json.dump((_SERIALIZER_VERSION, data), f)
 
-    def __init__(self, path, **kwargs):
+    def __init__(
+        self,
+        path,
+        *,
+        environment_path=None,
+        load_unsafe_extensions=False,
+        sys_path=None,
+        added_sys_path=(),
+        smart_sys_path=True,
+    ) -> None:
         """
         :param path: The base path for this project.
         :param environment_path: The Python executable path, typically the path
@@ -125,25 +134,22 @@ class Project:
             local directories. Otherwise you will have to rely on your packages
             being properly configured on the ``sys.path``.
         """
-        def py2_comp(path, environment_path=None, load_unsafe_extensions=False,
-                     sys_path=None, added_sys_path=(), smart_sys_path=True):
-            if isinstance(path, str):
-                path = Path(path).absolute()
-            self._path = path
 
-            self._environment_path = environment_path
-            if sys_path is not None:
-                # Remap potential pathlib.Path entries
-                sys_path = list(map(str, sys_path))
-            self._sys_path = sys_path
-            self._smart_sys_path = smart_sys_path
-            self._load_unsafe_extensions = load_unsafe_extensions
-            self._django = False
+        if isinstance(path, str):
+            path = Path(path).absolute()
+        self._path = path
+
+        self._environment_path = environment_path
+        if sys_path is not None:
             # Remap potential pathlib.Path entries
-            self.added_sys_path = list(map(str, added_sys_path))
-            """The sys path that is going to be added at the end of the """
-
-        py2_comp(path, **kwargs)
+            sys_path = list(map(str, sys_path))
+        self._sys_path = sys_path
+        self._smart_sys_path = smart_sys_path
+        self._load_unsafe_extensions = load_unsafe_extensions
+        self._django = False
+        # Remap potential pathlib.Path entries
+        self.added_sys_path = list(map(str, added_sys_path))
+        """The sys path that is going to be added at the end of the """
 
     @property
     def path(self):
@@ -328,7 +334,8 @@ class Project:
             )
 
         # 2. Search for identifiers in the project.
-        for module_context in search_in_file_ios(inference_state, file_ios, name):
+        for module_context in search_in_file_ios(inference_state, file_ios,
+                                                 name, complete=complete):
             names = get_module_names(module_context.tree_node, all_scopes=all_scopes)
             names = [module_context.create_name(n) for n in names]
             names = _remove_imports(names)
@@ -345,9 +352,8 @@ class Project:
         # 3. Search for modules on sys.path
         sys_path = [
             p for p in self._get_sys_path(inference_state)
-            # Exclude folders that are handled by recursing of the Python
-            # folders.
-            if not p.startswith(str(self._path))
+            # Exclude the current folder which is handled by recursing the folders.
+            if p != self._path
         ]
         names = list(iter_module_names(inference_state, empty_module_context, sys_path))
         yield from search_in_module(
@@ -426,7 +432,6 @@ def get_default_project(path=None):
             probable_path = dir
 
     if probable_path is not None:
-        # TODO search for setup.py etc
         return Project(probable_path)
 
     if first_no_init_file is not None:

jedi/api/refactoring/__init__.py

@@ -5,6 +5,7 @@ from typing import Dict, Iterable, Tuple
 from parso import split_lines
 
 from jedi.api.exceptions import RefactoringError
+from jedi.inference.value.namespace import ImplicitNSName
 
 EXPRESSION_PARTS = (
     'or_test and_test not_test comparison '
@@ -42,11 +43,17 @@ class ChangedFile:
         if self._from_path is None:
             from_p = ''
         else:
-            from_p = self._from_path.relative_to(project_path)
+            try:
+                from_p = self._from_path.relative_to(project_path)
+            except ValueError:  # Happens it the path is not on th project_path
+                from_p = self._from_path
         if self._to_path is None:
             to_p = ''
         else:
-            to_p = self._to_path.relative_to(project_path)
+            try:
+                to_p = self._to_path.relative_to(project_path)
+            except ValueError:
+                to_p = self._to_path
         diff = difflib.unified_diff(
             old_lines, new_lines,
             fromfile=str(from_p),
@@ -96,7 +103,12 @@ class Refactoring:
                 to_path=calculate_to_path(path),
                 module_node=next(iter(map_)).get_root_node(),
                 node_to_str_map=map_
-            ) for path, map_ in sorted(self._file_to_node_changes.items())
+            )
+            # We need to use `or`, because the path can be None
+            for path, map_ in sorted(
+                self._file_to_node_changes.items(),
+                key=lambda x: x[0] or Path("")
+            )
         }
 
     def get_renames(self) -> Iterable[Tuple[Path, Path]]:
@@ -110,7 +122,7 @@ class Refactoring:
         project_path = self._inference_state.project.path
         for from_, to in self.get_renames():
             text += 'rename from %s\nrename to %s\n' \
-                % (from_.relative_to(project_path), to.relative_to(project_path))
+                % (_try_relative_to(from_, project_path), _try_relative_to(to, project_path))
 
         return text + ''.join(f.get_diff() for f in self.get_changed_files().values())
 
@@ -140,13 +152,16 @@ def rename(inference_state, definitions, new_name):
         raise RefactoringError("There is no name under the cursor")
 
     for d in definitions:
+        # This private access is ok in a way. It's not public to
+        # protect Jedi users from seeing it.
         tree_name = d._name.tree_name
-        if d.type == 'module' and tree_name is None:
-            p = None if d.module_path is None else Path(d.module_path)
+        if d.type == 'module' and tree_name is None and d.module_path is not None:
+            p = Path(d.module_path)
             file_renames.add(_calculate_rename(p, new_name))
+        elif isinstance(d._name, ImplicitNSName):
+            for p in d._name._value.py__path__():
+                file_renames.add(_calculate_rename(Path(p), new_name))
         else:
-            # This private access is ok in a way. It's not public to
-            # protect Jedi users from seeing it.
             if tree_name is not None:
                 fmap = file_tree_name_map.setdefault(d.module_path, {})
                 fmap[tree_name] = tree_name.prefix + new_name
@@ -240,3 +255,10 @@ def _remove_indent_of_prefix(prefix):
     Removes the last indentation of a prefix, e.g. " \n \n " becomes " \n \n".
     """
     return ''.join(split_lines(prefix, keepends=True)[:-1])
+
+
+def _try_relative_to(path: Path, base: Path) -> Path:
+    try:
+        return path.relative_to(base)
+    except ValueError:
+        return path

jedi/api/strings.py

@@ -36,8 +36,11 @@ def complete_dict(module_context, code_lines, leaf, position, string, fuzzy):
             string = cut_value_at_position(leaf, position)
 
         context = module_context.create_context(bracket_leaf)
-        before_bracket_leaf = bracket_leaf.get_previous_leaf()
-        if before_bracket_leaf.type in ('atom', 'trailer', 'name'):
+
+        before_node = before_bracket_leaf = bracket_leaf.get_previous_leaf()
+        if before_node in (')', ']', '}'):
+            before_node = before_node.parent
+        if before_node.type in ('atom', 'trailer', 'name'):
             values = infer_call_of_leaf(context, before_bracket_leaf)
             return list(_completions_for_dicts(
                 module_context.inference_state,

jedi/debug.py

@@ -106,10 +106,7 @@ def dbg(message, *args, color='GREEN'):
         debug_function(color, i + 'dbg: ' + message % tuple(repr(a) for a in args))
 
 
-def warning(message, *args, **kwargs):
-    format = kwargs.pop('format', True)
-    assert not kwargs
-
+def warning(message, *args, format=True):
     if debug_function and enable_warning:
         i = ' ' * _debug_indent
         if format:

jedi/inference/__init__.py

@@ -90,7 +90,7 @@ class InferenceState:
         self.compiled_subprocess = environment.get_inference_state_subprocess(self)
         self.grammar = environment.get_grammar()
 
-        self.latest_grammar = parso.load_grammar(version='3.7')
+        self.latest_grammar = parso.load_grammar(version='3.13')
         self.memoize_cache = {}  # for memoize decorators
         self.module_cache = imports.ModuleCache()  # does the job of `sys.modules`.
         self.stub_module_cache = {}  # Dict[Tuple[str, ...], Optional[ModuleValue]]
@@ -99,10 +99,11 @@ class InferenceState:
         self.mixed_cache = {}  # see `inference.compiled.mixed._create()`
         self.analysis = []
         self.dynamic_params_depth = 0
+        self.do_dynamic_params_search = settings.dynamic_params
         self.is_analysis = False
         self.project = project
         self.access_cache = {}
-        self.allow_descriptor_getattr = False
+        self.allow_unsafe_executions = False
         self.flow_analysis_enabled = True
 
         self.reset_recursion_limitations()
@@ -125,7 +126,7 @@ class InferenceState:
     @inference_state_function_cache()
     def builtins_module(self):
         module_name = 'builtins'
-        builtins_module, = self.import_module((module_name,), sys_path=())
+        builtins_module, = self.import_module((module_name,), sys_path=[])
         return builtins_module
 
     @property  # type: ignore[misc]
@@ -181,8 +182,6 @@ class InferenceState:
 
     def parse_and_get_code(self, code=None, path=None,
                            use_latest_grammar=False, file_io=None, **kwargs):
-        if path is not None:
-            path = str(path)
         if code is None:
             if file_io is None:
                 file_io = FileIO(path)

jedi/inference/base_value.py

@@ -22,6 +22,10 @@ from jedi.cache import memoize_method
 sentinel = object()
 
 
+class HasNoContext(Exception):
+    pass
+
+
 class HelperValueMixin:
     def get_root_context(self):
         value = self
@@ -261,7 +265,7 @@ class Value(HelperValueMixin):
         return self.parent_context.is_stub()
 
     def _as_context(self):
-        raise NotImplementedError('Not all values need to be converted to contexts: %s', self)
+        raise HasNoContext
 
     @property
     def name(self):
@@ -293,7 +297,7 @@ class Value(HelperValueMixin):
             just the `_T` generic parameter.
 
         `value_set`: represents the actual argument passed to the parameter
-            we're inferrined for, or (for recursive calls) their types. In the
+            we're inferred for, or (for recursive calls) their types. In the
             above example this would first be the representation of the list
             `[1]` and then, when recursing, just of `1`.
         """

jedi/inference/compiled/access.py

@@ -8,6 +8,8 @@ import warnings
 import re
 import builtins
 import typing
+from pathlib import Path
+from typing import Optional, Tuple
 
 from jedi.inference.compiled.getattr_static import getattr_static
 
@@ -38,7 +40,7 @@ NOT_CLASS_TYPES = (
 MethodDescriptorType = type(str.replace)
 WrapperDescriptorType = type(set.__iter__)
 # `object.__subclasshook__` is an already executed descriptor.
-object_class_dict = type.__dict__["__dict__"].__get__(object)
+object_class_dict = type.__dict__["__dict__"].__get__(object)  # type: ignore[index]
 ClassMethodDescriptorType = type(object_class_dict['__subclasshook__'])
 
 _sentinel = object()
@@ -145,7 +147,7 @@ class AccessPath:
         self.accesses = accesses
 
 
-def create_access_path(inference_state, obj):
+def create_access_path(inference_state, obj) -> AccessPath:
     access = create_access(inference_state, obj)
     return AccessPath(access.get_access_path_tuples())
 
@@ -173,15 +175,15 @@ class DirectObjectAccess:
     def _create_access(self, obj):
         return create_access(self._inference_state, obj)
 
-    def _create_access_path(self, obj):
+    def _create_access_path(self, obj) -> AccessPath:
         return create_access_path(self._inference_state, obj)
 
     def py__bool__(self):
         return bool(self._obj)
 
-    def py__file__(self):
+    def py__file__(self) -> Optional[Path]:
         try:
-            return self._obj.__file__
+            return Path(self._obj.__file__)
         except AttributeError:
             return None
 
@@ -211,18 +213,39 @@ class DirectObjectAccess:
     def py__getitem__all_values(self):
         if isinstance(self._obj, dict):
             return [self._create_access_path(v) for v in self._obj.values()]
-        return self.py__iter__list()
+        if isinstance(self._obj, (list, tuple)):
+            return [self._create_access_path(v) for v in self._obj]
 
-    def py__simple_getitem__(self, index):
-        if type(self._obj) not in ALLOWED_GETITEM_TYPES:
+        if self.is_instance():
+            cls = DirectObjectAccess(self._inference_state, self._obj.__class__)
+            return cls.py__getitem__all_values()
+
+        try:
+            getitem = self._obj.__getitem__
+        except AttributeError:
+            pass
+        else:
+            annotation = DirectObjectAccess(self._inference_state, getitem).get_return_annotation()
+            if annotation is not None:
+                return [annotation]
+        return None
+
+    def py__simple_getitem__(self, index, *, safe=True):
+        if safe and type(self._obj) not in ALLOWED_GETITEM_TYPES:
             # Get rid of side effects, we won't call custom `__getitem__`s.
             return None
 
         return self._create_access_path(self._obj[index])
 
     def py__iter__list(self):
-        if not hasattr(self._obj, '__getitem__'):
+        try:
+            iter_method = self._obj.__iter__
+        except AttributeError:
             return None
+        else:
+            p = DirectObjectAccess(self._inference_state, iter_method).get_return_annotation()
+            if p is not None:
+                return [p]
 
         if type(self._obj) not in ALLOWED_GETITEM_TYPES:
             # Get rid of side effects, we won't call custom `__getitem__`s.
@@ -306,33 +329,37 @@ class DirectObjectAccess:
         except TypeError:
             return False
 
-    def is_allowed_getattr(self, name, unsafe=False):
+    def is_allowed_getattr(self, name, safe=True) -> Tuple[bool, bool, Optional[AccessPath]]:
         # TODO this API is ugly.
-        if unsafe:
-            # Unsafe is mostly used to check for __getattr__/__getattribute__.
-            # getattr_static works for properties, but the underscore methods
-            # are just ignored (because it's safer and avoids more code
-            # execution). See also GH #1378.
-
-            # Avoid warnings, see comment in the next function.
-            with warnings.catch_warnings(record=True):
-                warnings.simplefilter("always")
-                try:
-                    return hasattr(self._obj, name), False
-                except Exception:
-                    # Obviously has an attribute (propably a property) that
-                    # gets executed, so just avoid all exceptions here.
-                    return False, False
         try:
             attr, is_get_descriptor = getattr_static(self._obj, name)
         except AttributeError:
-            return False, False
+            if not safe:
+                # Unsafe is mostly used to check for __getattr__/__getattribute__.
+                # getattr_static works for properties, but the underscore methods
+                # are just ignored (because it's safer and avoids more code
+                # execution). See also GH #1378.
+
+                # Avoid warnings, see comment in the next function.
+                with warnings.catch_warnings(record=True):
+                    warnings.simplefilter("always")
+                    try:
+                        return hasattr(self._obj, name), False, None
+                    except Exception:
+                        # Obviously has an attribute (probably a property) that
+                        # gets executed, so just avoid all exceptions here.
+                        pass
+            return False, False, None
         else:
             if is_get_descriptor and type(attr) not in ALLOWED_DESCRIPTOR_ACCESS:
+                if isinstance(attr, property):
+                    if hasattr(attr.fget, '__annotations__'):
+                        a = DirectObjectAccess(self._inference_state, attr.fget)
+                        return True, True, a.get_return_annotation()
                 # In case of descriptors that have get methods we cannot return
                 # it's value, because that would mean code execution.
-                return True, True
-        return True, False
+                return True, True, None
+        return True, False, None
 
     def getattr_paths(self, name, default=_sentinel):
         try:
@@ -361,7 +388,7 @@ class DirectObjectAccess:
         except AttributeError:
             pass
         else:
-            if module is not None:
+            if module is not None and isinstance(module, str):
                 try:
                     __import__(module)
                     # For some modules like _sqlite3, the __module__ for classes is
@@ -492,7 +519,7 @@ class DirectObjectAccess:
             # the signature. In that case we just want a simple escape for now.
             raise ValueError
 
-    def get_return_annotation(self):
+    def get_return_annotation(self) -> Optional[AccessPath]:
         try:
             o = self._obj.__annotations__.get('return')
         except AttributeError:

jedi/inference/compiled/mixed.py

@@ -34,7 +34,7 @@ class MixedObject(ValueWrapper):
 
     This combined logic makes it possible to provide more powerful REPL
     completion. It allows side effects that are not noticable with the default
-    parser structure to still be completeable.
+    parser structure to still be completable.
 
     The biggest difference from CompiledValue to MixedObject is that we are
     generally dealing with Python code and not with C code. This will generate
@@ -142,9 +142,9 @@ class MixedObjectFilter(compiled.CompiledValueFilter):
         super().__init__(inference_state, compiled_value)
         self._tree_value = tree_value
 
-    def _create_name(self, name):
+    def _create_name(self, *args, **kwargs):
         return MixedName(
-            super()._create_name(name),
+            super()._create_name(*args, **kwargs),
             self._tree_value,
         )
 
@@ -187,7 +187,7 @@ def _find_syntax_node_name(inference_state, python_object):
     try:
         python_object = _get_object_to_check(python_object)
         path = inspect.getsourcefile(python_object)
-    except TypeError:
+    except (OSError, TypeError):
         # The type might not be known (e.g. class_with_dict.__weakref__)
         return None
     path = None if path is None else Path(path)
@@ -267,7 +267,7 @@ def _find_syntax_node_name(inference_state, python_object):
 @inference_state_function_cache()
 def _create(inference_state, compiled_value, module_context):
     # TODO accessing this is bad, but it probably doesn't matter that much,
-    # because we're working with interpreteters only here.
+    # because we're working with interpreters only here.
     python_object = compiled_value.access_handle.access._obj
     result = _find_syntax_node_name(inference_state, python_object)
     if result is None:

jedi/inference/compiled/subprocess/__init__.py

@@ -5,8 +5,26 @@ goals:
 1. Making it safer - Segfaults and RuntimeErrors as well as stdout/stderr can
    be ignored and dealt with.
 2. Make it possible to handle different Python versions as well as virtualenvs.
+
+The architecture here is briefly:
+ - For each Jedi `Environment` there is a corresponding subprocess which
+   operates within the target environment. If the subprocess dies it is replaced
+   at this level.
+ - `CompiledSubprocess` manages exactly one subprocess and handles communication
+   from the parent side.
+ - `Listener` runs within the subprocess, processing each request and yielding
+   results.
+ - `InterpreterEnvironment` provides an API which matches that of `Environment`,
+   but runs functionality inline rather than within a subprocess. It is thus
+   used both directly in places where a subprocess is unnecessary and/or
+   undesirable and also within subprocesses themselves.
+ - `InferenceStateSubprocess` (or `InferenceStateSameProcess`) provide high
+   level access to functionality within the subprocess from within the parent.
+   Each `InterpreterState` has an instance of one of these, provided by its
+   environment.
 """
 
+import collections
 import os
 import sys
 import queue
@@ -15,6 +33,7 @@ import traceback
 import weakref
 from functools import partial
 from threading import Thread
+from typing import Dict, TYPE_CHECKING
 
 from jedi._compatibility import pickle_dump, pickle_load
 from jedi import debug
@@ -24,6 +43,9 @@ from jedi.inference.compiled.access import DirectObjectAccess, AccessPath, \
     SignatureParam
 from jedi.api.exceptions import InternalError
 
+if TYPE_CHECKING:
+    from jedi.inference import InferenceState
+
 
 _MAIN_PATH = os.path.join(os.path.dirname(__file__), '__main__.py')
 PICKLE_PROTOCOL = 4
@@ -82,10 +104,9 @@ def _cleanup_process(process, thread):
 
 
 class _InferenceStateProcess:
-    def __init__(self, inference_state):
+    def __init__(self, inference_state: 'InferenceState') -> None:
         self._inference_state_weakref = weakref.ref(inference_state)
-        self._inference_state_id = id(inference_state)
-        self._handles = {}
+        self._handles: Dict[int, AccessHandle] = {}
 
     def get_or_create_access_handle(self, obj):
         id_ = id(obj)
@@ -115,11 +136,49 @@ class InferenceStateSameProcess(_InferenceStateProcess):
 
 
 class InferenceStateSubprocess(_InferenceStateProcess):
-    def __init__(self, inference_state, compiled_subprocess):
+    """
+    API to functionality which will run in a subprocess.
+
+    This mediates the interaction between an `InferenceState` and the actual
+    execution of functionality running within a `CompiledSubprocess`. Available
+    functions are defined in `.functions`, though should be accessed via
+    attributes on this class of the same name.
+
+    This class is responsible for indicating that the `InferenceState` within
+    the subprocess can be removed once the corresponding instance in the parent
+    goes away.
+    """
+
+    def __init__(
+        self,
+        inference_state: 'InferenceState',
+        compiled_subprocess: 'CompiledSubprocess',
+    ) -> None:
         super().__init__(inference_state)
         self._used = False
         self._compiled_subprocess = compiled_subprocess
 
+        # Opaque id we'll pass to the subprocess to identify the context (an
+        # `InferenceState`) which should be used for the request. This allows us
+        # to make subsequent requests which operate on results from previous
+        # ones, while keeping a single subprocess which can work with several
+        # contexts in the parent process. Once it is no longer needed(i.e: when
+        # this class goes away), we also use this id to indicate that the
+        # subprocess can discard the context.
+        #
+        # Note: this id is deliberately coupled to this class (and not to
+        # `InferenceState`) as this class manages access handle mappings which
+        # must correspond to those in the subprocess. This approach also avoids
+        # race conditions from successive `InferenceState`s with the same object
+        # id (as observed while adding support for Python 3.13).
+        #
+        # This value does not need to be the `id()` of this instance, we merely
+        # need to ensure that it enables the (visible) lifetime of the context
+        # within the subprocess to match that of this class. We therefore also
+        # depend on the semantics of `CompiledSubprocess.delete_inference_state`
+        # for correctness.
+        self._inference_state_id = id(self)
+
     def __getattr__(self, name):
         func = _get_function(name)
 
@@ -127,7 +186,7 @@ class InferenceStateSubprocess(_InferenceStateProcess):
             self._used = True
 
             result = self._compiled_subprocess.run(
-                self._inference_state_weakref(),
+                self._inference_state_id,
                 func,
                 args=args,
                 kwargs=kwargs,
@@ -163,12 +222,23 @@ class InferenceStateSubprocess(_InferenceStateProcess):
 
 
 class CompiledSubprocess:
+    """
+    A subprocess which runs inference within a target environment.
+
+    This class manages the interface to a single instance of such a process as
+    well as the lifecycle of the process itself. See `.__main__` and `Listener`
+    for the implementation of the subprocess and details of the protocol.
+
+    A single live instance of this is maintained by `jedi.api.environment.Environment`,
+    so that typically a single subprocess is used at a time.
+    """
+
     is_crashed = False
 
     def __init__(self, executable, env_vars=None):
         self._executable = executable
         self._env_vars = env_vars
-        self._inference_state_deletion_queue = queue.deque()
+        self._inference_state_deletion_queue = collections.deque()
         self._cleanup_callable = lambda: None
 
     def __repr__(self):
@@ -212,18 +282,18 @@ class CompiledSubprocess:
                                                   t)
         return process
 
-    def run(self, inference_state, function, args=(), kwargs={}):
+    def run(self, inference_state_id, function, args=(), kwargs={}):
         # Delete old inference_states.
         while True:
             try:
-                inference_state_id = self._inference_state_deletion_queue.pop()
+                delete_id = self._inference_state_deletion_queue.pop()
             except IndexError:
                 break
             else:
-                self._send(inference_state_id, None)
+                self._send(delete_id, None)
 
         assert callable(function)
-        return self._send(id(inference_state), function, args, kwargs)
+        return self._send(inference_state_id, function, args, kwargs)
 
     def get_sys_path(self):
         return self._send(None, functions.get_sys_path, (), {})
@@ -271,21 +341,65 @@ class CompiledSubprocess:
 
     def delete_inference_state(self, inference_state_id):
         """
-        Currently we are not deleting inference_state instantly. They only get
-        deleted once the subprocess is used again. It would probably a better
-        solution to move all of this into a thread. However, the memory usage
-        of a single inference_state shouldn't be that high.
+        Indicate that an inference state (in the subprocess) is no longer
+        needed.
+
+        The state corresponding to the given id will become inaccessible and the
+        id may safely be re-used to refer to a different context.
+
+        Note: it is not guaranteed that the corresponding state will actually be
+        deleted immediately.
         """
-        # With an argument - the inference_state gets deleted.
+        # Warning: if changing the semantics of context deletion see the comment
+        # in `InferenceStateSubprocess.__init__` regarding potential race
+        # conditions.
+
+        # Currently we are not deleting the related state instantly. They only
+        # get deleted once the subprocess is used again. It would probably a
+        # better solution to move all of this into a thread. However, the memory
+        # usage of a single inference_state shouldn't be that high.
         self._inference_state_deletion_queue.append(inference_state_id)
 
 
 class Listener:
+    """
+    Main loop for the subprocess which actually does the inference.
+
+    This class runs within the target environment. It listens to instructions
+    from the parent process, runs inference and returns the results.
+
+    The subprocess has a long lifetime and is expected to process several
+    requests, including for different `InferenceState` instances in the parent.
+    See `CompiledSubprocess` for the parent half of the system.
+
+    Communication is via pickled data sent serially over stdin and stdout.
+    Stderr is read only if the child process crashes.
+
+    The request protocol is a 4-tuple of:
+     * inference_state_id | None: an opaque identifier of the parent's
+       `InferenceState`. An `InferenceState` operating over an
+       `InterpreterEnvironment` is created within this process for each of
+       these, ensuring that each parent context has a corresponding context
+       here. This allows context to be persisted between requests. Unless
+       `None`, the local `InferenceState` will be passed to the given function
+       as the first positional argument.
+     * function | None: the function to run. This is expected to be a member of
+       `.functions`. `None` indicates that the corresponding inference state is
+       no longer needed and should be dropped.
+     * args: positional arguments to the `function`. If any of these are
+       `AccessHandle` instances they will be adapted to the local
+       `InferenceState` before being passed.
+     * kwargs: keyword arguments to the `function`. If any of these are
+       `AccessHandle` instances they will be adapted to the local
+       `InferenceState` before being passed.
+
+    The result protocol is a 3-tuple of either:
+     * (False, None, function result): if the function returns without error, or
+     * (True, traceback, exception): if the function raises an exception
+    """
+
     def __init__(self):
         self._inference_states = {}
-        # TODO refactor so we don't need to process anymore just handle
-        # controlling.
-        self._process = _InferenceStateProcess(Listener)
 
     def _get_inference_state(self, function, inference_state_id):
         from jedi.inference import InferenceState
@@ -307,6 +421,9 @@ class Listener:
         if inference_state_id is None:
             return function(*args, **kwargs)
         elif function is None:
+            # Warning: if changing the semantics of context deletion see the comment
+            # in `InferenceStateSubprocess.__init__` regarding potential race
+            # conditions.
             del self._inference_states[inference_state_id]
         else:
             inference_state = self._get_inference_state(function, inference_state_id)
@@ -347,7 +464,12 @@ class Listener:
 
 
 class AccessHandle:
-    def __init__(self, subprocess, access, id_):
+    def __init__(
+        self,
+        subprocess: _InferenceStateProcess,
+        access: DirectObjectAccess,
+        id_: int,
+    ) -> None:
         self.access = access
         self._subprocess = subprocess
         self.id = id_

jedi/inference/compiled/subprocess/__main__.py

@@ -21,11 +21,11 @@ class _ExactImporter(MetaPathFinder):
     def __init__(self, path_dct):
         self._path_dct = path_dct
 
-    def find_module(self, fullname, path=None):
+    def find_spec(self, fullname, path=None, target=None):
         if path is None and fullname in self._path_dct:
             p = self._path_dct[fullname]
-            loader = PathFinder.find_module(fullname, path=[p])
-            return loader
+            spec = PathFinder.find_spec(fullname, path=[p], target=target)
+            return spec
         return None
 
 

jedi/inference/compiled/subprocess/functions.py

@@ -2,12 +2,11 @@ import sys
 import os
 import inspect
 import importlib
-import warnings
 from pathlib import Path
-from zipimport import zipimporter
+from zipfile import ZipFile
+from zipimport import zipimporter, ZipImportError
 from importlib.machinery import all_suffixes
 
-from jedi._compatibility import cast_path
 from jedi.inference.compiled import access
 from jedi import debug
 from jedi import parser_utils
@@ -15,7 +14,7 @@ from jedi.file_io import KnownContentFileIO, ZipFileIO
 
 
 def get_sys_path():
-    return list(map(cast_path, sys.path))
+    return sys.path
 
 
 def load_module(inference_state, **kwargs):
@@ -93,15 +92,22 @@ def _iter_module_names(inference_state, paths):
     # Python modules/packages
     for path in paths:
         try:
-            dirs = os.scandir(path)
+            dir_entries = ((entry.name, entry.is_dir()) for entry in os.scandir(path))
         except OSError:
-            # The file might not exist or reading it might lead to an error.
-            debug.warning("Not possible to list directory: %s", path)
-            continue
-        for dir_entry in dirs:
-            name = dir_entry.name
+            try:
+                zip_import_info = zipimporter(path)
+                # Unfortunately, there is no public way to access zipimporter's
+                # private _files member. We therefore have to use a
+                # custom function to iterate over the files.
+                dir_entries = _zip_list_subdirectory(
+                    zip_import_info.archive, zip_import_info.prefix)
+            except ZipImportError:
+                # The file might not exist or reading it might lead to an error.
+                debug.warning("Not possible to list directory: %s", path)
+                continue
+        for name, is_dir in dir_entries:
             # First Namespaces then modules/stubs
-            if dir_entry.is_dir():
+            if is_dir:
                 # pycache is obviously not an interesting namespace. Also the
                 # name must be a valid identifier.
                 if name != '__pycache__' and name.isidentifier():
@@ -144,7 +150,11 @@ def _find_module(string, path=None, full_name=None, is_global_search=True):
 
         spec = find_spec(string, p)
         if spec is not None:
+            if spec.origin == "frozen":
+                continue
+
             loader = spec.loader
+
             if loader is None and not spec.has_location:
                 # This is a namespace package.
                 full_name = string if not path else full_name
@@ -156,17 +166,16 @@ def _find_module(string, path=None, full_name=None, is_global_search=True):
 
 
 def _find_module_py33(string, path=None, loader=None, full_name=None, is_global_search=True):
-    loader = loader or importlib.machinery.PathFinder.find_module(string, path)
+    if not loader:
+        spec = importlib.machinery.PathFinder.find_spec(string, path)
+        if spec is not None:
+            loader = spec.loader
 
     if loader is None and path is None:  # Fallback to find builtins
         try:
-            with warnings.catch_warnings(record=True):
-                # Mute "DeprecationWarning: Use importlib.util.find_spec()
-                # instead." While we should replace that in the future, it's
-                # probably good to wait until we deprecate Python 3.3, since
-                # it was added in Python 3.4 and find_loader hasn't been
-                # removed in 3.6.
-                loader = importlib.find_loader(string)
+            spec = importlib.util.find_spec(string)
+            if spec is not None:
+                loader = spec.loader
         except ValueError as e:
             # See #491. Importlib might raise a ValueError, to avoid this, we
             # just raise an ImportError to fix the issue.
@@ -190,7 +199,7 @@ def _from_loader(loader, string):
     except AttributeError:
         return None, is_package
     else:
-        module_path = cast_path(get_filename(string))
+        module_path = get_filename(string)
 
     # To avoid unicode and read bytes, "overwrite" loader.get_source if
     # possible.
@@ -212,7 +221,7 @@ def _from_loader(loader, string):
     if code is None:
         return None, is_package
     if isinstance(loader, zipimporter):
-        return ZipFileIO(module_path, code, Path(cast_path(loader.archive))), is_package
+        return ZipFileIO(module_path, code, Path(loader.archive)), is_package
 
     return KnownContentFileIO(module_path, code), is_package
 
@@ -230,6 +239,17 @@ def _get_source(loader, fullname):
                           name=fullname)
 
 
+def _zip_list_subdirectory(zip_path, zip_subdir_path):
+    zip_file = ZipFile(zip_path)
+    zip_subdir_path = Path(zip_subdir_path)
+    zip_content_file_paths = zip_file.namelist()
+    for raw_file_name in zip_content_file_paths:
+        file_path = Path(raw_file_name)
+        if file_path.parent == zip_subdir_path:
+            file_path = file_path.relative_to(zip_subdir_path)
+            yield file_path.name, raw_file_name.endswith("/")
+
+
 class ImplicitNSInfo:
     """Stores information returned from an implicit namespace spec"""
     def __init__(self, name, paths):

jedi/inference/compiled/value.py

@@ -5,10 +5,10 @@ import re
 from functools import partial
 from inspect import Parameter
 from pathlib import Path
+from typing import Optional
 
 from jedi import debug
 from jedi.inference.utils import to_list
-from jedi._compatibility import cast_path
 from jedi.cache import memoize_method
 from jedi.inference.filters import AbstractFilter
 from jedi.inference.names import AbstractNameDefinition, ValueNameMixin, \
@@ -51,7 +51,6 @@ class CompiledValue(Value):
     def py__call__(self, arguments):
         return_annotation = self.access_handle.get_return_annotation()
         if return_annotation is not None:
-            # TODO the return annotation may also be a string.
             return create_from_access_path(
                 self.inference_state,
                 return_annotation
@@ -163,11 +162,14 @@ class CompiledValue(Value):
     def py__simple_getitem__(self, index):
         with reraise_getitem_errors(IndexError, KeyError, TypeError):
             try:
-                access = self.access_handle.py__simple_getitem__(index)
+                access = self.access_handle.py__simple_getitem__(
+                    index,
+                    safe=not self.inference_state.allow_unsafe_executions
+                )
             except AttributeError:
                 return super().py__simple_getitem__(index)
         if access is None:
-            return NO_VALUES
+            return super().py__simple_getitem__(index)
 
         return ValueSet([create_from_access_path(self.inference_state, access)])
 
@@ -293,10 +295,7 @@ class CompiledModule(CompiledValue):
         return CompiledModuleContext(self)
 
     def py__path__(self):
-        paths = self.access_handle.py__path__()
-        if paths is None:
-            return None
-        return map(cast_path, paths)
+        return self.access_handle.py__path__()
 
     def is_package(self):
         return self.py__path__() is not None
@@ -309,19 +308,17 @@ class CompiledModule(CompiledValue):
             return ()
         return tuple(name.split('.'))
 
-    def py__file__(self):
-        path = cast_path(self.access_handle.py__file__())
-        if path is None:
-            return None
-        return Path(path)
+    def py__file__(self) -> Optional[Path]:
+        return self.access_handle.py__file__()  # type: ignore[no-any-return]
 
 
 class CompiledName(AbstractNameDefinition):
-    def __init__(self, inference_state, parent_value, name):
+    def __init__(self, inference_state, parent_value, name, is_descriptor):
         self._inference_state = inference_state
         self.parent_context = parent_value.as_context()
         self._parent_value = parent_value
         self.string_name = name
+        self.is_descriptor = is_descriptor
 
     def py__doc__(self):
         return self.infer_compiled_value().py__doc__()
@@ -348,6 +345,11 @@ class CompiledName(AbstractNameDefinition):
 
     @property
     def api_type(self):
+        if self.is_descriptor:
+            # In case of properties we want to avoid executions as much as
+            # possible. Since the api_type can be wrong for other reasons
+            # anyway, we just return instance here.
+            return "instance"
         return self.infer_compiled_value().api_type
 
     def infer(self):
@@ -438,9 +440,10 @@ class CompiledValueFilter(AbstractFilter):
 
     def get(self, name):
         access_handle = self.compiled_value.access_handle
+        safe = not self._inference_state.allow_unsafe_executions
         return self._get(
             name,
-            lambda name, unsafe: access_handle.is_allowed_getattr(name, unsafe),
+            lambda name: access_handle.is_allowed_getattr(name, safe=safe),
             lambda name: name in access_handle.dir(),
             check_has_attribute=True
         )
@@ -449,36 +452,40 @@ class CompiledValueFilter(AbstractFilter):
         """
         To remove quite a few access calls we introduced the callback here.
         """
-        if self._inference_state.allow_descriptor_getattr:
-            pass
-
-        has_attribute, is_descriptor = allowed_getattr_callback(
+        has_attribute, is_descriptor, property_return_annotation = allowed_getattr_callback(
             name,
-            unsafe=self._inference_state.allow_descriptor_getattr
         )
+        if property_return_annotation is not None:
+            values = create_from_access_path(
+                self._inference_state,
+                property_return_annotation
+            ).execute_annotation()
+            if values:
+                return [CompiledValueName(v, name) for v in values]
+
         if check_has_attribute and not has_attribute:
             return []
 
         if (is_descriptor or not has_attribute) \
-                and not self._inference_state.allow_descriptor_getattr:
+                and not self._inference_state.allow_unsafe_executions:
             return [self._get_cached_name(name, is_empty=True)]
 
         if self.is_instance and not in_dir_callback(name):
             return []
-        return [self._get_cached_name(name)]
+        return [self._get_cached_name(name, is_descriptor=is_descriptor)]
 
     @memoize_method
-    def _get_cached_name(self, name, is_empty=False):
+    def _get_cached_name(self, name, is_empty=False, *, is_descriptor=False):
         if is_empty:
             return EmptyCompiledName(self._inference_state, name)
         else:
-            return self._create_name(name)
+            return self._create_name(name, is_descriptor=is_descriptor)
 
     def values(self):
         from jedi.inference.compiled import builtin_from_name
         names = []
         needs_type_completions, dir_infos = self.compiled_value.access_handle.get_dir_infos()
-        # We could use `unsafe` here as well, especially as a parameter to
+        # We could use `safe=False` here as well, especially as a parameter to
         # get_dir_infos. But this would lead to a lot of property executions
         # that are probably not wanted. The drawback for this is that we
         # have a different name for `get` and `values`. For `get` we always
@@ -486,7 +493,7 @@ class CompiledValueFilter(AbstractFilter):
         for name in dir_infos:
             names += self._get(
                 name,
-                lambda name, unsafe: dir_infos[name],
+                lambda name: dir_infos[name],
                 lambda name: name in dir_infos,
             )
 
@@ -496,11 +503,12 @@ class CompiledValueFilter(AbstractFilter):
                 names += filter.values()
         return names
 
-    def _create_name(self, name):
+    def _create_name(self, name, is_descriptor):
         return CompiledName(
             self._inference_state,
             self.compiled_value,
-            name
+            name,
+            is_descriptor,
         )
 
     def __repr__(self):

jedi/inference/context.py

@@ -1,5 +1,7 @@
 from abc import abstractmethod
 from contextlib import contextmanager
+from pathlib import Path
+from typing import Optional
 
 from parso.tree import search_ancestor
 from parso.python.tree import Name
@@ -307,8 +309,8 @@ class FunctionContext(TreeContextMixin, ValueContext):
 
 
 class ModuleContext(TreeContextMixin, ValueContext):
-    def py__file__(self):
-        return self._value.py__file__()
+    def py__file__(self) -> Optional[Path]:
+        return self._value.py__file__()  # type: ignore[no-any-return]
 
     def get_filters(self, until_position=None, origin_scope=None):
         filters = self._value.get_filters(origin_scope)
@@ -325,7 +327,7 @@ class ModuleContext(TreeContextMixin, ValueContext):
         yield from filters
 
     def get_global_filter(self):
-        return GlobalNameFilter(self, self.tree_node)
+        return GlobalNameFilter(self)
 
     @property
     def string_names(self):
@@ -355,8 +357,8 @@ class NamespaceContext(TreeContextMixin, ValueContext):
     def string_names(self):
         return self._value.string_names
 
-    def py__file__(self):
-        return self._value.py__file__()
+    def py__file__(self) -> Optional[Path]:
+        return self._value.py__file__()  # type: ignore[no-any-return]
 
 
 class ClassContext(TreeContextMixin, ValueContext):
@@ -405,8 +407,8 @@ class CompiledModuleContext(CompiledContext):
     def string_names(self):
         return self._value.string_names
 
-    def py__file__(self):
-        return self._value.py__file__()
+    def py__file__(self) -> Optional[Path]:
+        return self._value.py__file__()  # type: ignore[no-any-return]
 
 
 def _get_global_filters_for_name(context, name_or_none, position):

jedi/inference/docstring_utils.py

@@ -0,0 +1,21 @@
+from jedi.inference.value import ModuleValue
+from jedi.inference.context import ModuleContext
+
+
+class DocstringModule(ModuleValue):
+    def __init__(self, in_module_context, **kwargs):
+        super().__init__(**kwargs)
+        self._in_module_context = in_module_context
+
+    def _as_context(self):
+        return DocstringModuleContext(self, self._in_module_context)
+
+
+class DocstringModuleContext(ModuleContext):
+    def __init__(self, module_value, in_module_context):
+        super().__init__(module_value)
+        self._in_module_context = in_module_context
+
+    def get_filters(self, origin_scope=None, until_position=None):
+        yield from super().get_filters(until_position=until_position)
+        yield from self._in_module_context.get_filters()

jedi/inference/docstrings.py

@@ -17,12 +17,10 @@ annotations.
 
 import re
 import warnings
-from textwrap import dedent
 
 from parso import parse, ParserSyntaxError
 
 from jedi import debug
-from jedi.common import indent_block
 from jedi.inference.cache import inference_state_method_cache
 from jedi.inference.base_value import iterator_to_value_set, ValueSet, \
     NO_VALUES
@@ -182,52 +180,40 @@ def _strip_rst_role(type_str):
 
 
 def _infer_for_statement_string(module_context, string):
-    code = dedent("""
-    def pseudo_docstring_stuff():
-        '''
-        Create a pseudo function for docstring statements.
-        Need this docstring so that if the below part is not valid Python this
-        is still a function.
-        '''
-    {}
-    """)
     if string is None:
         return []
 
-    for element in re.findall(r'((?:\w+\.)*\w+)\.', string):
-        # Try to import module part in dotted name.
-        # (e.g., 'threading' in 'threading.Thread').
-        string = 'import %s\n' % element + string
+    potential_imports = re.findall(r'((?:\w+\.)*\w+)\.', string)
+    # Try to import module part in dotted name.
+    # (e.g., 'threading' in 'threading.Thread').
+    imports = "\n".join(f"import {p}" for p in potential_imports)
+    string = f'{imports}\n{string}'
 
     debug.dbg('Parse docstring code %s', string, color='BLUE')
     grammar = module_context.inference_state.grammar
     try:
-        module = grammar.parse(code.format(indent_block(string)), error_recovery=False)
+        module = grammar.parse(string, error_recovery=False)
     except ParserSyntaxError:
         return []
     try:
-        funcdef = next(module.iter_funcdefs())
-        # First pick suite, then simple_stmt and then the node,
-        # which is also not the last item, because there's a newline.
-        stmt = funcdef.children[-1].children[-1].children[-2]
+        # It's not the last item, because that's an end marker.
+        stmt = module.children[-2]
     except (AttributeError, IndexError):
         return []
 
     if stmt.type not in ('name', 'atom', 'atom_expr'):
         return []
 
-    from jedi.inference.value import FunctionValue
-    function_value = FunctionValue(
-        module_context.inference_state,
-        module_context,
-        funcdef
+    # Here we basically use a fake module that also uses the filters in
+    # the actual module.
+    from jedi.inference.docstring_utils import DocstringModule
+    m = DocstringModule(
+        in_module_context=module_context,
+        inference_state=module_context.inference_state,
+        module_node=module,
+        code_lines=[],
     )
-    func_execution_context = function_value.as_context()
-    # Use the module of the param.
-    # TODO this module is not the module of the param in case of a function
-    # call. In that case it's the module of the function call.
-    # stuffed with content from a function call.
-    return list(_execute_types_in_stmt(func_execution_context, stmt))
+    return list(_execute_types_in_stmt(m.as_context(), stmt))
 
 
 def _execute_types_in_stmt(module_context, stmt):

jedi/inference/dynamic_params.py

@@ -66,11 +66,11 @@ def dynamic_param_lookup(function_value, param_index):
     have to look for all calls to ``func`` to find out what ``foo`` possibly
     is.
     """
-    funcdef = function_value.tree_node
-
-    if not settings.dynamic_params:
+    if not function_value.inference_state.do_dynamic_params_search:
         return NO_VALUES
 
+    funcdef = function_value.tree_node
+
     path = function_value.get_root_context().py__file__()
     if path is not None and is_stdlib_path(path):
         # We don't want to search for references in the stdlib. Usually people

jedi/inference/filters.py

@@ -12,7 +12,7 @@ from parso.python.tree import Name, UsedNamesMapping
 from jedi.inference import flow_analysis
 from jedi.inference.base_value import ValueSet, ValueWrapper, \
     LazyValueWrapper
-from jedi.parser_utils import get_cached_parent_scope
+from jedi.parser_utils import get_cached_parent_scope, get_parso_cache_node
 from jedi.inference.utils import to_list
 from jedi.inference.names import TreeNameDefinition, ParamName, \
     AnonymousParamName, AbstractNameDefinition, NameWrapper
@@ -54,11 +54,15 @@ class FilterWrapper:
         return self.wrap_names(self._wrapped_filter.values())
 
 
-def _get_definition_names(used_names, name_key):
+def _get_definition_names(parso_cache_node, used_names, name_key):
+    if parso_cache_node is None:
+        names = used_names.get(name_key, ())
+        return tuple(name for name in names if name.is_definition(include_setitem=True))
+
     try:
-        for_module = _definition_name_cache[used_names]
+        for_module = _definition_name_cache[parso_cache_node]
     except KeyError:
-        for_module = _definition_name_cache[used_names] = {}
+        for_module = _definition_name_cache[parso_cache_node] = {}
 
     try:
         return for_module[name_key]
@@ -70,18 +74,40 @@ def _get_definition_names(used_names, name_key):
         return result
 
 
-class AbstractUsedNamesFilter(AbstractFilter):
+class _AbstractUsedNamesFilter(AbstractFilter):
     name_class = TreeNameDefinition
 
-    def __init__(self, parent_context, parser_scope):
-        self._parser_scope = parser_scope
-        self._module_node = self._parser_scope.get_root_node()
-        self._used_names = self._module_node.get_used_names()
+    def __init__(self, parent_context, node_context=None):
+        if node_context is None:
+            node_context = parent_context
+        self._node_context = node_context
+        self._parser_scope = node_context.tree_node
+        module_context = node_context.get_root_context()
+        # It is quite hacky that we have to use that. This is for caching
+        # certain things with a WeakKeyDictionary. However, parso intentionally
+        # uses slots (to save memory) and therefore we end up with having to
+        # have a weak reference to the object that caches the tree.
+        #
+        # Previously we have tried to solve this by using a weak reference onto
+        # used_names. However that also does not work, because it has a
+        # reference from the module, which itself is referenced by any node
+        # through parents.
+        path = module_context.py__file__()
+        if path is None:
+            # If the path is None, there is no guarantee that parso caches it.
+            self._parso_cache_node = None
+        else:
+            self._parso_cache_node = get_parso_cache_node(
+                module_context.inference_state.latest_grammar
+                if module_context.is_stub() else module_context.inference_state.grammar,
+                path
+            )
+        self._used_names = module_context.tree_node.get_used_names()
         self.parent_context = parent_context
 
     def get(self, name):
         return self._convert_names(self._filter(
-            _get_definition_names(self._used_names, name),
+            _get_definition_names(self._parso_cache_node, self._used_names, name),
         ))
 
     def _convert_names(self, names):
@@ -92,7 +118,7 @@ class AbstractUsedNamesFilter(AbstractFilter):
             name
             for name_key in self._used_names
             for name in self._filter(
-                _get_definition_names(self._used_names, name_key),
+                _get_definition_names(self._parso_cache_node, self._used_names, name_key),
             )
         )
 
@@ -100,7 +126,7 @@ class AbstractUsedNamesFilter(AbstractFilter):
         return '<%s: %s>' % (self.__class__.__name__, self.parent_context)
 
 
-class ParserTreeFilter(AbstractUsedNamesFilter):
+class ParserTreeFilter(_AbstractUsedNamesFilter):
     def __init__(self, parent_context, node_context=None, until_position=None,
                  origin_scope=None):
         """
@@ -109,10 +135,7 @@ class ParserTreeFilter(AbstractUsedNamesFilter):
         value, but for some type inference it's important to have a local
         value of the other classes.
         """
-        if node_context is None:
-            node_context = parent_context
-        super().__init__(parent_context, node_context.tree_node)
-        self._node_context = node_context
+        super().__init__(parent_context, node_context)
         self._origin_scope = origin_scope
         self._until_position = until_position
 
@@ -126,7 +149,7 @@ class ParserTreeFilter(AbstractUsedNamesFilter):
         if parent.type == 'trailer':
             return False
         base_node = parent if parent.type in ('classdef', 'funcdef') else name
-        return get_cached_parent_scope(self._used_names, base_node) == self._parser_scope
+        return get_cached_parent_scope(self._parso_cache_node, base_node) == self._parser_scope
 
     def _check_flows(self, names):
         for name in sorted(names, key=lambda name: name.start_pos, reverse=True):
@@ -182,7 +205,7 @@ class AnonymousFunctionExecutionFilter(_FunctionExecutionFilter):
         return AnonymousParamName(self._function_value, name)
 
 
-class GlobalNameFilter(AbstractUsedNamesFilter):
+class GlobalNameFilter(_AbstractUsedNamesFilter):
     def get(self, name):
         try:
             names = self._used_names[name]

jedi/inference/gradual/annotation.py

@@ -196,13 +196,43 @@ def py__annotations__(funcdef):
     return dct
 
 
+def resolve_forward_references(context, all_annotations):
+    def resolve(node):
+        if node is None or node.type != 'string':
+            return node
+
+        node = _get_forward_reference_node(
+            context,
+            context.inference_state.compiled_subprocess.safe_literal_eval(
+                node.value,
+            ),
+        )
+
+        if node is None:
+            # There was a string, but it's not a valid annotation
+            return None
+
+        # The forward reference tree has an additional root node ('eval_input')
+        # that we don't want. Extract the node we do want, that is equivalent to
+        # the nodes returned by `py__annotations__` for a non-quoted node.
+        node = node.children[0]
+
+        return node
+
+    return {name: resolve(node) for name, node in all_annotations.items()}
+
+
 @inference_state_method_cache()
 def infer_return_types(function, arguments):
     """
     Infers the type of a function's return value,
     according to type annotations.
     """
-    all_annotations = py__annotations__(function.tree_node)
+    context = function.get_default_param_context()
+    all_annotations = resolve_forward_references(
+        context,
+        py__annotations__(function.tree_node),
+    )
     annotation = all_annotations.get("return", None)
     if annotation is None:
         # If there is no Python 3-type annotation, look for an annotation
@@ -217,11 +247,10 @@ def infer_return_types(function, arguments):
             return NO_VALUES
 
         return _infer_annotation_string(
-            function.get_default_param_context(),
+            context,
             match.group(1).strip()
         ).execute_annotation()
 
-    context = function.get_default_param_context()
     unknown_type_vars = find_unknown_type_vars(context, annotation)
     annotation_values = infer_annotation(context, annotation)
     if not unknown_type_vars:
@@ -373,6 +402,10 @@ def find_type_from_comment_hint_for(context, node, name):
 
 
 def find_type_from_comment_hint_with(context, node, name):
+    if len(node.children) > 4:
+        # In case there are multiple with_items, we do not want a type hint for
+        # now.
+        return []
     assert len(node.children[1].children) == 3, \
         "Can only be here when children[1] is 'foo() as f'"
     varlist = node.children[1].children[2]

jedi/inference/gradual/stub_value.py

@@ -86,6 +86,8 @@ class StubFilter(ParserTreeFilter):
         # Imports in stub files are only public if they have an "as"
         # export.
         definition = name.get_definition()
+        if definition is None:
+            return False
         if definition.type in ('import_from', 'import_name'):
             if name.parent.type not in ('import_as_name', 'dotted_as_name'):
                 return False

jedi/inference/gradual/typeshed.py

@@ -7,7 +7,6 @@ from pathlib import Path
 
 from jedi import settings
 from jedi.file_io import FileIO
-from jedi._compatibility import cast_path
 from jedi.parser_utils import get_cached_code_lines
 from jedi.inference.base_value import ValueSet, NO_VALUES
 from jedi.inference.gradual.stub_value import TypingModuleWrapper, StubModuleValue
@@ -44,7 +43,6 @@ def _create_stub_map(directory_path_info):
             return
 
         for entry in listed:
-            entry = cast_path(entry)
             path = os.path.join(directory_path_info.path, entry)
             if os.path.isdir(path):
                 init = os.path.join(path, '__init__.pyi')
@@ -169,7 +167,6 @@ def _try_to_load_stub(inference_state, import_names, python_value_set,
     if len(import_names) == 1:
         # foo-stubs
         for p in sys_path:
-            p = cast_path(p)
             init = os.path.join(p, *import_names) + '-stubs' + os.path.sep + '__init__.pyi'
             m = _try_to_load_stub_from_file(
                 inference_state,

jedi/inference/gradual/typing.py

@@ -32,7 +32,7 @@ _TYPE_ALIAS_TYPES = {
     'DefaultDict': 'collections.defaultdict',
     'Deque': 'collections.deque',
 }
-_PROXY_TYPES = 'Optional Union ClassVar'.split()
+_PROXY_TYPES = 'Optional Union ClassVar Annotated'.split()
 
 
 class TypingModuleName(NameWrapper):
@@ -113,7 +113,7 @@ class ProxyWithGenerics(BaseTypingClassWithGenerics):
         elif string_name == 'Type':
             # The type is actually already given in the index_value
             return self._generics_manager[0]
-        elif string_name == 'ClassVar':
+        elif string_name in ['ClassVar', 'Annotated']:
             # For now don't do anything here, ClassVars are always used.
             return self._generics_manager[0].execute_annotation()
 
@@ -294,6 +294,9 @@ class Callable(BaseTypingInstance):
             from jedi.inference.gradual.annotation import infer_return_for_callable
             return infer_return_for_callable(arguments, param_values, result_values)
 
+    def py__get__(self, instance, class_value):
+        return ValueSet([self])
+
 
 class Tuple(BaseTypingInstance):
     def _is_homogenous(self):
@@ -431,6 +434,9 @@ class NewType(Value):
         from jedi.inference.compiled.value import CompiledValueName
         return CompiledValueName(self, 'NewType')
 
+    def __repr__(self) -> str:
+        return '<NewType: %s>%s' % (self.tree_node, self._type_value_set)
+
 
 class CastFunction(ValueWrapper):
     @repack_with_argument_clinic('type, object, /')

jedi/inference/imports.py

@@ -422,20 +422,13 @@ def import_module(inference_state, import_names, parent_module_value, sys_path):
             # The module might not be a package.
             return NO_VALUES
 
-        for path in paths:
-            # At the moment we are only using one path. So this is
-            # not important to be correct.
-            if not isinstance(path, list):
-                path = [path]
-            file_io_or_ns, is_pkg = inference_state.compiled_subprocess.get_module_info(
-                string=import_names[-1],
-                path=path,
-                full_name=module_name,
-                is_global_search=False,
-            )
-            if is_pkg is not None:
-                break
-        else:
+        file_io_or_ns, is_pkg = inference_state.compiled_subprocess.get_module_info(
+            string=import_names[-1],
+            path=paths,
+            full_name=module_name,
+            is_global_search=False,
+        )
+        if is_pkg is None:
             return NO_VALUES
 
     if isinstance(file_io_or_ns, ImplicitNSInfo):

jedi/inference/names.py

@@ -248,7 +248,7 @@ class ValueNameMixin:
 
     def get_defining_qualified_value(self):
         context = self.parent_context
-        if context.is_module() or context.is_class():
+        if context is not None and (context.is_module() or context.is_class()):
             return self.parent_context.get_value()  # Might be None
         return None
 
@@ -341,6 +341,12 @@ class TreeNameDefinition(AbstractTreeName):
     def py__doc__(self):
         api_type = self.api_type
         if api_type in ('function', 'class', 'property'):
+            if self.parent_context.get_root_context().is_stub():
+                from jedi.inference.gradual.conversion import convert_names
+                names = convert_names([self], prefer_stub_to_compiled=False)
+                if self not in names:
+                    return _merge_name_docs(names)
+
             # Make sure the names are not TreeNameDefinitions anymore.
             return clean_scope_docstring(self.tree_name.get_definition())
 
@@ -408,6 +414,9 @@ class ParamNameInterface(_ParamMixin):
             return 2
         return 0
 
+    def infer_default(self):
+        return NO_VALUES
+
 
 class BaseTreeParamName(ParamNameInterface, AbstractTreeName):
     annotation_node = None

jedi/inference/recursion.py

@@ -12,7 +12,7 @@ count the function calls.
 Settings
 ~~~~~~~~~~
 
-Recursion settings are important if you don't want extremly
+Recursion settings are important if you don't want extremely
 recursive python code to go absolutely crazy.
 
 The default values are based on experiments while completing the |jedi| library

jedi/inference/references.py

@@ -180,26 +180,34 @@ def _check_fs(inference_state, file_io, regex):
     return m.as_context()
 
 
-def gitignored_lines(folder_io, file_io):
-    ignored_paths = set()
-    ignored_names = set()
+def gitignored_paths(folder_io, file_io):
+    ignored_paths_abs = set()
+    ignored_paths_rel = set()
+
     for l in file_io.read().splitlines():
-        if not l or l.startswith(b'#'):
+        if not l or l.startswith(b'#') or l.startswith(b'!') or b'*' in l:
             continue
 
-        p = l.decode('utf-8', 'ignore')
-        if p.startswith('/'):
-            name = p[1:]
-            if name.endswith(os.path.sep):
-                name = name[:-1]
-            ignored_paths.add(os.path.join(folder_io.path, name))
+        p = l.decode('utf-8', 'ignore').rstrip('/')
+        if '/' in p:
+            name = p.lstrip('/')
+            ignored_paths_abs.add(os.path.join(folder_io.path, name))
         else:
-            ignored_names.add(p)
-    return ignored_paths, ignored_names
+            name = p
+            ignored_paths_rel.add((folder_io.path, name))
+
+    return ignored_paths_abs, ignored_paths_rel
+
+
+def expand_relative_ignore_paths(folder_io, relative_paths):
+    curr_path = folder_io.path
+    return {os.path.join(curr_path, p[1]) for p in relative_paths if curr_path.startswith(p[0])}
 
 
 def recurse_find_python_folders_and_files(folder_io, except_paths=()):
     except_paths = set(except_paths)
+    except_paths_relative = set()
+
     for root_folder_io, folder_ios, file_ios in folder_io.walk():
         # Delete folders that we don't want to iterate over.
         for file_io in file_ios:
@@ -209,14 +217,21 @@ def recurse_find_python_folders_and_files(folder_io, except_paths=()):
                     yield None, file_io
 
             if path.name == '.gitignore':
-                ignored_paths, ignored_names = \
-                    gitignored_lines(root_folder_io, file_io)
-                except_paths |= ignored_paths
+                ignored_paths_abs, ignored_paths_rel = gitignored_paths(
+                    root_folder_io, file_io
+                )
+                except_paths |= ignored_paths_abs
+                except_paths_relative |= ignored_paths_rel
+
+        except_paths_relative_expanded = expand_relative_ignore_paths(
+            root_folder_io, except_paths_relative
+        )
 
         folder_ios[:] = [
             folder_io
             for folder_io in folder_ios
             if folder_io.path not in except_paths
+            and folder_io.path not in except_paths_relative_expanded
             and folder_io.get_base_name() not in _IGNORE_FOLDERS
         ]
         for folder_io in folder_ios:
@@ -282,12 +297,13 @@ def get_module_contexts_containing_name(inference_state, module_contexts, name,
                                   limit_reduction=limit_reduction)
 
 
-def search_in_file_ios(inference_state, file_io_iterator, name, limit_reduction=1):
+def search_in_file_ios(inference_state, file_io_iterator, name,
+                       limit_reduction=1, complete=False):
     parse_limit = _PARSED_FILE_LIMIT / limit_reduction
     open_limit = _OPENED_FILE_LIMIT / limit_reduction
     file_io_count = 0
     parsed_file_count = 0
-    regex = re.compile(r'\b' + re.escape(name) + r'\b')
+    regex = re.compile(r'\b' + re.escape(name) + (r'' if complete else r'\b'))
     for file_io in file_io_iterator:
         file_io_count += 1
         m = _check_fs(inference_state, file_io, regex)

jedi/inference/star_args.py

@@ -12,6 +12,8 @@ The signature here for bar should be `bar(b, c)` instead of bar(*args).
 """
 from inspect import Parameter
 
+from parso import tree
+
 from jedi.inference.utils import to_list
 from jedi.inference.names import ParamNameWrapper
 from jedi.inference.helpers import is_big_annoying_library
@@ -22,7 +24,11 @@ def _iter_nodes_for_param(param_name):
     from jedi.inference.arguments import TreeArguments
 
     execution_context = param_name.parent_context
-    function_node = execution_context.tree_node
+    # Walk up the parso tree to get the FunctionNode we want. We use the parso
+    # tree rather than going via the execution context so that we're agnostic of
+    # the specific scope we're evaluating within (i.e: module or function,
+    # etc.).
+    function_node = tree.search_ancestor(param_name.tree_name, 'funcdef', 'lambdef')
     module_node = function_node.get_root_node()
     start = function_node.children[-1].start_pos
     end = function_node.children[-1].end_pos

jedi/inference/syntax_tree.py

@@ -2,6 +2,7 @@
 Functions inferring the syntax tree.
 """
 import copy
+import itertools
 
 from parso.python import tree
 
@@ -328,8 +329,8 @@ def infer_atom(context, atom):
         c = atom.children
         # Parentheses without commas are not tuples.
         if c[0] == '(' and not len(c) == 2 \
-                and not(c[1].type == 'testlist_comp'
-                        and len(c[1].children) > 1):
+                and not (c[1].type == 'testlist_comp'
+                         and len(c[1].children) > 1):
             return context.infer_node(c[1])
 
         try:
@@ -492,8 +493,10 @@ def infer_factor(value_set, operator):
         elif operator == 'not':
             b = value.py__bool__()
             if b is None:  # Uncertainty.
-                return
-            yield compiled.create_simple_object(value.inference_state, not b)
+                yield list(value.inference_state.builtins_module.py__getattribute__('bool')
+                           .execute_annotation()).pop()
+            else:
+                yield compiled.create_simple_object(value.inference_state, not b)
         else:
             yield value
 
@@ -515,10 +518,20 @@ def _literals_to_types(inference_state, result):
 
 def _infer_comparison(context, left_values, operator, right_values):
     state = context.inference_state
+    if isinstance(operator, str):
+        operator_str = operator
+    else:
+        operator_str = str(operator.value)
     if not left_values or not right_values:
         # illegal slices e.g. cause left/right_result to be None
         result = (left_values or NO_VALUES) | (right_values or NO_VALUES)
         return _literals_to_types(state, result)
+    elif operator_str == "|" and all(
+        value.is_class() or value.is_compiled()
+        for value in itertools.chain(left_values, right_values)
+    ):
+        # ^^^ A naive hack for PEP 604
+        return ValueSet.from_sets((left_values, right_values))
     else:
         # I don't think there's a reasonable chance that a string
         # operation is still correct, once we pass something like six
@@ -634,7 +647,7 @@ def _infer_comparison_part(inference_state, context, left, operator, right):
             _bool_to_value(inference_state, False)
         ])
     elif str_operator in ('in', 'not in'):
-        return NO_VALUES
+        return inference_state.builtins_module.py__getattribute__('bool').execute_annotation()
 
     def check(obj):
         """Checks if a Jedi object is either a float or an int."""
@@ -684,8 +697,15 @@ def tree_name_to_values(inference_state, context, tree_name):
 
             if expr_stmt.type == "expr_stmt" and expr_stmt.children[1].type == "annassign":
                 correct_scope = parser_utils.get_parent_scope(name) == context.tree_node
+                ann_assign = expr_stmt.children[1]
                 if correct_scope:
                     found_annotation = True
+                    if (
+                        (ann_assign.children[1].type == 'name')
+                        and (ann_assign.children[1].value == tree_name.value)
+                        and context.parent_context
+                    ):
+                        context = context.parent_context
                     value_set |= annotation.infer_annotation(
                         context, expr_stmt.children[1].children[1]
                     ).execute_annotation()
@@ -738,6 +758,13 @@ def tree_name_to_values(inference_state, context, tree_name):
         types = infer_expr_stmt(context, node, tree_name)
     elif typ == 'with_stmt':
         value_managers = context.infer_node(node.get_test_node_from_name(tree_name))
+        if node.parent.type == 'async_stmt':
+            # In the case of `async with` statements, we need to
+            # first get the coroutine from the `__aenter__` method,
+            # then "unwrap" via the `__await__` method
+            enter_methods = value_managers.py__getattribute__('__aenter__')
+            coro = enter_methods.execute_with_values()
+            return coro.py__await__().py__stop_iteration_returns()
         enter_methods = value_managers.py__getattribute__('__enter__')
         return enter_methods.execute_with_values()
     elif typ in ('import_from', 'import_name'):

jedi/inference/sys_path.py

@@ -186,7 +186,6 @@ def _get_buildout_script_paths(search_path: Path):
     directory that look like python files.
 
     :param search_path: absolute path to the module.
-    :type search_path: str
     """
     project_root = _get_parent_dir_with_file(search_path, 'buildout.cfg')
     if not project_root:
@@ -205,7 +204,7 @@ def _get_buildout_script_paths(search_path: Path):
         except (UnicodeDecodeError, IOError) as e:
             # Probably a binary file; permission error or race cond. because
             # file got deleted. Ignore it.
-            debug.warning(e)
+            debug.warning(str(e))
             continue
 
 

jedi/inference/value/dynamic_arrays.py

@@ -16,7 +16,7 @@ settings will stop this process.
 
 It is important to note that:
 
-1. Array modfications work only in the current module.
+1. Array modifications work only in the current module.
 2. Jedi only checks Array additions; ``list.pop``, etc are ignored.
 """
 from jedi import debug

jedi/inference/value/function.py

@@ -344,7 +344,8 @@ class BaseFunctionExecutionContext(ValueContext, TreeContextMixin):
                     GenericClass(c, TupleGenericManager(generics)) for c in async_classes
                 ).execute_annotation()
         else:
-            if self.is_generator():
+            # If there are annotations, prefer them over anything else.
+            if self.is_generator() and not self.infer_annotations():
                 return ValueSet([iterable.Generator(inference_state, self)])
             else:
                 return self.get_return_values()

jedi/inference/value/iterable.py

@@ -342,6 +342,8 @@ class SequenceLiteralValue(Sequence):
         else:
             with reraise_getitem_errors(TypeError, KeyError, IndexError):
                 node = self.get_tree_entries()[index]
+            if node == ':' or node.type == 'subscript':
+                return NO_VALUES
             return self._defining_context.infer_node(node)
 
     def py__iter__(self, contextualized_node=None):
@@ -407,16 +409,6 @@ class SequenceLiteralValue(Sequence):
             else:
                 return [array_node]
 
-    def exact_key_items(self):
-        """
-        Returns a generator of tuples like dict.items(), where the key is
-        resolved (as a string) and the values are still lazy values.
-        """
-        for key_node, value in self.get_tree_entries():
-            for key in self._defining_context.infer_node(key_node):
-                if is_string(key):
-                    yield key.get_safe_value(), LazyTreeValue(self._defining_context, value)
-
     def __repr__(self):
         return "<%s of %s>" % (self.__class__.__name__, self.atom)
 
@@ -472,6 +464,16 @@ class DictLiteralValue(_DictMixin, SequenceLiteralValue, _DictKeyMixin):
 
         return ValueSet([FakeList(self.inference_state, lazy_values)])
 
+    def exact_key_items(self):
+        """
+        Returns a generator of tuples like dict.items(), where the key is
+        resolved (as a string) and the values are still lazy values.
+        """
+        for key_node, value in self.get_tree_entries():
+            for key in self._defining_context.infer_node(key_node):
+                if is_string(key):
+                    yield key.get_safe_value(), LazyTreeValue(self._defining_context, value)
+
     def _dict_values(self):
         return ValueSet.from_sets(
             self._defining_context.infer_node(v)

jedi/inference/value/klass.py

@@ -78,6 +78,8 @@ class ClassName(TreeNameDefinition):
         type_ = super().api_type
         if type_ == 'function':
             definition = self.tree_name.get_definition()
+            if definition is None:
+                return type_
             if function_is_property(definition):
                 # This essentially checks if there is an @property before
                 # the function. @property could be something different, but
@@ -114,25 +116,10 @@ class ClassFilter(ParserTreeFilter):
         while node is not None:
             if node == self._parser_scope or node == self.parent_context:
                 return True
-            node = get_cached_parent_scope(self._used_names, node)
+            node = get_cached_parent_scope(self._parso_cache_node, node)
         return False
 
     def _access_possible(self, name):
-        # Filter for ClassVar variables
-        # TODO this is not properly done, yet. It just checks for the string
-        # ClassVar in the annotation, which can be quite imprecise. If we
-        # wanted to do this correct, we would have to infer the ClassVar.
-        if not self._is_instance:
-            expr_stmt = name.get_definition()
-            if expr_stmt is not None and expr_stmt.type == 'expr_stmt':
-                annassign = expr_stmt.children[1]
-                if annassign.type == 'annassign':
-                    # If there is an =, the variable is obviously also
-                    # defined on the class.
-                    if 'ClassVar' not in annassign.children[1].get_code() \
-                            and '=' not in annassign.children:
-                        return False
-
         # Filter for name mangling of private variables like __foo
         return not name.value.startswith('__') or name.value.endswith('__') \
             or self._equals_origin_scope()

jedi/inference/value/module.py

@@ -64,7 +64,7 @@ class ModuleMixin(SubModuleDictMixin):
                 parent_context=self.as_context(),
                 origin_scope=origin_scope
             ),
-            GlobalNameFilter(self.as_context(), self.tree_node),
+            GlobalNameFilter(self.as_context()),
         )
         yield DictFilter(self.sub_modules_dict())
         yield DictFilter(self._module_attributes_dict())
@@ -148,7 +148,7 @@ class ModuleValue(ModuleMixin, TreeValue):
         if file_io is None:
             self._path: Optional[Path] = None
         else:
-            self._path = Path(file_io.path)
+            self._path = file_io.path
         self.string_names = string_names  # Optional[Tuple[str, ...]]
         self.code_lines = code_lines
         self._is_package = is_package

jedi/inference/value/namespace.py

@@ -1,3 +1,6 @@
+from pathlib import Path
+from typing import Optional
+
 from jedi.inference.cache import inference_state_method_cache
 from jedi.inference.filters import DictFilter
 from jedi.inference.names import ValueNameMixin, AbstractNameDefinition
@@ -41,7 +44,7 @@ class ImplicitNamespaceValue(Value, SubModuleDictMixin):
         string_name = self.py__package__()[-1]
         return ImplicitNSName(self, string_name)
 
-    def py__file__(self):
+    def py__file__(self) -> Optional[Path]:
         return None
 
     def py__package__(self):

jedi/parser_utils.py

@@ -216,11 +216,14 @@ def is_scope(node):
 def _get_parent_scope_cache(func):
     cache = WeakKeyDictionary()
 
-    def wrapper(used_names, node, include_flows=False):
+    def wrapper(parso_cache_node, node, include_flows=False):
+        if parso_cache_node is None:
+            return func(node, include_flows)
+
         try:
-            for_module = cache[used_names]
+            for_module = cache[parso_cache_node]
         except KeyError:
-            for_module = cache[used_names] = {}
+            for_module = cache[parso_cache_node] = {}
 
         try:
             return for_module[node]
@@ -270,7 +273,18 @@ def get_cached_code_lines(grammar, path):
     Basically access the cached code lines in parso. This is not the nicest way
     to do this, but we avoid splitting all the lines again.
     """
-    return parser_cache[grammar._hashed][path].lines
+    return get_parso_cache_node(grammar, path).lines
+
+
+def get_parso_cache_node(grammar, path):
+    """
+    This is of course not public. But as long as I control parso, this
+    shouldn't be a problem. ~ Dave
+
+    The reason for this is mostly caching. This is obviously also a sign of a
+    broken caching architecture.
+    """
+    return parser_cache[grammar._hashed][path]
 
 
 def cut_value_at_position(leaf, position):
@@ -306,7 +320,7 @@ def expr_is_dotted(node):
     return node.type == 'name'
 
 
-def _function_is_x_method(*method_names):
+def _function_is_x_method(decorator_checker):
     def wrapper(function_node):
         """
         This is a heuristic. It will not hold ALL the times, but it will be
@@ -316,12 +330,16 @@ def _function_is_x_method(*method_names):
         """
         for decorator in function_node.get_decorators():
             dotted_name = decorator.children[1]
-            if dotted_name.get_code() in method_names:
+            if decorator_checker(dotted_name.get_code()):
                 return True
         return False
     return wrapper
 
 
-function_is_staticmethod = _function_is_x_method('staticmethod')
-function_is_classmethod = _function_is_x_method('classmethod')
-function_is_property = _function_is_x_method('property', 'cached_property')
+function_is_staticmethod = _function_is_x_method(lambda m: m == "staticmethod")
+function_is_classmethod = _function_is_x_method(lambda m: m == "classmethod")
+function_is_property = _function_is_x_method(
+    lambda m: m == "property"
+    or m == "cached_property"
+    or (m.endswith(".setter"))
+)

jedi/plugins/pytest.py

@@ -1,8 +1,10 @@
+import sys
+from typing import List
 from pathlib import Path
 
 from parso.tree import search_ancestor
 from jedi.inference.cache import inference_state_method_cache
-from jedi.inference.imports import load_module_from_path
+from jedi.inference.imports import goto_import, load_module_from_path
 from jedi.inference.filters import ParserTreeFilter
 from jedi.inference.base_value import NO_VALUES, ValueSet
 from jedi.inference.helpers import infer_call_of_leaf
@@ -31,7 +33,15 @@ def execute(callback):
 def infer_anonymous_param(func):
     def get_returns(value):
         if value.tree_node.annotation is not None:
-            return value.execute_with_values()
+            result = value.execute_with_values()
+            if any(v.name.get_qualified_names(include_module_names=True)
+                   == ('typing', 'Generator')
+                   for v in result):
+                return ValueSet.from_sets(
+                    v.py__getattribute__('__next__').execute_annotation()
+                    for v in result
+                )
+            return result
 
         # In pytest we need to differentiate between generators and normal
         # returns.
@@ -43,6 +53,9 @@ def infer_anonymous_param(func):
             return function_context.get_return_values()
 
     def wrapper(param_name):
+        # parameters with an annotation do not need special handling
+        if param_name.annotation_node:
+            return func(param_name)
         is_pytest_param, param_name_is_function_name = \
             _is_a_pytest_param_and_inherited(param_name)
         if is_pytest_param:
@@ -120,6 +133,36 @@ def _is_pytest_func(func_name, decorator_nodes):
         or any('fixture' in n.get_code() for n in decorator_nodes)
 
 
+def _find_pytest_plugin_modules() -> List[List[str]]:
+    """
+    Finds pytest plugin modules hooked by setuptools entry points
+
+    See https://docs.pytest.org/en/stable/how-to/writing_plugins.html#setuptools-entry-points
+    """
+    if sys.version_info >= (3, 8):
+        from importlib.metadata import entry_points
+
+        if sys.version_info >= (3, 10):
+            pytest_entry_points = entry_points(group="pytest11")
+        else:
+            pytest_entry_points = entry_points().get("pytest11", ())
+
+        if sys.version_info >= (3, 9):
+            return [ep.module.split(".") for ep in pytest_entry_points]
+        else:
+            # Python 3.8 doesn't have `EntryPoint.module`. Implement equivalent
+            # to what Python 3.9 does (with additional None check to placate `mypy`)
+            matches = [
+                ep.pattern.match(ep.value)
+                for ep in pytest_entry_points
+            ]
+            return [x.group('module').split(".") for x in matches if x]
+
+    else:
+        from pkg_resources import iter_entry_points
+        return [ep.module_name.split(".") for ep in iter_entry_points(group="pytest11")]
+
+
 @inference_state_method_cache()
 def _iter_pytest_modules(module_context, skip_own_module=False):
     if not skip_own_module:
@@ -129,32 +172,74 @@ def _iter_pytest_modules(module_context, skip_own_module=False):
     if file_io is not None:
         folder = file_io.get_parent_folder()
         sys_path = module_context.inference_state.get_sys_path()
+
+        # prevent an infinite loop when reaching the root of the current drive
+        last_folder = None
+
         while any(folder.path.startswith(p) for p in sys_path):
             file_io = folder.get_file_io('conftest.py')
             if Path(file_io.path) != module_context.py__file__():
                 try:
                     m = load_module_from_path(module_context.inference_state, file_io)
-                    yield m.as_context()
+                    conftest_module = m.as_context()
+                    yield conftest_module
+
+                    plugins_list = m.tree_node.get_used_names().get("pytest_plugins")
+                    if plugins_list:
+                        name = conftest_module.create_name(plugins_list[0])
+                        yield from _load_pytest_plugins(module_context, name)
                 except FileNotFoundError:
                     pass
             folder = folder.get_parent_folder()
 
-    for names in _PYTEST_FIXTURE_MODULES:
+            # prevent an infinite for loop if the same parent folder is return twice
+            if last_folder is not None and folder.path == last_folder.path:
+                break
+            last_folder = folder  # keep track of the last found parent name
+
+    for names in _PYTEST_FIXTURE_MODULES + _find_pytest_plugin_modules():
         for module_value in module_context.inference_state.import_module(names):
             yield module_value.as_context()
 
 
+def _load_pytest_plugins(module_context, name):
+    from jedi.inference.helpers import get_str_or_none
+
+    for inferred in name.infer():
+        for seq_value in inferred.py__iter__():
+            for value in seq_value.infer():
+                fq_name = get_str_or_none(value)
+                if fq_name:
+                    names = fq_name.split(".")
+                    for module_value in module_context.inference_state.import_module(names):
+                        yield module_value.as_context()
+
+
 class FixtureFilter(ParserTreeFilter):
     def _filter(self, names):
         for name in super()._filter(names):
-            funcdef = name.parent
-            # Class fixtures are not supported
-            if funcdef.type == 'funcdef':
-                decorated = funcdef.parent
-                if decorated.type == 'decorated' and self._is_fixture(decorated):
+            # look for fixture definitions of imported names
+            if name.parent.type == "import_from":
+                imported_names = goto_import(self.parent_context, name)
+                if any(
+                    self._is_fixture(iname.parent_context, iname.tree_name)
+                    for iname in imported_names
+                    # discard imports of whole modules, that have no tree_name
+                    if iname.tree_name
+                ):
                     yield name
 
-    def _is_fixture(self, decorated):
+            elif self._is_fixture(self.parent_context, name):
+                yield name
+
+    def _is_fixture(self, context, name):
+        funcdef = name.parent
+        # Class fixtures are not supported
+        if funcdef.type != "funcdef":
+            return False
+        decorated = funcdef.parent
+        if decorated.type != "decorated":
+            return False
         decorators = decorated.children[0]
         if decorators.type == 'decorators':
             decorators = decorators.children
@@ -171,11 +256,12 @@ class FixtureFilter(ParserTreeFilter):
                     last_leaf = last_trailer.get_last_leaf()
                     if last_leaf == ')':
                         values = infer_call_of_leaf(
-                            self.parent_context, last_leaf, cut_own_trailer=True)
+                            context, last_leaf, cut_own_trailer=True
+                        )
                     else:
-                        values = self.parent_context.infer_node(dotted_name)
+                        values = context.infer_node(dotted_name)
                 else:
-                    values = self.parent_context.infer_node(dotted_name)
+                    values = context.infer_node(dotted_name)
                 for value in values:
                     if value.name.get_qualified_names(include_module_names=True) \
                             == ('_pytest', 'fixtures', 'fixture'):

jedi/plugins/stdlib.py

@@ -803,6 +803,17 @@ _implemented = {
         # For now this works at least better than Jedi trying to understand it.
         'dataclass': _dataclass
     },
+    # attrs exposes declaration interface roughly compatible with dataclasses
+    # via attrs.define, attrs.frozen and attrs.mutable
+    # https://www.attrs.org/en/stable/names.html
+    'attr': {
+        'define': _dataclass,
+        'frozen': _dataclass,
+    },
+    'attrs': {
+        'define': _dataclass,
+        'frozen': _dataclass,
+    },
     'os.path': {
         'dirname': _create_string_input_function(os.path.dirname),
         'abspath': _create_string_input_function(os.path.abspath),

jedi/settings.py

@@ -143,6 +143,15 @@ This improves autocompletion for libraries that use ``setattr`` or
 ``globals()`` modifications a lot.
 """
 
+allow_unsafe_interpreter_executions = True
+"""
+Controls whether descriptors are evaluated when using an Interpreter. This is
+something you might want to control when using Jedi from a Repl (e.g. IPython)
+
+Generally this setting allows Jedi to execute __getitem__ and descriptors like
+`property`.
+"""
+
 # ----------------
 # Caching Validity
 # ----------------

jedi/utils.py

@@ -2,7 +2,7 @@
 Utilities for end-users.
 """
 
-import __main__  # type: ignore[import]
+import __main__
 from collections import namedtuple
 import logging
 import traceback

setup.cfg

@@ -21,13 +21,25 @@ per-file-ignores =
   jedi/__init__.py:F401
   jedi/inference/compiled/__init__.py:F401
   jedi/inference/value/__init__.py:F401
-exclude = jedi/third_party/* .tox/*
+exclude =
+  .tox/*
+  jedi/third_party/*
+  test/completion/*
+  test/examples/*
+  test/refactor/*
+  test/static_analysis/*
 
 [pycodestyle]
 max-line-length = 100
 
 
 [mypy]
+# Exclude our copies of external stubs
+exclude = ^jedi/third_party
+
+show_error_codes = true
+enable_error_code = ignore-without-code
+
 # Ensure generics are explicit about what they are (e.g: `List[str]` rather than
 # just `List`)
 disallow_any_generics = True

setup.py

@@ -27,24 +27,61 @@ setup(name='jedi',
       maintainer=__AUTHOR__,
       maintainer_email=__AUTHOR_EMAIL__,
       url='https://github.com/davidhalter/jedi',
+      project_urls={
+          "Documentation": 'https://jedi.readthedocs.io/en/latest/',
+      },
       license='MIT',
       keywords='python completion refactoring vim',
       long_description=readme,
       packages=find_packages(exclude=['test', 'test.*']),
       python_requires='>=3.6',
-      install_requires=['parso>=0.8.0,<0.9.0'],
+      # Python 3.13 grammars are added to parso in 0.8.4
+      install_requires=['parso>=0.8.4,<0.9.0'],
       extras_require={
           'testing': [
-              'pytest<6.0.0',
+              'pytest<9.0.0',
               # docopt for sith doctests
               'docopt',
               # coloroma for colored debug output
               'colorama',
-              'Django<3.1',  # For now pin this.
+              'Django',
+              'attrs',
           ],
           'qa': [
-              'flake8==3.8.3',
-              'mypy==0.782',
+              # latest version supporting Python 3.6
+              'flake8==5.0.4',
+              # latest version supporting Python 3.6
+              'mypy==0.971',
+              # Arbitrary pins, latest at the time of pinning
+              'types-setuptools==67.2.0.1',
+          ],
+          'docs': [
+              # Just pin all of these.
+              'Jinja2==2.11.3',
+              'MarkupSafe==1.1.1',
+              'Pygments==2.8.1',
+              'alabaster==0.7.12',
+              'babel==2.9.1',
+              'chardet==4.0.0',
+              'commonmark==0.8.1',
+              'docutils==0.17.1',
+              'future==0.18.2',
+              'idna==2.10',
+              'imagesize==1.2.0',
+              'mock==1.0.1',
+              'packaging==20.9',
+              'pyparsing==2.4.7',
+              'pytz==2021.1',
+              'readthedocs-sphinx-ext==2.1.4',
+              'recommonmark==0.5.0',
+              'requests==2.25.1',
+              'six==1.15.0',
+              'snowballstemmer==2.1.0',
+              'sphinx==1.8.5',
+              'sphinx-rtd-theme==0.4.3',
+              'sphinxcontrib-serializinghtml==1.1.4',
+              'sphinxcontrib-websupport==1.2.4',
+              'urllib3==1.26.4',
           ],
       },
       package_data={'jedi': ['*.pyi', 'third_party/typeshed/LICENSE',
@@ -61,6 +98,10 @@ setup(name='jedi',
           'Programming Language :: Python :: 3.7',
           'Programming Language :: Python :: 3.8',
           'Programming Language :: Python :: 3.9',
+          'Programming Language :: Python :: 3.10',
+          'Programming Language :: Python :: 3.11',
+          'Programming Language :: Python :: 3.12',
+          'Programming Language :: Python :: 3.13',
           'Topic :: Software Development :: Libraries :: Python Modules',
           'Topic :: Text Editors :: Integrated Development Environments (IDE)',
           'Topic :: Utilities',

test/completion/arrays.py

@@ -44,6 +44,8 @@ b[int():]
 
 #? list()
 b[:]
+#? int()
+b[:, :-1]
 
 #? 3
 b[:]
@@ -67,6 +69,20 @@ class _StrangeSlice():
 #? slice()
 _StrangeSlice()[1:2]
 
+for x in b[:]:
+    #? int()
+    x
+
+for x in b[:, :-1]:
+    #?
+    x
+
+class Foo:
+    def __getitem__(self, item):
+        return item
+
+#?
+Foo()[:, :-1][0]
 
 # -----------------
 # iterable multiplication

test/completion/async_.py

@@ -26,11 +26,6 @@ async def y():
     x().__await__().__next
     return 2
 
-async def x2():
-    async with open('asdf') as f:
-        #? ['readlines']
-        f.readlines
-
 class A():
     @staticmethod
     async def b(c=1, d=2):
@@ -52,8 +47,6 @@ async def awaitable_test():
     #? str()
     foo
 
-# python >= 3.6
-
 async def asgen():
     yield 1
     await asyncio.sleep(0)
@@ -105,3 +98,22 @@ async def f():
     f = await C().async_for_classmethod()
     #? C()
     f
+
+
+class AsyncCtxMgr:
+    def some_method():
+        pass
+
+    async def __aenter__(self):
+        return self
+
+    async def __aexit__(self, *args):
+        pass
+
+
+async def asyncctxmgr():
+    async with AsyncCtxMgr() as acm:
+        #? AsyncCtxMgr()
+        acm
+        #? ['some_method']
+        acm.som

test/completion/basic.py

@@ -413,6 +413,10 @@ with Foo() as f3:
 with Foo() as f3:
     f3
 
+with open("a"), open("b") as bfile:
+    #? ['flush']
+   bfile.flush
+
 # -----------------
 # Avoiding multiple definitions
 # -----------------
@@ -420,3 +424,13 @@ with Foo() as f3:
 some_array = ['', '']
 #! ['def upper']
 some_array[some_not_defined_index].upper
+
+# -----------------
+# operator
+# -----------------
+
+#? bool()
+res = 'f' in 'foo'; res
+
+#? bool()
+res = not {}; res

test/completion/conftest.py

@@ -23,7 +23,13 @@ def inheritance_fixture():
 
 
 @pytest.fixture
-def testdir(testdir):
-    #? ['chdir']
-    testdir.chdir
-    return testdir
+def capsysbinary(capsysbinary):
+    #? ['close']
+    capsysbinary.clos
+    return capsysbinary
+
+
+# used when fixtures are defined in multiple files
+pytest_plugins = [
+    "completion.fixture_module",
+]

test/completion/context.py

@@ -21,11 +21,9 @@ class Y(X):
     #? []
     def __doc__
 
-    # This might or might not be what we wanted, currently properties are also
-    # used like this. IMO this is not wanted ~dave.
-    #? ['__class__']
-    def __class__
     #? []
+    def __class__
+    #? ['__class__']
     __class__
 
 

test/completion/django.py

@@ -5,6 +5,7 @@ import uuid
 from django.db import models
 from django.contrib.auth.models import User
 from django.db.models.query_utils import DeferredAttribute
+from django.db.models.manager import BaseManager
 
 
 class TagManager(models.Manager):

test/completion/docstring.py

@@ -284,6 +284,13 @@ def doctest_with_space():
         import_issu
     """
 
+def doctest_issue_github_1748():
+    """From GitHub #1748
+    #? 10 []
+    This. Al
+    """
+    pass
+
 
 def docstring_rst_identifiers():
     """

test/completion/fixture_module.py

@@ -0,0 +1,6 @@
+# Exists only for completion/pytest.py
+import pytest
+
+@pytest.fixture
+def my_module_fixture():
+    return 1.0

test/completion/fstring.py

@@ -1,5 +1,3 @@
-# python >= 3.6
-
 class Foo:
     bar = 1
 

test/completion/generators.py

@@ -309,3 +309,8 @@ def annotation2() -> Iterator[float]:
 next(annotation1())
 #? float()
 next(annotation2())
+
+
+# annotations should override generator inference
+#? float()
+annotation1()

test/completion/lambdas.py

@@ -110,4 +110,4 @@ class Test(object):
 # nocond lambdas make no sense at all.
 
 #? int()
-[a for a in [1,2] if lambda: 3][0]
+[a for a in [1,2] if (lambda: 3)][0]

test/completion/namespace1/pkg1/pkg2/mod1.py

@@ -0,0 +1 @@
+mod1_name = 'mod1'

test/completion/namespace2/pkg1/pkg2/mod2.py

@@ -0,0 +1 @@
+mod2_name = 'mod2'

test/completion/ns_path.py

@@ -0,0 +1,18 @@
+import sys
+import os
+from os.path import dirname
+
+sys.path.insert(0, os.path.join(dirname(__file__), 'namespace2'))
+sys.path.insert(0, os.path.join(dirname(__file__), 'namespace1'))
+
+#? ['mod1']
+import pkg1.pkg2.mod1
+
+#? ['mod2']
+import pkg1.pkg2.mod2
+
+#? ['mod1_name']
+pkg1.pkg2.mod1.mod1_name
+
+#? ['mod2_name']
+pkg1.pkg2.mod2.mod2_name

test/completion/on_import.py

@@ -23,11 +23,9 @@ def builtin_test():
 import sqlite3
 
 # classes is a local module that has an __init__.py and can therefore not be
-# found. test can be found.
+# found.
 #? []
 import classes
-#? ['test']
-import test
 
 #? ['timedelta']
 from datetime import timedel
@@ -78,7 +76,7 @@ from import_tree.pkg.mod1 import not_existant,
 #? 22 ['mod1', 'base']
 from import_tree.pkg. import mod1
 #? 17 ['mod1', 'mod2', 'random', 'pkg', 'references', 'rename1', 'rename2', 'classes', 'globals', 'recurse_class1', 'recurse_class2', 'invisible_pkg', 'flow_import']
-from import_tree. import pkg
+from import_tree. import new_pkg
 
 #? 18 ['pkg']
 from import_tree.p import pkg

test/completion/pep0484_basic.py

@@ -180,6 +180,11 @@ def argskwargs(*args: int, **kwargs: float):
     #? float()
     kwargs['']
 
+class Test:
+    str: str = 'abc'
+
+#? ['upper']
+Test.str.upp
 
 class NotCalledClass:
     def __init__(self, x):

test/completion/pep0484_decorators.py

@@ -0,0 +1,50 @@
+""" Pep-0484 type hinted decorators """
+
+from typing import Callable
+
+
+def decorator(func):
+    def wrapper(*a, **k):
+        return str(func(*a, **k))
+    return wrapper
+
+
+def typed_decorator(func: Callable[..., int]) -> Callable[..., str]:
+    ...
+
+# Functions
+
+@decorator
+def plain_func() -> int:
+    return 4
+
+#? str()
+plain_func()
+
+
+@typed_decorator
+def typed_func() -> int:
+    return 4
+
+#? str()
+typed_func()
+
+
+# Methods
+
+class X:
+    @decorator
+    def plain_method(self) -> int:
+        return 4
+
+    @typed_decorator
+    def typed_method(self) -> int:
+        return 4
+
+inst = X()
+
+#? str()
+inst.plain_method()
+
+#? str()
+inst.typed_method()

test/completion/pep0484_generic_mismatches.py

@@ -27,13 +27,13 @@ class PlainClass(object):
 
 
 tpl = ("1", 2)
-tpl_typed = ("2", 3)  # type: Tuple[str, int]
+tpl_typed: Tuple[str, int] = ("2", 3)
 
 collection = {"a": 1}
-collection_typed = {"a": 1}  # type: Dict[str, int]
+collection_typed: Dict[str, int] = {"a": 1}
 
-list_of_ints = [42]  # type: List[int]
-list_of_funcs = [foo]  # type: List[Callable[[T], T]]
+list_of_ints: List[int] = [42]
+list_of_funcs: List[Callable[[T], T]] = [foo]
 
 custom_generic = CustomGeneric(123.45)
 

test/completion/pep0484_generic_parameters.py

@@ -19,12 +19,12 @@ T_co = TypeVar('T_co', covariant=True)
 V = TypeVar('V')
 
 
-just_float = 42.  # type: float
-optional_float = 42.  # type: Optional[float]
-list_of_ints = [42]  # type: List[int]
-list_of_floats = [42.]  # type: List[float]
-list_of_optional_floats = [x or None for x in list_of_floats]  # type: List[Optional[float]]
-list_of_ints_and_strs = [42, 'abc']  # type: List[Union[int, str]]
+just_float: float = 42.
+optional_float: Optional[float] = 42.
+list_of_ints: List[int] = [42]
+list_of_floats: List[float] = [42.]
+list_of_optional_floats: List[Optional[float]] = [x or None for x in list_of_floats]
+list_of_ints_and_strs: List[Union[int, str]] = [42, 'abc']
 
 # Test that simple parameters are handled
 def list_t_to_list_t(the_list: List[T]) -> List[T]:
@@ -48,7 +48,7 @@ for z in list_t_to_list_t(list_of_ints_and_strs):
     z
 
 
-list_of_int_type = [int]  # type: List[Type[int]]
+list_of_int_type: List[Type[int]] = [int]
 
 # Test that nested parameters are handled
 def list_optional_t_to_list_t(the_list: List[Optional[T]]) -> List[T]:
@@ -85,7 +85,7 @@ def optional_list_t_to_list_t(x: Optional[List[T]]) -> List[T]:
     return x if x is not None else []
 
 
-optional_list_float = None  # type: Optional[List[float]]
+optional_list_float: Optional[List[float]] = None
 for xc in optional_list_t_to_list_t(optional_list_float):
     #? float()
     xc
@@ -134,7 +134,7 @@ def list_tuple_t_to_tuple_list_t(the_list: List[Tuple[T]]) -> Tuple[List[T], ...
     return tuple(list(x) for x in the_list)
 
 
-list_of_int_tuples = [(x,) for x in list_of_ints]  # type: List[Tuple[int]]
+list_of_int_tuples: List[Tuple[int]] = [(x,) for x in list_of_ints]
 
 for b in list_tuple_t_to_tuple_list_t(list_of_int_tuples):
     #? int()
@@ -145,7 +145,7 @@ def list_tuple_t_elipsis_to_tuple_list_t(the_list: List[Tuple[T, ...]]) -> Tuple
     return tuple(list(x) for x in the_list)
 
 
-list_of_int_tuple_elipsis = [tuple(list_of_ints)]  # type: List[Tuple[int, ...]]
+list_of_int_tuple_elipsis: List[Tuple[int, ...]] = [tuple(list_of_ints)]
 
 for b in list_tuple_t_elipsis_to_tuple_list_t(list_of_int_tuple_elipsis):
     #? int()
@@ -157,7 +157,7 @@ def foo(x: int) -> int:
     return x
 
 
-list_of_funcs = [foo]  # type: List[Callable[[int], int]]
+list_of_funcs: List[Callable[[int], int]] = [foo]
 
 def list_func_t_to_list_func_type_t(the_list: List[Callable[[T], T]]) -> List[Callable[[Type[T]], T]]:
     def adapt(func: Callable[[T], T]) -> Callable[[Type[T]], T]:
@@ -176,7 +176,7 @@ def bar(*a, **k) -> int:
     return len(a) + len(k)
 
 
-list_of_funcs_2 = [bar]  # type: List[Callable[..., int]]
+list_of_funcs_2: List[Callable[..., int]] = [bar]
 
 def list_func_t_passthrough(the_list: List[Callable[..., T]]) -> List[Callable[..., T]]:
     return the_list
@@ -187,7 +187,7 @@ for b in list_func_t_passthrough(list_of_funcs_2):
     b(None, x="x")
 
 
-mapping_int_str = {42: 'a'}  # type: Dict[int, str]
+mapping_int_str: Dict[int, str] = {42: 'a'}
 
 # Test that mappings (that have more than one parameter) are handled
 def invert_mapping(mapping: Mapping[K, V]) -> Mapping[V, K]:
@@ -210,11 +210,11 @@ first(mapping_int_str)
 #? str()
 first("abc")
 
-some_str = NotImplemented  # type: str
+some_str: str = NotImplemented
 #? str()
 first(some_str)
 
-annotated = [len]  # type: List[ Callable[[Sequence[float]], int] ]
+annotated: List[ Callable[[Sequence[float]], int] ] = [len]
 #? int()
 first(annotated)()
 
@@ -237,7 +237,7 @@ for b in values(mapping_int_str):
 #
 # Tests that user-defined generic types are handled
 #
-list_ints = [42]  # type: List[int]
+list_ints: List[int] = [42]
 
 class CustomGeneric(Generic[T_co]):
     def __init__(self, val: T_co) -> None:
@@ -248,7 +248,7 @@ class CustomGeneric(Generic[T_co]):
 def custom(x: CustomGeneric[T]) -> T:
     return x.val
 
-custom_instance = CustomGeneric(42)  # type: CustomGeneric[int]
+custom_instance: CustomGeneric[int] = CustomGeneric(42)
 
 #? int()
 custom(custom_instance)
@@ -275,7 +275,7 @@ for x5 in wrap_custom(list_ints):
 
 
 # Test extraction of type from a nested custom generic type
-list_custom_instances = [CustomGeneric(42)]  # type: List[CustomGeneric[int]]
+list_custom_instances: List[CustomGeneric[int]] = [CustomGeneric(42)]
 
 def unwrap_custom(iterable: Iterable[CustomGeneric[T]]) -> List[T]:
     return [x.val for x in iterable]
@@ -303,7 +303,7 @@ for xg in unwrap_custom(CustomGeneric(s) for s in 'abc'):
 
 
 # Test extraction of type from type parameer nested within a custom generic type
-custom_instance_list_int = CustomGeneric([42])  # type: CustomGeneric[List[int]]
+custom_instance_list_int: CustomGeneric[List[int]] = CustomGeneric([42])
 
 def unwrap_custom2(instance: CustomGeneric[Iterable[T]]) -> List[T]:
     return list(instance.val)
@@ -326,7 +326,7 @@ class Specialised(Mapping[int, str]):
     pass
 
 
-specialised_instance = NotImplemented  # type: Specialised
+specialised_instance: Specialised = NotImplemented
 
 #? int()
 first(specialised_instance)
@@ -341,7 +341,7 @@ class ChildOfSpecialised(Specialised):
     pass
 
 
-child_of_specialised_instance = NotImplemented  # type: ChildOfSpecialised
+child_of_specialised_instance: ChildOfSpecialised = NotImplemented
 
 #? int()
 first(child_of_specialised_instance)
@@ -355,13 +355,13 @@ class CustomPartialGeneric1(Mapping[str, T]):
     pass
 
 
-custom_partial1_instance = NotImplemented  # type: CustomPartialGeneric1[int]
+custom_partial1_instance: CustomPartialGeneric1[int] = NotImplemented
 
 #? str()
 first(custom_partial1_instance)
 
 
-custom_partial1_unbound_instance = NotImplemented  # type: CustomPartialGeneric1
+custom_partial1_unbound_instance: CustomPartialGeneric1 = NotImplemented
 
 #? str()
 first(custom_partial1_unbound_instance)
@@ -371,7 +371,7 @@ class CustomPartialGeneric2(Mapping[T, str]):
     pass
 
 
-custom_partial2_instance = NotImplemented  # type: CustomPartialGeneric2[int]
+custom_partial2_instance: CustomPartialGeneric2[int] = NotImplemented
 
 #? int()
 first(custom_partial2_instance)
@@ -380,7 +380,7 @@ first(custom_partial2_instance)
 values(custom_partial2_instance)[0]
 
 
-custom_partial2_unbound_instance = NotImplemented  # type: CustomPartialGeneric2
+custom_partial2_unbound_instance: CustomPartialGeneric2 = NotImplemented
 
 #? []
 first(custom_partial2_unbound_instance)

test/completion/pep0484_generic_passthroughs.py

@@ -19,16 +19,16 @@ TTypeAny = TypeVar('TTypeAny', bound=Type[Any])
 TCallable = TypeVar('TCallable', bound=Callable[..., Any])
 
 untyped_list_str = ['abc', 'def']
-typed_list_str = ['abc', 'def']  # type: List[str]
+typed_list_str: List[str] = ['abc', 'def']
 
 untyped_tuple_str = ('abc',)
-typed_tuple_str = ('abc',)  # type: Tuple[str]
+typed_tuple_str: Tuple[str] = ('abc',)
 
 untyped_tuple_str_int = ('abc', 4)
-typed_tuple_str_int = ('abc', 4)  # type: Tuple[str, int]
+typed_tuple_str_int: Tuple[str, int] = ('abc', 4)
 
-variadic_tuple_str = ('abc',)  # type: Tuple[str, ...]
-variadic_tuple_str_int = ('abc', 4)  # type: Tuple[Union[str, int], ...]
+variadic_tuple_str: Tuple[str, ...] = ('abc',)
+variadic_tuple_str_int: Tuple[Union[str, int], ...] = ('abc', 4)
 
 
 def untyped_passthrough(x):
@@ -58,6 +58,16 @@ def typed_bound_generic_passthrough(x: TList) -> TList:
 
     return x
 
+# Forward references are more likely with custom types, however this aims to
+# test just the handling of the quoted type rather than any other part of the
+# machinery.
+def typed_quoted_return_generic_passthrough(x: T) -> 'List[T]':
+    return [x]
+
+def typed_quoted_input_generic_passthrough(x: 'Tuple[T]') -> T:
+    x
+    return x[0]
+
 
 for a in untyped_passthrough(untyped_list_str):
     #? str()
@@ -146,6 +156,23 @@ for q in typed_bound_generic_passthrough(typed_list_str):
     q
 
 
+for r in typed_quoted_return_generic_passthrough("something"):
+    #? str()
+    r
+
+for s in typed_quoted_return_generic_passthrough(42):
+    #? int()
+    s
+
+
+#? str()
+typed_quoted_input_generic_passthrough(("something",))
+
+#? int()
+typed_quoted_input_generic_passthrough((42,))
+
+
+
 class CustomList(List):
     def get_first(self):
         return self[0]

test/completion/pep0484_overload.py

@@ -1,10 +1,9 @@
-# python >= 3.6
 from typing import List, Dict, overload, Tuple, TypeVar
 
 lst: list
 list_alias: List
 list_str: List[str]
-list_str: List[int]
+list_int: List[int]
 
 # -------------------------
 # With base classes

test/completion/pep0484_typing.py

@@ -2,18 +2,14 @@
 Test the typing library, with docstrings and annotations
 """
 import typing
+from typing import Sequence, MutableSequence, List, Iterable, Iterator, \
+    AbstractSet, Tuple, Mapping, Dict, Union, Optional
+
 class B:
     pass
 
-def we_can_has_sequence(p, q, r, s, t, u):
-    """
-    :type p: typing.Sequence[int]
-    :type q: typing.Sequence[B]
-    :type r: typing.Sequence[int]
-    :type s: typing.Sequence["int"]
-    :type t: typing.MutableSequence[dict]
-    :type u: typing.List[float]
-    """
+def we_can_has_sequence(p: Sequence[int], q: Sequence[B], r: Sequence[int],
+                        s: Sequence["int"], t: MutableSequence[dict], u: List[float]):
     #? ["count"]
     p.c
     #? int()
@@ -43,13 +39,8 @@ def we_can_has_sequence(p, q, r, s, t, u):
     #? float()
     u[1]
 
-def iterators(ps, qs, rs, ts):
-    """
-    :type ps: typing.Iterable[int]
-    :type qs: typing.Iterator[str]
-    :type rs: typing.Sequence["ForwardReference"]
-    :type ts: typing.AbstractSet["float"]
-    """
+def iterators(ps: Iterable[int], qs: Iterator[str], rs:
+              Sequence["ForwardReference"], ts: AbstractSet["float"]):
     for p in ps:
         #? int()
         p
@@ -79,22 +70,13 @@ def iterators(ps, qs, rs, ts):
         #? float()
         t
 
-def sets(p, q):
-    """
-    :type p: typing.AbstractSet[int]
-    :type q: typing.MutableSet[float]
-    """
+def sets(p: AbstractSet[int], q: typing.MutableSet[float]):
     #? []
     p.a
     #? ["add"]
     q.a
 
-def tuple(p, q, r):
-    """
-    :type p: typing.Tuple[int]
-    :type q: typing.Tuple[int, str, float]
-    :type r: typing.Tuple[B, ...]
-    """
+def tuple(p: Tuple[int], q: Tuple[int, str, float], r: Tuple[B, ...]):
     #? int()
     p[0]
     #? ['index']
@@ -127,16 +109,14 @@ class Key:
 class Value:
     pass
 
-def mapping(p, q, d, dd, r, s, t):
-    """
-    :type p: typing.Mapping[Key, Value]
-    :type q: typing.MutableMapping[Key, Value]
-    :type d: typing.Dict[Key, Value]
-    :type dd: typing.DefaultDict[Key, Value]
-    :type r: typing.KeysView[Key]
-    :type s: typing.ValuesView[Value]
-    :type t: typing.ItemsView[Key, Value]
-    """
+def mapping(
+        p: Mapping[Key, Value],
+        q: typing.MutableMapping[Key, Value],
+        d: Dict[Key, Value],
+        dd: typing.DefaultDict[Key, Value],
+        r: typing.KeysView[Key],
+        s: typing.ValuesView[Value],
+        t: typing.ItemsView[Key, Value]):
     #? []
     p.setd
     #? ["setdefault"]
@@ -198,14 +178,12 @@ def mapping(p, q, d, dd, r, s, t):
         #? Value()
         value
 
-def union(p, q, r, s, t):
-    """
-    :type p: typing.Union[int]
-    :type q: typing.Union[int, int]
-    :type r: typing.Union[int, str, "int"]
-    :type s: typing.Union[int, typing.Union[str, "typing.Union['float', 'dict']"]]
-    :type t: typing.Union[int, None]
-    """
+def union(
+    p: Union[int],
+    q: Union[int, int],
+    r: Union[int, str, "int"],
+    s: Union[int, typing.Union[str, "typing.Union['float', 'dict']"]],
+    t: Union[int, None]):
     #? int()
     p
     #? int()
@@ -217,9 +195,8 @@ def union(p, q, r, s, t):
     #? int() None
     t
 
-def optional(p):
+def optional(p: Optional[int]):
     """
-    :type p: typing.Optional[int]
     Optional does not do anything special. However it should be recognised
     as being of that type. Jedi doesn't do anything with the extra into that
     it can be None as well
@@ -234,10 +211,7 @@ class TestDict(typing.Dict[str, int]):
     def setdud(self):
         pass
 
-def testdict(x):
-    """
-    :type x: TestDict
-    """
+def testdict(x: TestDict):
     #? ["setdud", "setdefault"]
     x.setd
     for key in x.keys():
@@ -262,10 +236,7 @@ y = WrappingType(0) # Per https://github.com/davidhalter/jedi/issues/1015#issuec
 #? str()
 y
 
-def testnewtype(y):
-    """
-    :type y: WrappingType
-    """
+def testnewtype(y: WrappingType):
     #? str()
     y
     #? ["upper"]
@@ -273,10 +244,7 @@ def testnewtype(y):
 
 WrappingType2 = typing.NewType()
 
-def testnewtype2(y):
-    """
-    :type y: WrappingType2
-    """
+def testnewtype2(y: WrappingType2):
     #?
     y
     #? []
@@ -297,10 +265,7 @@ class TestDefaultDict(typing.DefaultDict[str, int]):
     def setdud(self):
         pass
 
-def testdict(x):
-    """
-    :type x: TestDefaultDict
-    """
+def testdict(x: TestDefaultDict):
     #? ["setdud", "setdefault"]
     x.setd
     for key in x.keys():

test/completion/pep0526_variables.py

@@ -1,9 +1,6 @@
 """
-PEP 526 introduced a new way of using type annotations on variables. It was
-introduced in Python 3.6.
+PEP 526 introduced a way of using type annotations on variables.
 """
-# python >= 3.6
-
 import typing
 
 asdf = ''
@@ -47,7 +44,7 @@ class Foo():
     baz: typing.ClassVar[str]
 
 
-#?
+#? int()
 Foo.bar
 #? int()
 Foo().bar
@@ -61,6 +58,7 @@ class VarClass:
     var_instance2: float
     var_class1: typing.ClassVar[str] = 1
     var_class2: typing.ClassVar[bytes]
+    var_class3 = None
 
     def __init__(self):
         #? int()
@@ -73,15 +71,21 @@ class VarClass:
         d.var_class2
         #? []
         d.int
-        #? ['var_class1', 'var_class2', 'var_instance1', 'var_instance2']
+        #? ['var_class1', 'var_class2', 'var_instance1', 'var_instance2', 'var_class3']
         self.var_
 
+class VarClass2(VarClass):
+    var_class3: typing.ClassVar[int]
 
-#? ['var_class1', 'var_class2', 'var_instance1']
+    def __init__(self):
+        #? int()
+        self.var_class3
+
+#? ['var_class1', 'var_class2', 'var_instance1', 'var_class3', 'var_instance2']
 VarClass.var_
 #? int()
 VarClass.var_instance1
-#?
+#? float()
 VarClass.var_instance2
 #? str()
 VarClass.var_class1
@@ -91,7 +95,7 @@ VarClass.var_class2
 VarClass.int
 
 d = VarClass()
-#? ['var_class1', 'var_class2', 'var_instance1', 'var_instance2']
+#? ['var_class1', 'var_class2', 'var_class3', 'var_instance1', 'var_instance2']
 d.var_
 #? int()
 d.var_instance1

test/completion/pep0593_annotations.py

@@ -0,0 +1,26 @@
+# python >= 3.9
+
+from typing import Annotated
+
+# This is just a dummy and very meaningless thing to use with to the Annotated
+# type hint
+class Foo:
+    pass
+
+class A:
+    pass
+
+
+def annotated_function_params(
+    basic: Annotated[str, Foo()],
+    obj: A,
+    annotated_obj: Annotated[A, Foo()],
+):
+    #? str()
+    basic
+
+    #? A()
+    obj
+
+    #? A()
+    annotated_obj

test/completion/pep0604.py

@@ -0,0 +1,46 @@
+from pep0484_generic_parameters import list_t_to_list_t
+
+list_of_ints_and_strs: list[int | str]
+
+# Test that unions are handled
+x2 = list_t_to_list_t(list_of_ints_and_strs)[0]
+#? int() str()
+x2
+
+for z in list_t_to_list_t(list_of_ints_and_strs):
+    #? int() str()
+    z
+
+
+from pep0484_generic_passthroughs import (
+    typed_variadic_tuple_generic_passthrough,
+)
+
+variadic_tuple_str_int: tuple[int | str, ...]
+
+for m in typed_variadic_tuple_generic_passthrough(variadic_tuple_str_int):
+    #? str() int()
+    m
+
+
+def func_returns_byteslike() -> bytes | bytearray:
+    pass
+
+#? bytes() bytearray()
+func_returns_byteslike()
+
+
+pep604_optional_1: int | str | None
+pep604_optional_2: None | bytes
+
+#? int() str() None
+pep604_optional_1
+
+#? None bytes()
+pep604_optional_2
+
+
+pep604_in_str: "int | bytes"
+
+#? int() bytes()
+pep604_in_str

test/completion/pytest.py

@@ -1,3 +1,5 @@
+from typing import Generator
+
 import pytest
 from pytest import fixture
 
@@ -64,6 +66,11 @@ def lala(my_fixture):
 def lala(my_fixture):
     pass
 
+# overriding types of a fixture should be possible
+def test_x(my_yield_fixture: str):
+    #? str()
+    my_yield_fixture
+
 # -----------------
 # completion
 # -----------------
@@ -89,6 +96,9 @@ def test_x(my_con
 #? 18 ['my_conftest_fixture']
 def test_x(my_conftest_fixture):
     return
+#? ['my_module_fixture']
+def test_x(my_modu
+    return
 
 #? []
 def lala(my_con
@@ -132,9 +142,6 @@ def test_p(monkeypatch):
 #? ['capsysbinary']
 def test_p(capsysbin
 
-#? ['tmpdir', 'tmpdir_factory']
-def test_p(tmpdi
-
 
 def close_parens():
     pass
@@ -164,3 +171,40 @@ def test_inheritance_fixture(inheritance_fixture, caplog):
 @pytest.fixture
 def caplog(caplog):
     yield caplog
+
+# -----------------
+# Generator with annotation
+# -----------------
+
+@pytest.fixture
+def with_annot() -> Generator[float, None, None]:
+    pass
+
+def test_with_annot(inheritance_fixture, with_annot):
+    #? float()
+    with_annot
+
+# -----------------
+# pytest external plugins
+# -----------------
+
+#? ['admin_user', 'admin_client']
+def test_z(admin
+
+#! 15 ['def admin_client']
+def test_p(admin_client):
+    #? ['login', 'logout']
+    admin_client.log
+
+@pytest.fixture
+@some_decorator
+#? ['admin_user']
+def bla(admin_u
+    return
+
+@pytest.fixture
+@some_decorator
+#! 12 ['def admin_user']
+def bla(admin_user):
+    pass
+

test/conftest.py

@@ -1,7 +1,6 @@
 import os
 import sys
 import subprocess
-from itertools import count
 
 import pytest
 
@@ -10,9 +9,6 @@ from . import run
 from . import refactor
 from jedi import InterpreterEnvironment, get_system_environment
 from jedi.inference.compiled.value import create_from_access_path
-from jedi.inference.imports import _load_python_module
-from jedi.file_io import KnownContentFileIO
-from jedi.inference.base_value import ValueSet
 from jedi.api.interpreter import MixedModuleContext
 
 # For interpreter tests sometimes the path of this directory is in the sys
@@ -163,19 +159,6 @@ def create_compiled_object(inference_state):
     )
 
 
-@pytest.fixture
-def module_injector():
-    counter = count()
-
-    def module_injector(inference_state, names, code):
-        assert isinstance(names, tuple)
-        file_io = KnownContentFileIO('/foo/bar/module-injector-%s.py' % next(counter), code)
-        v = _load_python_module(inference_state, file_io, names)
-        inference_state.module_cache.add(names, ValueSet([v]))
-
-    return module_injector
-
-
 @pytest.fixture(params=[False, True])
 def class_findable(monkeypatch, request):
     if not request.param:

test/examples/pytest_plugin_package/pytest_plugin/fixtures.py

@@ -0,0 +1,6 @@
+from pytest import fixture
+
+
+@fixture()
+def admin_user():
+    pass

test/examples/pytest_plugin_package/pytest_plugin/plugin.py

@@ -0,0 +1,16 @@
+import pytest
+
+from .fixtures import admin_user  # noqa
+
+
+@pytest.fixture()
+def admin_client():
+    return Client()
+
+
+class Client:
+    def login(self, **credentials):
+        ...
+
+    def logout(self):
+        ...

test/run.py

@@ -104,10 +104,14 @@ import os
 import re
 import sys
 import operator
-from ast import literal_eval
+if sys.version_info < (3, 8):
+    literal_eval = eval
+else:
+    from ast import literal_eval
 from io import StringIO
 from functools import reduce
 from unittest.mock import ANY
+from pathlib import Path
 
 import parso
 from _pytest.outcomes import Skipped
@@ -122,6 +126,7 @@ from jedi.api.environment import get_default_environment, get_system_environment
 from jedi.inference.gradual.conversion import convert_values
 from jedi.inference.analysis import Warning
 
+test_dir = Path(__file__).absolute().parent
 
 TEST_COMPLETIONS = 0
 TEST_INFERENCE = 1
@@ -173,6 +178,7 @@ class IntegrationTestCase(BaseTestCase):
         self.start = start
         self.line = line
         self.path = path
+        self._project = jedi.Project(test_dir)
 
     @property
     def module_name(self):
@@ -188,7 +194,12 @@ class IntegrationTestCase(BaseTestCase):
                                    self.line_nr_test, self.line.rstrip())
 
     def script(self, environment):
-        return jedi.Script(self.source, path=self.path, environment=environment)
+        return jedi.Script(
+            self.source,
+            path=self.path,
+            environment=environment,
+            project=self._project
+        )
 
     def run(self, compare_cb, environment=None):
         testers = {
@@ -198,7 +209,7 @@ class IntegrationTestCase(BaseTestCase):
             TEST_REFERENCES: self.run_get_references,
         }
         if (self.path.endswith('pytest.py') or self.path.endswith('conftest.py')) \
-                and environment.executable != os.path.realpath(sys.executable):
+                and os.path.realpath(environment.executable) != os.path.realpath(sys.executable):
             # It's not guarantueed that pytest is installed in test
             # environments, if we're not running in the same environment that
             # we're already in, so just skip that case.
@@ -263,7 +274,7 @@ class IntegrationTestCase(BaseTestCase):
         self.correct = self.correct.strip()
         compare = sorted(
             (('stub:' if r.is_stub() else '')
-             + re.sub(r'^test\.completion\.', '', r.module_name),
+             + re.sub(r'^completion\.', '', r.module_name),
              r.line,
              r.column)
             for r in result

test/test_api/test_api.py

@@ -321,10 +321,19 @@ def test_docstrings_for_completions(Script):
         assert isinstance(c.docstring(), str)
 
 
+def test_completions_order_most_resemblance_on_top(Script):
+    """Test that the completion which resembles the in-typing the most will come first."""
+    code = "from pathlib import Path\npath = Path('hello.txt')\n\npat"
+    script = Script(code)
+    # User is typing "pat" and "path" is closer to it than "Path".
+    assert ['path', 'Path'] == [comp.name for comp in script.complete()]
+
+
 def test_fuzzy_completion(Script):
     script = Script('string =  "hello"\nstring.upper')
-    assert ['isupper',
-            'upper'] == [comp.name for comp in script.complete(fuzzy=True)]
+    # 'isupper' is included because it is fuzzily matched.
+    assert ['upper',
+            'isupper'] == [comp.name for comp in script.complete(fuzzy=True)]
 
 
 def test_math_fuzzy_completion(Script, environment):

test/test_api/test_call_signatures.py

@@ -650,6 +650,7 @@ def test_cursor_after_signature(Script, column):
         ('abs(chr  ( \nclass y: pass', 1, 8, 'abs', 0),
         ('abs(chr  ( \nclass y: pass', 1, 9, 'abs', 0),
         ('abs(chr  ( \nclass y: pass', 1, 10, 'chr', 0),
+        ('abs(foo.bar=3)', 1, 13, 'abs', 0),
     ]
 )
 def test_base_signatures(Script, code, line, column, name, index):

test/test_api/test_classes.py

@@ -188,10 +188,7 @@ def test_functions_should_have_params(Script):
                 assert c.get_signatures()
 
 
-def test_hashlib_params(Script, environment):
-    if environment.version_info < (3,):
-        pytest.skip()
-
+def test_hashlib_params(Script):
     script = Script('from hashlib import sha256')
     c, = script.complete()
     sig, = c.get_signatures()
@@ -351,8 +348,8 @@ def test_parent_on_comprehension(Script):
 
 def test_type(Script):
     for c in Script('a = [str()]; a[0].').complete():
-        if c.name == '__class__' and False:  # TODO fix.
-            assert c.type == 'class'
+        if c.name == '__class__':
+            assert c.type == 'property'
         else:
             assert c.type in ('function', 'statement')
 

test/test_api/test_completion.py

@@ -1,11 +1,16 @@
 from os.path import join, sep as s, dirname, expanduser
 import os
 from textwrap import dedent
+from itertools import count
+from pathlib import Path
 
 import pytest
 
 from ..helpers import root_dir
 from jedi.api.helpers import _start_match, _fuzzy_match
+from jedi.inference.imports import _load_python_module
+from jedi.file_io import KnownContentFileIO
+from jedi.inference.base_value import ValueSet
 
 
 def test_in_whitespace(Script):
@@ -400,6 +405,22 @@ def test_ellipsis_completion(Script):
     assert Script('...').complete() == []
 
 
+@pytest.fixture
+def module_injector():
+    counter = count()
+
+    def module_injector(inference_state, names, code):
+        assert isinstance(names, tuple)
+        file_io = KnownContentFileIO(
+            Path('foo/bar/module-injector-%s.py' % next(counter)).absolute(),
+            code
+        )
+        v = _load_python_module(inference_state, file_io, names)
+        inference_state.module_cache.add(names, ValueSet([v]))
+
+    return module_injector
+
+
 def test_completion_cache(Script, module_injector):
     """
     For some modules like numpy, tensorflow or pandas we cache docstrings and
@@ -436,3 +457,7 @@ def test_module_completions(Script, module):
         # Just make sure that there are no errors
         c.type
         c.docstring()
+
+
+def test_whitespace_at_end_after_dot(Script):
+    assert 'strip' in [c.name for c in Script('str. ').complete()]

test/test_api/test_context.py

@@ -17,7 +17,7 @@ def func1(x, y):
 def func2():
  what ?
 i = 3
-    
+
 def func3():
     1'''
 cls_code = '''\

test/test_api/test_documentation.py

@@ -28,6 +28,11 @@ def test_import_keyword(Script):
     # unrelated to #44
 
 
+def test_import_keyword_after_newline(Script):
+    d, = Script("import x\nimport y").help(line=2, column=0)
+    assert d.docstring().startswith('The "import" statement')
+
+
 def test_import_keyword_with_gotos(goto_or_infer):
     assert not goto_or_infer("import x", column=0)
 
@@ -37,6 +42,17 @@ def test_operator_doc(Script):
     assert len(d.docstring()) > 100
 
 
+@pytest.mark.parametrize(
+    'code, help_part', [
+        ('str', 'Create a new string object'),
+        ('str.strip', 'Return a copy of the string'),
+    ]
+)
+def test_stdlib_doc(Script, code, help_part):
+    h, = Script(code).help()
+    assert help_part in h.docstring(raw=True)
+
+
 def test_lambda(Script):
     d, = Script('lambda x: x').help(column=0)
     assert d.type == 'keyword'

test/test_api/test_interpreter.py

@@ -8,6 +8,7 @@ import typing
 import pytest
 
 import jedi
+import jedi.settings
 from jedi.inference.compiled import mixed
 from importlib import import_module
 
@@ -25,7 +26,7 @@ def get_completion(source, namespace):
 
 
 def test_builtin_details():
-    import keyword
+    import keyword  # noqa: F401
 
     class EmptyClass:
         pass
@@ -52,9 +53,9 @@ def test_numpy_like_non_zero():
     """
 
     class NumpyNonZero:
-
         def __zero__(self):
             raise ValueError('Numpy arrays would raise and tell you to use .any() or all()')
+
         def __bool__(self):
             raise ValueError('Numpy arrays would raise and tell you to use .any() or all()')
 
@@ -101,26 +102,28 @@ def test_side_effect_completion():
     assert foo.name == 'foo'
 
 
-def _assert_interpreter_complete(source, namespace, completions,
-                                 **kwds):
+def _assert_interpreter_complete(source, namespace, completions, *, check_type=False, **kwds):
     script = jedi.Interpreter(source, [namespace], **kwds)
     cs = script.complete()
     actual = [c.name for c in cs]
+    if check_type:
+        for c in cs:
+            c.type
     assert sorted(actual) == sorted(completions)
 
 
 def test_complete_raw_function():
-    from os.path import join
+    from os.path import join  # noqa: F401
     _assert_interpreter_complete('join("").up', locals(), ['upper'])
 
 
 def test_complete_raw_function_different_name():
-    from os.path import join as pjoin
+    from os.path import join as pjoin  # noqa: F401
     _assert_interpreter_complete('pjoin("").up', locals(), ['upper'])
 
 
 def test_complete_raw_module():
-    import os
+    import os  # noqa: F401
     _assert_interpreter_complete('os.path.join("a").up', locals(), ['upper'])
 
 
@@ -219,7 +222,7 @@ def test__getattr__completions(allow_unsafe_getattr, class_is_findable):
 
 @pytest.fixture(params=[False, True])
 def allow_unsafe_getattr(request, monkeypatch):
-    monkeypatch.setattr(jedi.Interpreter, '_allow_descriptor_getattr_default', request.param)
+    monkeypatch.setattr(jedi.settings, 'allow_unsafe_interpreter_executions', request.param)
     return request.param
 
 
@@ -278,7 +281,7 @@ def test_param_completion():
     def foo(bar):
         pass
 
-    lambd = lambda xyz: 3
+    lambd = lambda xyz: 3  # noqa: E731
 
     _assert_interpreter_complete('foo(bar', locals(), ['bar='])
     _assert_interpreter_complete('lambd(xyz', locals(), ['xyz='])
@@ -292,7 +295,7 @@ def test_endless_yield():
 
 
 def test_completion_params():
-    foo = lambda a, b=3: None
+    foo = lambda a, b=3: None  # noqa: E731
 
     script = jedi.Interpreter('foo', [locals()])
     c, = script.complete()
@@ -307,8 +310,9 @@ def test_completion_param_annotations():
     # Need to define this function not directly in Python. Otherwise Jedi is too
     # clever and uses the Python code instead of the signature object.
     code = 'def foo(a: 1, b: str, c: int = 1.0) -> bytes: pass'
-    exec(code, locals())
-    script = jedi.Interpreter('foo', [locals()])
+    exec_locals = {}
+    exec(code, exec_locals)
+    script = jedi.Interpreter('foo', [exec_locals])
     c, = script.complete()
     sig, = c.get_signatures()
     a, b, c = sig.params
@@ -320,7 +324,7 @@ def test_completion_param_annotations():
     assert b.description == 'param b: str'
     assert c.description == 'param c: int=1.0'
 
-    d, = jedi.Interpreter('foo()', [locals()]).infer()
+    d, = jedi.Interpreter('foo()', [exec_locals]).infer()
     assert d.name == 'bytes'
 
 
@@ -406,7 +410,7 @@ def test_dir_magic_method(allow_unsafe_getattr):
 def test_name_not_findable():
     class X():
         if 0:
-            NOT_FINDABLE
+            NOT_FINDABLE  # noqa: F821
 
         def hidden(self):
             return
@@ -419,7 +423,7 @@ def test_name_not_findable():
 
 
 def test_stubs_working():
-    from multiprocessing import cpu_count
+    from multiprocessing import cpu_count  # noqa: F401
     defs = jedi.Interpreter("cpu_count()", [locals()]).infer()
     assert [d.name for d in defs] == ['int']
 
@@ -522,14 +526,17 @@ def test_partial_signatures(code, expected, index):
     c = functools.partial(func, 1, c=2)
 
     sig, = jedi.Interpreter(code, [locals()]).get_signatures()
-    assert sig.name == 'partial'
     assert [p.name for p in sig.params] == expected
     assert index == sig.index
 
+    if sys.version_info < (3, 13):
+        # Python 3.13.0b3 makes functools.partial be a descriptor, which breaks
+        # Jedi's `py__name__` detection; see https://github.com/davidhalter/jedi/issues/2012
+        assert sig.name == 'partial'
+
 
 def test_type_var():
     """This was an issue before, see Github #1369"""
-    import typing
     x = typing.TypeVar('myvar')
     def_, = jedi.Interpreter('x', [locals()]).infer()
     assert def_.name == 'TypeVar'
@@ -603,16 +610,19 @@ def test_dict_getitem(code, types):
 @pytest.mark.parametrize(
     'code, expected', [
         ('DunderCls()[0]', 'int'),
+        ('dunder[0]', 'int'),
         ('next(DunderCls())', 'float'),
+        ('next(dunder)', 'float'),
         ('for x in DunderCls(): x', 'str'),
+        # ('for x in dunder: x', 'str'),
     ]
 )
-def test_dunders(class_is_findable, code, expected):
+def test_dunders(class_is_findable, code, expected, allow_unsafe_getattr):
     from typing import Iterator
 
     class DunderCls:
         def __getitem__(self, key) -> int:
-            pass
+            return 1
 
         def __iter__(self, key) -> Iterator[str]:
             pass
@@ -623,6 +633,8 @@ def test_dunders(class_is_findable, code, expected):
     if not class_is_findable:
         DunderCls.__name__ = 'asdf'
 
+    dunder = DunderCls()
+
     n, = jedi.Interpreter(code, [locals()]).infer()
     assert n.name == expected
 
@@ -651,10 +663,12 @@ def bar():
         ({'return': 'typing.Union[str, int]'}, ['int', 'str'], ''),
         ({'return': 'typing.Union["str", int]'},
          ['int', 'str'] if sys.version_info >= (3, 9) else ['int'], ''),
-        ({'return': 'typing.Union["str", 1]'}, [], ''),
+        ({'return': 'typing.Union["str", 1]'},
+         ['str'] if sys.version_info >= (3, 11) else [], ''),
         ({'return': 'typing.Optional[str]'}, ['NoneType', 'str'], ''),
         ({'return': 'typing.Optional[str, int]'}, [], ''),  # Takes only one arg
-        ({'return': 'typing.Any'}, [], ''),
+        ({'return': 'typing.Any'},
+         ['_AnyMeta'] if sys.version_info >= (3, 11) else [], ''),
 
         ({'return': 'typing.Tuple[int, str]'},
          ['Tuple' if sys.version_info[:2] == (3, 6) else 'tuple'], ''),
@@ -677,7 +691,7 @@ def bar():
     ]
 )
 def test_string_annotation(annotations, result, code):
-    x = lambda foo: 1
+    x = lambda foo: 1  # noqa: E731
     x.__annotations__ = annotations
     defs = jedi.Interpreter(code or 'x()', [locals()]).infer()
     assert [d.name for d in defs] == result
@@ -706,3 +720,142 @@ def test_negate():
     assert x.name == 'int'
     value, = x._name.infer()
     assert value.get_safe_value() == -3
+
+
+def test_complete_not_findable_class_source():
+    class TestClass():
+        ta = 1
+        ta1 = 2
+
+    # Simulate the environment where the class is defined in
+    # an interactive session and therefore inspect module
+    # cannot find its source code and raises OSError (Py 3.10+) or TypeError.
+    TestClass.__module__ = "__main__"
+    # There is a pytest __main__ module we have to remove temporarily.
+    module = sys.modules.pop("__main__")
+    try:
+        interpreter = jedi.Interpreter("TestClass.", [locals()])
+        completions = interpreter.complete(column=10, line=1)
+    finally:
+        sys.modules["__main__"] = module
+
+    assert "ta" in [c.name for c in completions]
+    assert "ta1" in [c.name for c in completions]
+
+
+def test_param_infer_default():
+    abs_sig, = jedi.Interpreter('abs(', [{'abs': abs}]).get_signatures()
+    param, = abs_sig.params
+    assert param.name == 'x'
+    assert param.infer_default() == []
+
+
+@pytest.mark.parametrize(
+    'code, expected', [
+        ("random.triangular(", ['high=', 'low=', 'mode=']),
+        ("random.triangular(low=1, ", ['high=', 'mode=']),
+        ("random.triangular(high=1, ", ['low=', 'mode=']),
+        ("random.triangular(low=1, high=2, ", ['mode=']),
+        ("random.triangular(low=1, mode=2, ", ['high=']),
+    ],
+)
+def test_keyword_param_completion(code, expected):
+    import random  # noqa: F401
+    completions = jedi.Interpreter(code, [locals()]).complete()
+    assert expected == [c.name for c in completions if c.name.endswith('=')]
+
+
+@pytest.mark.parametrize('class_is_findable', [False, True])
+def test_avoid_descriptor_executions_if_not_necessary(class_is_findable):
+    counter = 0
+
+    class AvoidDescriptor(object):
+        @property
+        def prop(self):
+            nonlocal counter
+            counter += 1
+            return self
+
+    if not class_is_findable:
+        AvoidDescriptor.__name__ = "something_somewhere"
+    namespace = {'b': AvoidDescriptor()}
+    expected = ['prop']
+    _assert_interpreter_complete('b.pro', namespace, expected, check_type=True)
+    assert counter == 0
+    _assert_interpreter_complete('b.prop.pro', namespace, expected, check_type=True)
+    assert counter == 1
+
+
+class Hello:
+    its_me = 1
+
+
+@pytest.mark.parametrize('class_is_findable', [False, True])
+def test_try_to_use_return_annotation_for_property(class_is_findable):
+    class WithProperties(object):
+        @property
+        def with_annotation1(self) -> str:
+            raise BaseException
+
+        @property
+        def with_annotation2(self) -> 'str':
+            raise BaseException
+
+        @property
+        def with_annotation3(self) -> Hello:
+            raise BaseException
+
+        @property
+        def with_annotation4(self) -> 'Hello':
+            raise BaseException
+
+        @property
+        def with_annotation_garbage1(self) -> 'asldjflksjdfljdslkjfsl':  # noqa
+            return Hello()
+
+        @property
+        def with_annotation_garbage2(self) -> 'sdf$@@$5*+8':  # noqa
+            return Hello()
+
+        @property
+        def without_annotation(self):
+            return ""
+
+    if not class_is_findable:
+        WithProperties.__name__ = "something_somewhere"
+        Hello.__name__ = "something_somewhere_else"
+
+    namespace = {'p': WithProperties()}
+    _assert_interpreter_complete('p.without_annotation.upp', namespace, ['upper'])
+    _assert_interpreter_complete('p.with_annotation1.upp', namespace, ['upper'])
+    _assert_interpreter_complete('p.with_annotation2.upp', namespace, ['upper'])
+    _assert_interpreter_complete('p.with_annotation3.its', namespace, ['its_me'])
+    _assert_interpreter_complete('p.with_annotation4.its', namespace, ['its_me'])
+    # This is a fallback, if the annotations don't help
+    _assert_interpreter_complete('p.with_annotation_garbage1.its', namespace, ['its_me'])
+    _assert_interpreter_complete('p.with_annotation_garbage2.its', namespace, ['its_me'])
+
+
+def test_nested__getitem__():
+    d = {'foo': {'bar': 1}}
+    _assert_interpreter_complete('d["fo', locals(), ['"foo"'])
+    _assert_interpreter_complete('d["foo"]["ba', locals(), ['"bar"'])
+    _assert_interpreter_complete('(d["foo"])["ba', locals(), ['"bar"'])
+    _assert_interpreter_complete('((d["foo"]))["ba', locals(), ['"bar"'])
+
+
+@pytest.mark.parametrize('class_is_findable', [False, True])
+def test_custom__getitem__(class_is_findable, allow_unsafe_getattr):
+    class CustomGetItem:
+        def __getitem__(self, x: int):
+            return "asdf"
+
+    if not class_is_findable:
+        CustomGetItem.__name__ = "something_somewhere"
+
+    namespace = {'c': CustomGetItem()}
+    if not class_is_findable and not allow_unsafe_getattr:
+        expected = []
+    else:
+        expected = ['upper']
+    _assert_interpreter_complete('c["a"].up', namespace, expected)

test/test_api/test_names.py

@@ -189,3 +189,9 @@ def test_no_error(get_names):
 def test_is_side_effect(get_names, code, index, is_side_effect):
     names = get_names(code, references=True, all_scopes=True)
     assert names[index].is_side_effect() == is_side_effect
+
+
+def test_no_defined_names(get_names):
+    definition, = get_names("x = (1, 2)")
+
+    assert not definition.defined_names()

test/test_api/test_project.py

@@ -68,6 +68,10 @@ def test_load_save_project(tmpdir):
          dict(all_scopes=True)),
         ('some_search_test_var', ['test_api.test_project.test_search.some_search_test_var'],
          dict(complete=True, all_scopes=True)),
+        # Make sure that the searched name is not part of the file, by
+        # splitting it up.
+        ('some_search_test_v' + 'a', ['test_api.test_project.test_search.some_search_test_var'],
+         dict(complete=True, all_scopes=True)),
 
         ('sample_int', ['helpers.sample_int'], {}),
         ('sample_int', ['helpers.sample_int'], dict(all_scopes=True)),
@@ -140,13 +144,17 @@ def test_load_save_project(tmpdir):
     ]
 )
 def test_search(string, full_names, kwargs):
-    some_search_test_var = 1.0
+    some_search_test_var = 1.0  # noqa: F841
     project = Project(test_dir)
     if kwargs.pop('complete', False) is True:
         defs = project.complete_search(string, **kwargs)
     else:
         defs = project.search(string, **kwargs)
-    assert sorted([('stub:' if d.is_stub() else '') + d.full_name for d in defs]) == full_names
+    actual = sorted([
+        ('stub:' if d.is_stub() else '') + (d.full_name or d.name)
+        for d in defs
+    ])
+    assert actual == full_names
 
 
 @pytest.mark.parametrize(
@@ -165,8 +173,8 @@ def test_complete_search(Script, string, completions, all_scopes):
 
 @pytest.mark.parametrize(
     'path,expected', [
-        (Path(__file__).parents[2], True), # The path of the project
-        (Path(__file__).parents[1], False), # The path of the tests, not a project
+        (Path(__file__).parents[2], True),  # The path of the project
+        (Path(__file__).parents[1], False),  # The path of the tests, not a project
         (Path.home(), None)
     ]
 )
@@ -174,7 +182,7 @@ def test_is_potential_project(path, expected):
 
     if expected is None:
         try:
-            expected = _CONTAINS_POTENTIAL_PROJECT in os.listdir(path)
+            expected = bool(set(_CONTAINS_POTENTIAL_PROJECT) & set(os.listdir(path)))
         except OSError:
             expected = False
 

test/test_api/test_refactoring.py

@@ -1,10 +1,13 @@
 import os
+import shutil
 from textwrap import dedent
 from pathlib import Path
+import platform
 
 import pytest
 
 import jedi
+from test.helpers import get_example_dir
 
 
 @pytest.fixture()
@@ -51,6 +54,47 @@ def test_rename_mod(Script, dir_with_content):
         ''').format(dir=dir_with_content)
 
 
+@pytest.mark.skipif('sys.version_info[:2] < (3, 8)', message="Python 3.8 introduces dirs_exist_ok")
+def test_namespace_package(Script, tmpdir):
+    origin = get_example_dir('implicit_namespace_package')
+    shutil.copytree(origin, tmpdir.strpath, dirs_exist_ok=True)
+    sys_path = [
+        os.path.join(tmpdir.strpath, 'ns1'),
+        os.path.join(tmpdir.strpath, 'ns2')
+    ]
+    script_path = os.path.join(tmpdir.strpath, 'script.py')
+    script = Script(
+        'import pkg\n',
+        path=script_path,
+        project=jedi.Project(os.path.join(tmpdir.strpath, 'does-not-exist'), sys_path=sys_path),
+    )
+    refactoring = script.rename(line=1, new_name='new_pkg')
+    refactoring.apply()
+    old1 = os.path.join(sys_path[0], "pkg")
+    new1 = os.path.join(sys_path[0], "new_pkg")
+    old2 = os.path.join(sys_path[1], "pkg")
+    new2 = os.path.join(sys_path[1], "new_pkg")
+    assert not os.path.exists(old1)
+    assert os.path.exists(new1)
+    assert not os.path.exists(old2)
+    assert os.path.exists(new2)
+
+    changed, = iter(refactoring.get_changed_files().values())
+    assert changed.get_new_code() == "import new_pkg\n"
+
+    assert refactoring.get_diff() == dedent(f'''\
+        rename from {old1}
+        rename to {new1}
+        rename from {old2}
+        rename to {new2}
+        --- {script_path}
+        +++ {script_path}
+        @@ -1,2 +1,2 @@
+        -import pkg
+        +import new_pkg
+        ''').format(dir=dir_with_content)
+
+
 def test_rename_none_path(Script):
     refactoring = Script('foo', path=None).rename(new_name='bar')
     with pytest.raises(jedi.RefactoringError, match='on a Script with path=None'):
@@ -69,4 +113,24 @@ def test_diff_without_ending_newline(Script):
          b
         -a
         +c
+        ''')  # noqa: W291
+
+
+def test_diff_path_outside_of_project(Script):
+    if platform.system().lower() == 'windows':
+        abs_path = r'D:\unknown_dir\file.py'
+    else:
+        abs_path = '/unknown_dir/file.py'
+    script = Script(
+        code='foo = 1',
+        path=abs_path,
+        project=jedi.get_default_project()
+    )
+    diff = script.rename(line=1, column=0, new_name='bar').get_diff()
+    assert diff == dedent(f'''\
+        --- {abs_path}
+        +++ {abs_path}
+        @@ -1 +1 @@
+        -foo = 1
+        +bar = 1
         ''')