docs: Add semantic release information and issue template

chore: Add Semantic Release GitHub action
Install Semantic Release
2021-11-16 22:42:31 -08:00 · 2021-11-16 22:42:03 -08:00 · 2021-11-16 22:30:16 -08:00 · 2021-11-05 14:22:15 +00:00 · 2021-10-26 00:34:33 +01:00 · 2021-08-25 12:53:02 +02:00
53 changed files with 7851 additions and 1639 deletions
--- a/.github/PULL_REQUEST_TEMPLATE.md
+++ b/.github/PULL_REQUEST_TEMPLATE.md
@@ -0,0 +1,9 @@
+## Description
+
+<!-- Thanks for contributing to betterproto! Add a thorough explanation of what your changes do below this line: -->
+
+## Checklist
+
+- [ ] This PR targets the `rc` branch (**not** `master`).
+- [ ] [If this should release a new version to PyPI when merged] The title of the PR follows the [Angular Conventional Commit](https://www.conventionalcommits.org/) syntax (`feat:` or `fix:`, with `BREAKING CHANGE:` in the commit message body if appropriate), and clearly describes the fix or feature.
+- [ ] Documentation is updated (`README.md` and docstrings).
--- a/.github/workflows/release.yml
+++ b/.github/workflows/release.yml
@@ -16,12 +16,18 @@ jobs:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v2
+        with:
+          persist-credentials: false
      - name: Set up Python 3.8
        uses: actions/setup-python@v2
        with:
          python-version: 3.8
      - name: Install poetry
        run: python -m pip install poetry
+      - name: Semantic Release
+        uses: cycjimmy/semantic-release-action@v2
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
      - name: Build package
        run: poetry build
      - name: Publish package to PyPI
--- a/.gitignore
+++ b/.gitignore
@@ -6,6 +6,7 @@
 .pytest_cache
 .python-version
 build/
+node_modules/
 tests/output_*
 **/__pycache__
 dist
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,25 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0

 - Versions suffixed with `b*` are in `beta` and can be installed with `pip install --pre betterproto`.

+## [2.0.0b3] - 2021-04-07
+
+- Generate grpclib service stubs [#170](https://github.com/danielgtaylor/python-betterproto/pull/170)
+- Add \_\_version\_\_ attribute to package [#134](https://github.com/danielgtaylor/python-betterproto/pull/134)
+- Use betterproto generated messages in the plugin [#161](https://github.com/danielgtaylor/python-betterproto/pull/161)
+- Sort the list of sources in generated file headers [#164](https://github.com/danielgtaylor/python-betterproto/pull/164)
+- Micro-optimization: use tuples instead of lists for conditions [#228](https://github.com/danielgtaylor/python-betterproto/pull/228)
+- Improve datestring parsing [#213](https://github.com/danielgtaylor/python-betterproto/pull/213)
+
+- Fix serialization of repeated fields with empty messages [#180](https://github.com/danielgtaylor/python-betterproto/pull/180)
+- Fix compilation of fields named 'bytes' or 'str' [#226](https://github.com/danielgtaylor/python-betterproto/pull/226)
+- Fix json serialization of infinite and nan floats/doubles [#215](https://github.com/danielgtaylor/python-betterproto/pull/215)
+- Fix template bug resulting in empty \_\_post_init\_\_ methods [#162](https://github.com/danielgtaylor/python-betterproto/pull/162)
+- Fix serialization of zero-value messages in a oneof group [#176](https://github.com/danielgtaylor/python-betterproto/pull/176)
+- Fix missing typing and datetime imports [#183](https://github.com/danielgtaylor/python-betterproto/pull/183)
+- Fix code generation for empty services [#222](https://github.com/danielgtaylor/python-betterproto/pull/222)
+- Fix Message.to_dict and from_dict handling of repeated timestamps and durations [#211](https://github.com/danielgtaylor/python-betterproto/pull/211)
+- Fix incorrect routes in generated client when service is not in a package [#177](https://github.com/danielgtaylor/python-betterproto/pull/177)
+
 ## [2.0.0b2] - 2020-11-24

 - Add support for deprecated message and fields [#126](https://github.com/danielgtaylor/python-betterproto/pull/126)
--- a/README.md
+++ b/README.md
@@ -1,6 +1,7 @@
 # Better Protobuf / gRPC Support for Python

 ![](https://github.com/danielgtaylor/python-betterproto/workflows/CI/badge.svg)
+> :octocat: If you're reading this on github, please be aware that it might mention unreleased features! See the latest released README on [pypi](https://pypi.org/project/betterproto/).

 This project aims to provide an improved experience when using Protobuf / gRPC in a modern Python environment by making use of modern language features and generating readable, understandable, idiomatic Python code. It will not support legacy features or environments (e.g. Protobuf 2). The following are supported:

@@ -159,6 +160,12 @@ service Echo {
 }
 ```

+Generate echo proto file:
+
+```
+python -m grpc_tools.protoc -I . --python_betterproto_out=. echo.proto
+```
+
 A client can be implemented as follows:
 ```python
 import asyncio
@@ -192,6 +199,37 @@ EchoStreamResponse(value='hello')
 EchoStreamResponse(value='hello')
 ```

+This project also produces server-facing stubs that can be used to implement a Python
+gRPC server.
+To use them, simply subclass the base class in the generated files and override the
+service methods:
+
+```python
+import asyncio
+from echo import EchoBase, EchoResponse, EchoStreamResponse
+from grpclib.server import Server
+from typing import AsyncIterator
+
+
+class EchoService(EchoBase):
+    async def echo(self, value: str, extra_times: int) -> "EchoResponse":
+        return EchoResponse([value for _ in range(extra_times)])
+
+    async def echo_stream(self, value: str, extra_times: int) -> AsyncIterator["EchoStreamResponse"]:
+        for _ in range(extra_times):
+            yield EchoStreamResponse(value)
+
+
+async def main():
+    server = Server([EchoService()])
+    await server.start("127.0.0.1", 50051)
+    await server.wait_closed()
+
+if __name__ == '__main__':
+    loop = asyncio.get_event_loop()
+    loop.run_until_complete(main())
+```
+
 ### JSON

 Both serializing and parsing are supported to/from JSON and Python dictionaries using the following methods:
@@ -413,9 +451,9 @@ Assuming your `google.protobuf` source files (included with all releases of `pro

 ```sh
 protoc \
-    --plugin=protoc-gen-custom=betterproto/plugin.py \
+    --plugin=protoc-gen-custom=src/betterproto/plugin/main.py \
    --custom_opt=INCLUDE_GOOGLE \
-    --custom_out=betterproto/lib \
+    --custom_out=src/betterproto/lib \
    -I /usr/local/include/ \
    /usr/local/include/google/protobuf/*.proto
 ```
@@ -460,6 +498,16 @@ protoc \
 - [x] Automate running tests
 - [ ] Cleanup!

+
+## Release
+
+New versions are versioned and released using [Semantic Release](https://github.com/semantic-release/semantic-release). When new commits
+using the Angular Conventional Commits syntax land on `master`  or `rc`, those commits are used to determine what new version to release.
+
+All Pull Requests must target the `rc` branch; when merged into `rc` they will publish new release candidate (`rc`) versions to PyPI
+automatically. When maintainers want to publish a new full release, they simply merge `rc` into `master`. This flow ensures that features
+and fixes are published quickly and continuously rather than awaiting a manual release process.
+
 ## Community

 Join us on [Slack](https://join.slack.com/t/betterproto/shared_invite/zt-f0n0uolx-iN8gBNrkPxtKHTLpG3o1OQ)!
--- a/benchmarks/benchmarks.py
+++ b/benchmarks/benchmarks.py
@@ -1,6 +1,8 @@
 import betterproto
 from dataclasses import dataclass

+from typing import List
+

@dataclass
 class TestMessage(betterproto.Message):
@@ -9,6 +11,29 @@ class TestMessage(betterproto.Message):
    baz: float = betterproto.float_field(2)


+@dataclass
+class TestNestedChildMessage(betterproto.Message):
+    str_key: str = betterproto.string_field(0)
+    bytes_key: bytes = betterproto.bytes_field(1)
+    bool_key: bool = betterproto.bool_field(2)
+    float_key: float = betterproto.float_field(3)
+    int_key: int = betterproto.uint64_field(4)
+
+
+@dataclass
+class TestNestedMessage(betterproto.Message):
+    foo: TestNestedChildMessage = betterproto.message_field(0)
+    bar: TestNestedChildMessage = betterproto.message_field(1)
+    baz: TestNestedChildMessage = betterproto.message_field(2)
+
+
+@dataclass
+class TestRepeatedMessage(betterproto.Message):
+    foo_repeat: List[str] = betterproto.string_field(0)
+    bar_repeat: List[int] = betterproto.int64_field(1)
+    baz_repeat: List[bool] = betterproto.bool_field(2)
+
+
 class BenchMessage:
    """Test creation and usage a proto message."""

@@ -16,6 +41,30 @@ class BenchMessage:
        self.cls = TestMessage
        self.instance = TestMessage()
        self.instance_filled = TestMessage(0, "test", 0.0)
+        self.instance_filled_bytes = bytes(self.instance_filled)
+        self.instance_filled_nested = TestNestedMessage(
+            TestNestedChildMessage("foo", bytearray(b"test1"), True, 0.1234, 500),
+            TestNestedChildMessage("bar", bytearray(b"test2"), True, 3.1415, -302),
+            TestNestedChildMessage("baz", bytearray(b"test3"), False, 1e5, 300),
+        )
+        self.instance_filled_nested_bytes = bytes(self.instance_filled_nested)
+        self.instance_filled_repeated = TestRepeatedMessage(
+            [
+                "test1",
+                "test2",
+                "test3",
+                "test4",
+                "test5",
+                "test6",
+                "test7",
+                "test8",
+                "test9",
+                "test10",
+            ],
+            [2, -100, 0, 500000, 600, -425678, 1000000000, -300, 1, -694214214466],
+            [True, False, False, False, True, True, False, True, False, False],
+        )
+        self.instance_filled_repeated_bytes = bytes(self.instance_filled_repeated)

    def time_overhead(self):
        """Overhead in class definition."""
@@ -50,6 +99,26 @@ class BenchMessage:
        """Time serializing a message to wire."""
        bytes(self.instance_filled)

+    def time_deserialize(self):
+        """Time deserialize a message."""
+        TestMessage().parse(self.instance_filled_bytes)
+
+    def time_serialize_nested(self):
+        """Time serializing a nested message to wire."""
+        bytes(self.instance_filled_nested)
+
+    def time_deserialize_nested(self):
+        """Time deserialize a nested message."""
+        TestNestedMessage().parse(self.instance_filled_nested_bytes)
+
+    def time_serialize_repeated(self):
+        """Time serializing a repeated message to wire."""
+        bytes(self.instance_filled_repeated)
+
+    def time_deserialize_repeated(self):
+        """Time deserialize a repeated message."""
+        TestRepeatedMessage().parse(self.instance_filled_repeated_bytes)
+

 class MemSuite:
    def setup(self):
--- a/docs/index.rst
+++ b/docs/index.rst
@@ -12,7 +12,7 @@ Features:
 - Generated messages are both binary & JSON serializable
 - Messages use relevant python types, e.g. ``Enum``, ``datetime`` and ``timedelta``
  objects
- ``async``/``await`` support for gRPC Clients
+- ``async``/``await`` support for gRPC Clients and Servers
 - Generates modern, readable, idiomatic python code

 Contents:
--- a/docs/quick-start.rst
+++ b/docs/quick-start.rst
@@ -100,7 +100,7 @@ Async gRPC Support
 ++++++++++++++++++

 The generated code includes `grpclib <https://grpclib.readthedocs.io/en/latest>`_ based
-stub (client) classes for rpc services declared in the input proto files.
+stub (client and server) classes for rpc services declared in the input proto files.
 It is enabled by default.


@@ -160,6 +160,36 @@ The generated client can be used like so:
    EchoStreamResponse(value='hello')


+The server-facing stubs can be used to implement a Python
+gRPC server.
+To use them, simply subclass the base class in the generated files and override the
+service methods:
+
+.. code-block:: python
+
+    from echo import EchoBase
+    from grpclib.server import Server
+    from typing import AsyncIterator
+
+
+    class EchoService(EchoBase):
+        async def echo(self, value: str, extra_times: int) -> "EchoResponse":
+            return value
+
+        async def echo_stream(
+            self, value: str, extra_times: int
+        ) -> AsyncIterator["EchoStreamResponse"]:
+            for _ in range(extra_times):
+                yield value
+
+
+    async def start_server():
+        HOST = "127.0.0.1"
+        PORT = 1337
+        server = Server([EchoService()])
+        await server.start(HOST, PORT)
+        await server.serve_forever()
+
 JSON
 ++++
 Message objects include :meth:`betterproto.Message.to_json` and
--- a/package.json
+++ b/package.json
@@ -0,0 +1,52 @@
+{
+    "name": "python-betterproto-semantic-release",
+    "version": "1.0.0",
+    "description": "Encapsulate dependencies needed to use semantic-release",
+    "dependencies": {
+        "@semantic-release/exec": "^5.0.0",
+        "@semantic-release/git": "^9.0.0",
+        "@semantic-release/gitlab": "^6.0.4",
+        "conventional-changelog-eslint": "^3.0.8",
+        "semantic-release": "^17.1.1"
+    },
+    "release": {
+        "branches": [
+            "master",
+            {
+                "name": "rc",
+                "prerelease": true
+            }
+        ],
+        "plugins": [
+            [
+                "@semantic-release/commit-analyzer",
+                {
+                    "preset": "angular"
+                }
+            ],
+            [
+                "@semantic-release/release-notes-generator",
+                {
+                    "preset": "angular"
+                }
+            ],
+            [
+                "@semantic-release/exec",
+                {
+                    "prepareCmd": "poetry version ${nextRelease.version}"
+                }
+            ],
+            "@semantic-release/github",
+            [
+                "@semantic-release/git",
+                {
+                    "assets": [
+                        "pyproject.toml"
+                    ],
+                    "message": "Release v${nextRelease.version} [skip ci]"
+                }
+            ]
+        ],
+        "repositoryUrl": "ssh://git@github.com/danielgtaylor/python-betterproto.git"
+    }
+}
--- a/poetry.lock
+++ b/poetry.lock
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "betterproto"
-version = "2.0.0b2"
+version = "2.0.0b3"
 description = "A better Protobuf / gRPC generator & library"
 authors = ["Daniel G. Taylor <danielgtaylor@gmail.com>"]
 readme = "README.md"
@@ -13,49 +13,93 @@ packages = [

 [tool.poetry.dependencies]
 python = "^3.6"
-backports-datetime-fromisoformat = { version = "^1.0.0", python = "<3.7" }
 black = { version = ">=19.3b0", optional = true }
 dataclasses = { version = "^0.7", python = ">=3.6, <3.7" }
 grpclib = "^0.4.1"
 jinja2 = { version = "^2.11.2", optional = true }
-protobuf = { version = "^3.12.2", optional = true }
+python-dateutil = "^2.8"

 [tool.poetry.dev-dependencies]
+asv = "^0.4.2"
 black = "^20.8b1"
 bpython = "^0.19"
 grpcio-tools = "^1.30.0"
 jinja2 = "^2.11.2"
 mypy = "^0.770"
-poethepoet = "^0.5.0"
+poethepoet = ">=0.9.0"
 protobuf = "^3.12.2"
 pytest = "^5.4.2"
 pytest-asyncio = "^0.12.0"
 pytest-cov = "^2.9.0"
 pytest-mock = "^3.1.1"
-tox = "^3.15.1"
 sphinx = "3.1.2"
 sphinx-rtd-theme = "0.5.0"
-asv = "^0.4.2"
+tomlkit = "^0.7.0"
+tox = "^3.15.1"
+

 [tool.poetry.scripts]
 protoc-gen-python_betterproto = "betterproto.plugin:main"

 [tool.poetry.extras]
-compiler = ["black", "jinja2", "protobuf"]
+compiler = ["black", "jinja2"]
+

-[tool.poe.tasks]
 # Dev workflow tasks
-generate    = { script = "tests.generate:main", help = "Generate test cases (do this once before running test)" }
-test        = { cmd = "pytest --cov src", help = "Run tests" }
-types       = { cmd = "mypy src --ignore-missing-imports", help = "Check types with mypy" }
-format      = { cmd = "black . --exclude tests/output_",   help = "Apply black formatting to source code" }
-clean       = { cmd = "rm -rf .coverage .mypy_cache .pytest_cache dist betterproto.egg-info **/__pycache__ tests/output_*", help = "Clean out generated files from the workspace" }
-docs        = { cmd = "sphinx-build docs docs/build", help = "Build the sphinx docs"}
-bench       = { shell = "asv run master^! && asv run HEAD^! && asv compare master HEAD", help = "Benchmark current commit vs. master branch"}
+
+[tool.poe.tasks.generate]
+script = "tests.generate:main"
+help = "Generate test cases (do this once before running test)"
+
+[tool.poe.tasks.test]
+cmd = "pytest"
+help = "Run tests"
+
+[tool.poe.tasks.types]
+cmd  = "mypy src --ignore-missing-imports"
+help = "Check types with mypy"
+
+[tool.poe.tasks.format]
+cmd  = "black . --exclude tests/output_"
+help = "Apply black formatting to source code"
+
+[tool.poe.tasks.docs]
+cmd  = "sphinx-build docs docs/build"
+help = "Build the sphinx docs"
+
+[tool.poe.tasks.bench]
+shell = "asv run master^! && asv run HEAD^! && asv compare master HEAD"
+help  = "Benchmark current commit vs. master branch"
+
+[tool.poe.tasks.clean]
+cmd  = """
+rm -rf .asv .coverage .mypy_cache .pytest_cache
+       dist betterproto.egg-info **/__pycache__
+       testsoutput_*
+"""
+help = "Clean out generated files from the workspace"
+
+[tool.poe.tasks.generate_lib]
+cmd = """
+protoc
+    --plugin=protoc-gen-custom=src/betterproto/plugin/main.py
+    --custom_opt=INCLUDE_GOOGLE
+    --custom_out=src/betterproto/lib
+    -I /usr/local/include/
+    /usr/local/include/google/protobuf/**/*.proto
+"""
+help = "Regenerate the types in betterproto.lib.google"

 # CI tasks
-full-test   = { shell = "poe generate && tox", help = "Run tests with multiple pythons" }
-check-style = { cmd = "black . --check --diff --exclude tests/output_", help = "Check if code style is correct"}
+
+[tool.poe.tasks.full-test]
+shell = "poe generate && tox"
+help = "Run tests with multiple pythons"
+
+[tool.poe.tasks.check-style]
+cmd = "black . --check --diff --exclude tests/output_"
+help = "Check if code style is correct"
+

 [tool.black]
 target-version = ['py36']
--- a/src/betterproto/init.py
+++ b/src/betterproto/init.py
@@ -2,17 +2,20 @@ import dataclasses
 import enum
 import inspect
 import json
+import math
 import struct
 import sys
 import typing
 from abc import ABC
 from base64 import b64decode, b64encode
 from datetime import datetime, timedelta, timezone
+from dateutil.parser import isoparse
 from typing import (
    Any,
    Callable,
    Dict,
    Generator,
+    Iterable,
    List,
    Optional,
    Set,
@@ -23,15 +26,10 @@ from typing import (
 )

 from ._types import T
+from ._version import __version__
 from .casing import camel_case, safe_snake_case, snake_case
 from .grpc.grpclib_client import ServiceStub

-if sys.version_info[:2] < (3, 7):
-    # Apply backport of datetime.fromisoformat from 3.7
-    from backports.datetime_fromisoformat import MonkeyPatch
-
-    MonkeyPatch.patch_fromisoformat()
-

 # Proto 3 data types
 TYPE_ENUM = "enum"
@@ -117,6 +115,12 @@ def datetime_default_gen() -> datetime:
 DATETIME_ZERO = datetime_default_gen()


+# Special protobuf json doubles
+INFINITY = "Infinity"
+NEG_INFINITY = "-Infinity"
+NAN = "NaN"
+
+
 class Casing(enum.Enum):
    """Casing constants for serialization."""

@@ -269,7 +273,7 @@ class Enum(enum.IntEnum):
            The member was not found in the Enum.
        """
        try:
-            return cls._member_map_[name]
+            return cls._member_map_[name]  # type: ignore
        except KeyError as e:
            raise ValueError(f"Unknown value {name} for enum {cls.__name__}") from e

@@ -304,16 +308,16 @@ def encode_varint(value: int) -> bytes:

 def _preprocess_single(proto_type: str, wraps: str, value: Any) -> bytes:
    """Adjusts values before serialization."""
-    if proto_type in [
+    if proto_type in (
        TYPE_ENUM,
        TYPE_BOOL,
        TYPE_INT32,
        TYPE_INT64,
        TYPE_UINT32,
        TYPE_UINT64,
-    ]:
+    ):
        return encode_varint(value)
-    elif proto_type in [TYPE_SINT32, TYPE_SINT64]:
+    elif proto_type in (TYPE_SINT32, TYPE_SINT64):
        # Handle zig-zag encoding.
        return encode_varint(value << 1 if value >= 0 else (value << 1) ^ (~0))
    elif proto_type in FIXED_TYPES:
@@ -373,6 +377,51 @@ def _serialize_single(
    return bytes(output)


+def _parse_float(value: Any) -> float:
+    """Parse the given value to a float
+
+    Parameters
+    ----------
+    value : Any
+        Value to parse
+
+    Returns
+    -------
+    float
+        Parsed value
+    """
+    if value == INFINITY:
+        return float("inf")
+    if value == NEG_INFINITY:
+        return -float("inf")
+    if value == NAN:
+        return float("nan")
+    return float(value)
+
+
+def _dump_float(value: float) -> Union[float, str]:
+    """Dump the given float to JSON
+
+    Parameters
+    ----------
+    value : float
+        Value to dump
+
+    Returns
+    -------
+    Union[float, str]
+        Dumped valid, either a float or the strings
+        "Infinity" or "-Infinity"
+    """
+    if value == float("inf"):
+        return INFINITY
+    if value == -float("inf"):
+        return NEG_INFINITY
+    if value == float("nan"):
+        return NAN
+    return value
+
+
 def decode_varint(buffer: bytes, pos: int) -> Tuple[int, int]:
    """
    Decode a single varint value from a byte buffer. Returns the value and the
@@ -474,13 +523,13 @@ class ProtoClassMetadata:

    @staticmethod
    def _get_default_gen(
-        cls: Type["Message"], fields: List[dataclasses.Field]
+        cls: Type["Message"], fields: Iterable[dataclasses.Field]
    ) -> Dict[str, Callable[[], Any]]:
        return {field.name: cls._get_field_default_gen(field) for field in fields}

    @staticmethod
    def _get_cls_by_field(
-        cls: Type["Message"], fields: List[dataclasses.Field]
+        cls: Type["Message"], fields: Iterable[dataclasses.Field]
    ) -> Dict[str, Type]:
        field_cls = {}

@@ -568,7 +617,18 @@ class Message(ABC):
                other_val = other._get_field_default(field_name)

            if self_val != other_val:
-                return False
+                # We consider two nan values to be the same for the
+                # purposes of comparing messages (otherwise a message
+                # is not equal to itself)
+                if (
+                    isinstance(self_val, float)
+                    and isinstance(other_val, float)
+                    and math.isnan(self_val)
+                    and math.isnan(other_val)
+                ):
+                    continue
+                else:
+                    return False

        return True

@@ -628,7 +688,7 @@ class Message(ABC):
        meta = getattr(self.__class__, "_betterproto_meta", None)
        if not meta:
            meta = ProtoClassMetadata(self.__class__)
-            self.__class__._betterproto_meta = meta
+            self.__class__._betterproto_meta = meta  # type: ignore
        return meta

    def __bytes__(self) -> bytes:
@@ -679,9 +739,18 @@ class Message(ABC):
                    output += _serialize_single(meta.number, TYPE_BYTES, buf)
                else:
                    for item in value:
-                        output += _serialize_single(
-                            meta.number, meta.proto_type, item, wraps=meta.wraps or ""
+                        output += (
+                            _serialize_single(
+                                meta.number,
+                                meta.proto_type,
+                                item,
+                                wraps=meta.wraps or "",
+                            )
+                            # if it's an empty message it still needs to be represented
+                            # as an item in the repeated list
+                            or b"\n\x00"
                        )
+
            elif isinstance(value, dict):
                for k, v in value.items():
                    assert meta.map_types
@@ -704,7 +773,7 @@ class Message(ABC):
                    meta.number,
                    meta.proto_type,
                    value,
-                    serialize_empty=serialize_empty,
+                    serialize_empty=serialize_empty or bool(selected_in_group),
                    wraps=meta.wraps or "",
                )

@@ -781,23 +850,23 @@ class Message(ABC):
    ) -> Any:
        """Adjusts values after parsing."""
        if wire_type == WIRE_VARINT:
-            if meta.proto_type in [TYPE_INT32, TYPE_INT64]:
+            if meta.proto_type in (TYPE_INT32, TYPE_INT64):
                bits = int(meta.proto_type[3:])
                value = value & ((1 << bits) - 1)
                signbit = 1 << (bits - 1)
                value = int((value ^ signbit) - signbit)
-            elif meta.proto_type in [TYPE_SINT32, TYPE_SINT64]:
+            elif meta.proto_type in (TYPE_SINT32, TYPE_SINT64):
                # Undo zig-zag encoding
                value = (value >> 1) ^ (-(value & 1))
            elif meta.proto_type == TYPE_BOOL:
                # Booleans use a varint encoding, so convert it to true/false.
                value = value > 0
-        elif wire_type in [WIRE_FIXED_32, WIRE_FIXED_64]:
+        elif wire_type in (WIRE_FIXED_32, WIRE_FIXED_64):
            fmt = _pack_fmt(meta.proto_type)
            value = struct.unpack(fmt, value)[0]
        elif wire_type == WIRE_LEN_DELIM:
            if meta.proto_type == TYPE_STRING:
-                value = value.decode("utf-8")
+                value = str(value, "utf-8")
            elif meta.proto_type == TYPE_MESSAGE:
                cls = self._betterproto.cls_by_field[field_name]

@@ -856,10 +925,10 @@ class Message(ABC):
                pos = 0
                value = []
                while pos < len(parsed.value):
-                    if meta.proto_type in [TYPE_FLOAT, TYPE_FIXED32, TYPE_SFIXED32]:
+                    if meta.proto_type in (TYPE_FLOAT, TYPE_FIXED32, TYPE_SFIXED32):
                        decoded, pos = parsed.value[pos : pos + 4], pos + 4
                        wire_type = WIRE_FIXED_32
-                    elif meta.proto_type in [TYPE_DOUBLE, TYPE_FIXED64, TYPE_SFIXED64]:
+                    elif meta.proto_type in (TYPE_DOUBLE, TYPE_FIXED64, TYPE_SFIXED64):
                        decoded, pos = parsed.value[pos : pos + 8], pos + 8
                        wire_type = WIRE_FIXED_64
                    else:
@@ -961,7 +1030,15 @@ class Message(ABC):
                        output[cased_name] = value
                elif field_is_repeated:
                    # Convert each item.
-                    value = [i.to_dict(casing, include_default_values) for i in value]
+                    cls = self._betterproto.cls_by_field[field_name]
+                    if cls == datetime:
+                        value = [_Timestamp.timestamp_to_json(i) for i in value]
+                    elif cls == timedelta:
+                        value = [_Duration.delta_to_json(i) for i in value]
+                    else:
+                        value = [
+                            i.to_dict(casing, include_default_values) for i in value
+                        ]
                    if value or include_default_values:
                        output[cased_name] = value
                elif (
@@ -1000,7 +1077,7 @@ class Message(ABC):
                        output[cased_name] = b64encode(value).decode("utf8")
                elif meta.proto_type == TYPE_ENUM:
                    if field_is_repeated:
-                        enum_class: Type[Enum] = field_types[field_name].__args__[0]
+                        enum_class = field_types[field_name].__args__[0]
                        if isinstance(value, typing.Iterable) and not isinstance(
                            value, str
                        ):
@@ -1009,8 +1086,13 @@ class Message(ABC):
                            # transparently upgrade single value to repeated
                            output[cased_name] = [enum_class(value).name]
                    else:
-                        enum_class: Type[Enum] = field_types[field_name]  # noqa
+                        enum_class = field_types[field_name]  # noqa
                        output[cased_name] = enum_class(value).name
+                elif meta.proto_type in (TYPE_FLOAT, TYPE_DOUBLE):
+                    if field_is_repeated:
+                        output[cased_name] = [_dump_float(n) for n in value]
+                    else:
+                        output[cased_name] = _dump_float(value)
                else:
                    output[cased_name] = value
        return output
@@ -1042,10 +1124,17 @@ class Message(ABC):
                    v = getattr(self, field_name)
                    if isinstance(v, list):
                        cls = self._betterproto.cls_by_field[field_name]
-                        for item in value[key]:
-                            v.append(cls().from_dict(item))
+                        if cls == datetime:
+                            v = [isoparse(item) for item in value[key]]
+                        elif cls == timedelta:
+                            v = [
+                                timedelta(seconds=float(item[:-1]))
+                                for item in value[key]
+                            ]
+                        else:
+                            v = [cls().from_dict(item) for item in value[key]]
                    elif isinstance(v, datetime):
-                        v = datetime.fromisoformat(value[key].replace("Z", "+00:00"))
+                        v = isoparse(value[key])
                        setattr(self, field_name, v)
                    elif isinstance(v, timedelta):
                        v = timedelta(seconds=float(value[key][:-1]))
@@ -1079,6 +1168,11 @@ class Message(ABC):
                            v = [enum_cls.from_string(e) for e in v]
                        elif isinstance(v, str):
                            v = enum_cls.from_string(v)
+                    elif meta.proto_type in (TYPE_FLOAT, TYPE_DOUBLE):
+                        if isinstance(value[key], list):
+                            v = [_parse_float(n) for n in value[key]]
+                        else:
+                            v = _parse_float(value[key])

                if v is not None:
                    setattr(self, field_name, v)
@@ -1161,6 +1255,7 @@ from .lib.google.protobuf import (  # noqa
    BytesValue,
    DoubleValue,
    Duration,
+    EnumValue,
    FloatValue,
    Int32Value,
    Int64Value,
@@ -1179,7 +1274,7 @@ class _Duration(Duration):
    def delta_to_json(delta: timedelta) -> str:
        parts = str(delta.total_seconds()).split(".")
        if len(parts) > 1:
-            while len(parts[1]) not in [3, 6, 9]:
+            while len(parts[1]) not in (3, 6, 9):
                parts[1] = f"{parts[1]}0"
        return f"{'.'.join(parts)}s"

@@ -1208,33 +1303,19 @@ class _Timestamp(Timestamp):
        return f"{result}.{nanos:09d}"


-class _WrappedMessage(Message):
-    """
-    Google protobuf wrapper types base class. JSON representation is just the
-    value itself.
-    """
-
-    value: Any
-
-    def to_dict(self, casing: Casing = Casing.CAMEL) -> Any:
-        return self.value
-
-    def from_dict(self: T, value: Any) -> T:
-        if value is not None:
-            self.value = value
-        return self
-
-
 def _get_wrapper(proto_type: str) -> Type:
    """Get the wrapper message class for a wrapped type."""
+
+    # TODO: include ListValue and NullValue?
    return {
        TYPE_BOOL: BoolValue,
-        TYPE_INT32: Int32Value,
-        TYPE_UINT32: UInt32Value,
-        TYPE_INT64: Int64Value,
-        TYPE_UINT64: UInt64Value,
-        TYPE_FLOAT: FloatValue,
-        TYPE_DOUBLE: DoubleValue,
-        TYPE_STRING: StringValue,
        TYPE_BYTES: BytesValue,
+        TYPE_DOUBLE: DoubleValue,
+        TYPE_FLOAT: FloatValue,
+        TYPE_ENUM: EnumValue,
+        TYPE_INT32: Int32Value,
+        TYPE_INT64: Int64Value,
+        TYPE_STRING: StringValue,
+        TYPE_UINT32: UInt32Value,
+        TYPE_UINT64: UInt64Value,
    }[proto_type]
--- a/src/betterproto/_version.py
+++ b/src/betterproto/_version.py
@@ -0,0 +1,3 @@
+from pkg_resources import get_distribution
+
+__version__ = get_distribution("betterproto").version
--- a/src/betterproto/casing.py
+++ b/src/betterproto/casing.py
@@ -133,6 +133,16 @@ def lowercase_first(value: str) -> str:
    return value[0:1].lower() + value[1:]


+def is_reserved_name(value: str) -> bool:
+    if keyword.iskeyword(value):
+        return True
+
+    if value in ("bytes", "str"):
+        return True
+
+    return False
+
+
 def sanitize_name(value: str) -> str:
    # https://www.python.org/dev/peps/pep-0008/#descriptive-naming-styles
-    return f"{value}_" if keyword.iskeyword(value) else value
+    return f"{value}_" if is_reserved_name(value) else value
--- a/src/betterproto/compile/importing.py
+++ b/src/betterproto/compile/importing.py
@@ -100,8 +100,9 @@ def reference_descendent(
    current_package: List[str], imports: Set[str], py_package: List[str], py_type: str
 ) -> str:
    """
-    Returns a reference to a python type in a package that is a descendent of the current package,
-    and adds the required import that is aliased to avoid name conflicts.
+    Returns a reference to a python type in a package that is a descendent of the
+    current package, and adds the required import that is aliased to avoid name
+    conflicts.
    """
    importing_descendent = py_package[len(current_package) :]
    string_from = ".".join(importing_descendent[:-1])
@@ -119,8 +120,9 @@ def reference_ancestor(
    current_package: List[str], imports: Set[str], py_package: List[str], py_type: str
 ) -> str:
    """
-    Returns a reference to a python type in a package which is an ancestor to the current package,
-    and adds the required import that is aliased (if possible) to avoid name conflicts.
+    Returns a reference to a python type in a package which is an ancestor to the
+    current package, and adds the required import that is aliased (if possible) to avoid
+    name conflicts.

    Adds trailing __ to avoid name mangling (python.org/dev/peps/pep-0008/#id34).
    """
@@ -141,10 +143,10 @@ def reference_cousin(
    current_package: List[str], imports: Set[str], py_package: List[str], py_type: str
 ) -> str:
    """
-    Returns a reference to a python type in a package that is not descendent, ancestor or sibling,
-    and adds the required import that is aliased to avoid name conflicts.
+    Returns a reference to a python type in a package that is not descendent, ancestor
+    or sibling, and adds the required import that is aliased to avoid name conflicts.
    """
-    shared_ancestry = os.path.commonprefix([current_package, py_package])
+    shared_ancestry = os.path.commonprefix([current_package, py_package])  # type: ignore
    distance_up = len(current_package) - len(shared_ancestry)
    string_from = f".{'.' * distance_up}" + ".".join(
        py_package[len(shared_ancestry) : -1]
--- a/src/betterproto/grpc/grpclib_server.py
+++ b/src/betterproto/grpc/grpclib_server.py
@@ -0,0 +1,30 @@
+from abc import ABC
+from collections.abc import AsyncIterable
+from typing import Callable, Any, Dict
+
+import grpclib
+import grpclib.server
+
+
+class ServiceBase(ABC):
+    """
+    Base class for async gRPC servers.
+    """
+
+    async def _call_rpc_handler_server_stream(
+        self,
+        handler: Callable,
+        stream: grpclib.server.Stream,
+        request_kwargs: Dict[str, Any],
+    ) -> None:
+
+        response_iter = handler(**request_kwargs)
+        # check if response is actually an AsyncIterator
+        # this might be false if the method just returns without
+        # yielding at least once
+        # in that case, we just interpret it as an empty iterator
+        if isinstance(response_iter, AsyncIterable):
+            async for response_message in response_iter:
+                await stream.send_message(response_message)
+        else:
+            response_iter.close()
--- a/src/betterproto/grpc/util/async_channel.py
+++ b/src/betterproto/grpc/util/async_channel.py
@@ -70,7 +70,7 @@ class AsyncChannel(AsyncIterable[T]):
    """

    def __init__(self, *, buffer_limit: int = 0, close: bool = False):
-        self._queue: asyncio.Queue[Union[T, object]] = asyncio.Queue(buffer_limit)
+        self._queue: asyncio.Queue[T] = asyncio.Queue(buffer_limit)
        self._closed = False
        self._waiting_receivers: int = 0
        # Track whether flush has been invoked so it can only happen once
--- a/src/betterproto/lib/google/protobuf/init.py
+++ b/src/betterproto/lib/google/protobuf/init.py
--- a/src/betterproto/lib/google/protobuf/compiler/init.py
+++ b/src/betterproto/lib/google/protobuf/compiler/init.py
@@ -0,0 +1,113 @@
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# sources: google/protobuf/compiler/plugin.proto
+# plugin: python-betterproto
+from dataclasses import dataclass
+from typing import List
+
+import betterproto
+
+
+@dataclass(eq=False, repr=False)
+class Version(betterproto.Message):
+    """The version number of protocol compiler."""
+
+    major: int = betterproto.int32_field(1)
+    minor: int = betterproto.int32_field(2)
+    patch: int = betterproto.int32_field(3)
+    # A suffix for alpha, beta or rc release, e.g., "alpha-1", "rc2". It should
+    # be empty for mainline stable releases.
+    suffix: str = betterproto.string_field(4)
+
+
+@dataclass(eq=False, repr=False)
+class CodeGeneratorRequest(betterproto.Message):
+    """An encoded CodeGeneratorRequest is written to the plugin's stdin."""
+
+    # The .proto files that were explicitly listed on the command-line.  The code
+    # generator should generate code only for these files.  Each file's
+    # descriptor will be included in proto_file, below.
+    file_to_generate: List[str] = betterproto.string_field(1)
+    # The generator parameter passed on the command-line.
+    parameter: str = betterproto.string_field(2)
+    # FileDescriptorProtos for all files in files_to_generate and everything they
+    # import.  The files will appear in topological order, so each file appears
+    # before any file that imports it. protoc guarantees that all proto_files
+    # will be written after the fields above, even though this is not technically
+    # guaranteed by the protobuf wire format.  This theoretically could allow a
+    # plugin to stream in the FileDescriptorProtos and handle them one by one
+    # rather than read the entire set into memory at once.  However, as of this
+    # writing, this is not similarly optimized on protoc's end -- it will store
+    # all fields in memory at once before sending them to the plugin. Type names
+    # of fields and extensions in the FileDescriptorProto are always fully
+    # qualified.
+    proto_file: List[
+        "betterproto_lib_google_protobuf.FileDescriptorProto"
+    ] = betterproto.message_field(15)
+    # The version number of protocol compiler.
+    compiler_version: "Version" = betterproto.message_field(3)
+
+
+@dataclass(eq=False, repr=False)
+class CodeGeneratorResponse(betterproto.Message):
+    """The plugin writes an encoded CodeGeneratorResponse to stdout."""
+
+    # Error message.  If non-empty, code generation failed.  The plugin process
+    # should exit with status code zero even if it reports an error in this way.
+    # This should be used to indicate errors in .proto files which prevent the
+    # code generator from generating correct code.  Errors which indicate a
+    # problem in protoc itself -- such as the input CodeGeneratorRequest being
+    # unparseable -- should be reported by writing a message to stderr and
+    # exiting with a non-zero status code.
+    error: str = betterproto.string_field(1)
+    file: List["CodeGeneratorResponseFile"] = betterproto.message_field(15)
+
+
+@dataclass(eq=False, repr=False)
+class CodeGeneratorResponseFile(betterproto.Message):
+    """Represents a single generated file."""
+
+    # The file name, relative to the output directory.  The name must not contain
+    # "." or ".." components and must be relative, not be absolute (so, the file
+    # cannot lie outside the output directory).  "/" must be used as the path
+    # separator, not "\". If the name is omitted, the content will be appended to
+    # the previous file.  This allows the generator to break large files into
+    # small chunks, and allows the generated text to be streamed back to protoc
+    # so that large files need not reside completely in memory at one time.  Note
+    # that as of this writing protoc does not optimize for this -- it will read
+    # the entire CodeGeneratorResponse before writing files to disk.
+    name: str = betterproto.string_field(1)
+    # If non-empty, indicates that the named file should already exist, and the
+    # content here is to be inserted into that file at a defined insertion point.
+    # This feature allows a code generator to extend the output produced by
+    # another code generator.  The original generator may provide insertion
+    # points by placing special annotations in the file that look like:
+    # @@protoc_insertion_point(NAME) The annotation can have arbitrary text
+    # before and after it on the line, which allows it to be placed in a comment.
+    # NAME should be replaced with an identifier naming the point -- this is what
+    # other generators will use as the insertion_point.  Code inserted at this
+    # point will be placed immediately above the line containing the insertion
+    # point (thus multiple insertions to the same point will come out in the
+    # order they were added). The double-@ is intended to make it unlikely that
+    # the generated code could contain things that look like insertion points by
+    # accident. For example, the C++ code generator places the following line in
+    # the .pb.h files that it generates:   //
+    # @@protoc_insertion_point(namespace_scope) This line appears within the
+    # scope of the file's package namespace, but outside of any particular class.
+    # Another plugin can then specify the insertion_point "namespace_scope" to
+    # generate additional classes or other declarations that should be placed in
+    # this scope. Note that if the line containing the insertion point begins
+    # with whitespace, the same whitespace will be added to every line of the
+    # inserted text.  This is useful for languages like Python, where indentation
+    # matters.  In these languages, the insertion point comment should be
+    # indented the same amount as any inserted code will need to be in order to
+    # work correctly in that context. The code generator that generates the
+    # initial file and the one which inserts into it must both run as part of a
+    # single invocation of protoc. Code generators are executed in the order in
+    # which they appear on the command line. If |insertion_point| is present,
+    # |name| must also be present.
+    insertion_point: str = betterproto.string_field(2)
+    # The file contents.
+    content: str = betterproto.string_field(15)
+
+
+import betterproto.lib.google.protobuf as betterproto_lib_google_protobuf
--- a/src/betterproto/plugin/main.py
+++ b/src/betterproto/plugin/main.py
@@ -3,9 +3,13 @@
 import os
 import sys

-from google.protobuf.compiler import plugin_pb2 as plugin
+from betterproto.lib.google.protobuf.compiler import (
+    CodeGeneratorRequest,
+    CodeGeneratorResponse,
+)

 from betterproto.plugin.parser import generate_code
+from betterproto.plugin.models import monkey_patch_oneof_index


 def main() -> None:
@@ -13,16 +17,19 @@ def main() -> None:
    # Read request message from stdin
    data = sys.stdin.buffer.read()

+    # Apply Work around for proto2/3 difference in protoc messages
+    monkey_patch_oneof_index()
+
    # Parse request
-    request = plugin.CodeGeneratorRequest()
-    request.ParseFromString(data)
+    request = CodeGeneratorRequest()
+    request.parse(data)

    dump_file = os.getenv("BETTERPROTO_DUMP")
    if dump_file:
        dump_request(dump_file, request)

    # Create response
-    response = plugin.CodeGeneratorResponse()
+    response = CodeGeneratorResponse()

    # Generate code
    generate_code(request, response)
@@ -34,7 +41,7 @@ def main() -> None:
    sys.stdout.buffer.write(output)


-def dump_request(dump_file: str, request: plugin.CodeGeneratorRequest) -> None:
+def dump_request(dump_file: str, request: CodeGeneratorRequest) -> None:
    """
    For developers: Supports running plugin.py standalone so its possible to debug it.
    Run protoc (or generate.py) with BETTERPROTO_DUMP="yourfile.bin" to write the request to a file.
--- a/src/betterproto/plugin/models.py
+++ b/src/betterproto/plugin/models.py
@@ -29,12 +29,37 @@ instantiating field `A` with parent message `B` should add a
 reference to `A` to `B`'s `fields` attribute.
 """

+
+import betterproto
+from betterproto import which_one_of
+from betterproto.casing import sanitize_name
+from betterproto.compile.importing import (
+    get_type_reference,
+    parse_source_type_name,
+)
+from betterproto.compile.naming import (
+    pythonize_class_name,
+    pythonize_field_name,
+    pythonize_method_name,
+)
+from betterproto.lib.google.protobuf import (
+    DescriptorProto,
+    EnumDescriptorProto,
+    FileDescriptorProto,
+    MethodDescriptorProto,
+    Field,
+    FieldDescriptorProto,
+    FieldDescriptorProtoType,
+    FieldDescriptorProtoLabel,
+)
+from betterproto.lib.google.protobuf.compiler import CodeGeneratorRequest
+
+
 import re
 import textwrap
 from dataclasses import dataclass, field
-from typing import Dict, Iterator, List, Optional, Set, Text, Type, Union
-
-import betterproto
+from typing import Dict, Iterable, Iterator, List, Optional, Set, Text, Type, Union
+import sys

 from ..casing import sanitize_name
 from ..compile.importing import get_type_reference, parse_source_type_name
@@ -44,26 +69,6 @@ from ..compile.naming import (
    pythonize_method_name,
 )

-try:
-    # betterproto[compiler] specific dependencies
-    from google.protobuf.compiler import plugin_pb2 as plugin
-    from google.protobuf.descriptor_pb2 import (
-        DescriptorProto,
-        EnumDescriptorProto,
-        FieldDescriptorProto,
-        FileDescriptorProto,
-        MethodDescriptorProto,
-    )
-except ImportError as err:
-    print(
-        "\033[31m"
-        f"Unable to import `{err.name}` from betterproto plugin! "
-        "Please ensure that you've installed betterproto as "
-        '`pip install "betterproto[compiler]"` so that compiler dependencies '
-        "are included."
-        "\033[0m"
-    )
-    raise SystemExit(1)

 # Create a unique placeholder to deal with
 # https://stackoverflow.com/questions/51575931/class-inheritance-in-python-3-7-dataclasses
@@ -71,54 +76,75 @@ PLACEHOLDER = object()

 # Organize proto types into categories
 PROTO_FLOAT_TYPES = (
-    FieldDescriptorProto.TYPE_DOUBLE,  # 1
-    FieldDescriptorProto.TYPE_FLOAT,  # 2
+    FieldDescriptorProtoType.TYPE_DOUBLE,  # 1
+    FieldDescriptorProtoType.TYPE_FLOAT,  # 2
 )
 PROTO_INT_TYPES = (
-    FieldDescriptorProto.TYPE_INT64,  # 3
-    FieldDescriptorProto.TYPE_UINT64,  # 4
-    FieldDescriptorProto.TYPE_INT32,  # 5
-    FieldDescriptorProto.TYPE_FIXED64,  # 6
-    FieldDescriptorProto.TYPE_FIXED32,  # 7
-    FieldDescriptorProto.TYPE_UINT32,  # 13
-    FieldDescriptorProto.TYPE_SFIXED32,  # 15
-    FieldDescriptorProto.TYPE_SFIXED64,  # 16
-    FieldDescriptorProto.TYPE_SINT32,  # 17
-    FieldDescriptorProto.TYPE_SINT64,  # 18
+    FieldDescriptorProtoType.TYPE_INT64,  # 3
+    FieldDescriptorProtoType.TYPE_UINT64,  # 4
+    FieldDescriptorProtoType.TYPE_INT32,  # 5
+    FieldDescriptorProtoType.TYPE_FIXED64,  # 6
+    FieldDescriptorProtoType.TYPE_FIXED32,  # 7
+    FieldDescriptorProtoType.TYPE_UINT32,  # 13
+    FieldDescriptorProtoType.TYPE_SFIXED32,  # 15
+    FieldDescriptorProtoType.TYPE_SFIXED64,  # 16
+    FieldDescriptorProtoType.TYPE_SINT32,  # 17
+    FieldDescriptorProtoType.TYPE_SINT64,  # 18
 )
-PROTO_BOOL_TYPES = (FieldDescriptorProto.TYPE_BOOL,)  # 8
-PROTO_STR_TYPES = (FieldDescriptorProto.TYPE_STRING,)  # 9
-PROTO_BYTES_TYPES = (FieldDescriptorProto.TYPE_BYTES,)  # 12
+PROTO_BOOL_TYPES = (FieldDescriptorProtoType.TYPE_BOOL,)  # 8
+PROTO_STR_TYPES = (FieldDescriptorProtoType.TYPE_STRING,)  # 9
+PROTO_BYTES_TYPES = (FieldDescriptorProtoType.TYPE_BYTES,)  # 12
 PROTO_MESSAGE_TYPES = (
-    FieldDescriptorProto.TYPE_MESSAGE,  # 11
-    FieldDescriptorProto.TYPE_ENUM,  # 14
+    FieldDescriptorProtoType.TYPE_MESSAGE,  # 11
+    FieldDescriptorProtoType.TYPE_ENUM,  # 14
 )
-PROTO_MAP_TYPES = (FieldDescriptorProto.TYPE_MESSAGE,)  # 11
+PROTO_MAP_TYPES = (FieldDescriptorProtoType.TYPE_MESSAGE,)  # 11
 PROTO_PACKED_TYPES = (
-    FieldDescriptorProto.TYPE_DOUBLE,  # 1
-    FieldDescriptorProto.TYPE_FLOAT,  # 2
-    FieldDescriptorProto.TYPE_INT64,  # 3
-    FieldDescriptorProto.TYPE_UINT64,  # 4
-    FieldDescriptorProto.TYPE_INT32,  # 5
-    FieldDescriptorProto.TYPE_FIXED64,  # 6
-    FieldDescriptorProto.TYPE_FIXED32,  # 7
-    FieldDescriptorProto.TYPE_BOOL,  # 8
-    FieldDescriptorProto.TYPE_UINT32,  # 13
-    FieldDescriptorProto.TYPE_SFIXED32,  # 15
-    FieldDescriptorProto.TYPE_SFIXED64,  # 16
-    FieldDescriptorProto.TYPE_SINT32,  # 17
-    FieldDescriptorProto.TYPE_SINT64,  # 18
+    FieldDescriptorProtoType.TYPE_DOUBLE,  # 1
+    FieldDescriptorProtoType.TYPE_FLOAT,  # 2
+    FieldDescriptorProtoType.TYPE_INT64,  # 3
+    FieldDescriptorProtoType.TYPE_UINT64,  # 4
+    FieldDescriptorProtoType.TYPE_INT32,  # 5
+    FieldDescriptorProtoType.TYPE_FIXED64,  # 6
+    FieldDescriptorProtoType.TYPE_FIXED32,  # 7
+    FieldDescriptorProtoType.TYPE_BOOL,  # 8
+    FieldDescriptorProtoType.TYPE_UINT32,  # 13
+    FieldDescriptorProtoType.TYPE_SFIXED32,  # 15
+    FieldDescriptorProtoType.TYPE_SFIXED64,  # 16
+    FieldDescriptorProtoType.TYPE_SINT32,  # 17
+    FieldDescriptorProtoType.TYPE_SINT64,  # 18
 )


+def monkey_patch_oneof_index():
+    """
+    The compiler message types are written for proto2, but we read them as proto3.
+    For this to work in the case of the oneof_index fields, which depend on being able
+    to tell whether they were set, we have to treat them as oneof fields. This method
+    monkey patches the generated classes after the fact to force this behaviour.
+    """
+    object.__setattr__(
+        FieldDescriptorProto.__dataclass_fields__["oneof_index"].metadata[
+            "betterproto"
+        ],
+        "group",
+        "oneof_index",
+    )
+    object.__setattr__(
+        Field.__dataclass_fields__["oneof_index"].metadata["betterproto"],
+        "group",
+        "oneof_index",
+    )
+
+
 def get_comment(
    proto_file: "FileDescriptorProto", path: List[int], indent: int = 4
 ) -> str:
    pad = " " * indent
-    for sci in proto_file.source_code_info.location:
-        if list(sci.path) == path and sci.leading_comments:
+    for sci_loc in proto_file.source_code_info.location:
+        if list(sci_loc.path) == path and sci_loc.leading_comments:
            lines = textwrap.wrap(
-                sci.leading_comments.strip().replace("\n", ""), width=79 - indent
+                sci_loc.leading_comments.strip().replace("\n", ""), width=79 - indent
            )

            if path[-2] == 2 and path[-4] != 6:
@@ -139,10 +165,13 @@ def get_comment(
 class ProtoContentBase:
    """Methods common to MessageCompiler, ServiceCompiler and ServiceMethodCompiler."""

+    source_file: FileDescriptorProto
    path: List[int]
    comment_indent: int = 4
    parent: Union["betterproto.Message", "OutputTemplate"]

+    __dataclass_fields__: Dict[str, object]
+
    def __post_init__(self) -> None:
        """Checks that no fake default fields were left as placeholders."""
        for field_name, field_val in self.__dataclass_fields__.items():
@@ -156,13 +185,6 @@ class ProtoContentBase:
            current = current.parent
        return current

-    @property
-    def proto_file(self) -> FieldDescriptorProto:
-        current = self
-        while not isinstance(current, OutputTemplate):
-            current = current.parent
-        return current.package_proto_obj
-
    @property
    def request(self) -> "PluginRequestCompiler":
        current = self
@@ -176,14 +198,14 @@ class ProtoContentBase:
        for this object.
        """
        return get_comment(
-            proto_file=self.proto_file, path=self.path, indent=self.comment_indent
+            proto_file=self.source_file, path=self.path, indent=self.comment_indent
        )


@dataclass
 class PluginRequestCompiler:

-    plugin_request_obj: plugin.CodeGeneratorRequest
+    plugin_request_obj: CodeGeneratorRequest
    output_packages: Dict[str, "OutputTemplate"] = field(default_factory=dict)

    @property
@@ -231,15 +253,15 @@ class OutputTemplate:
        return self.package_proto_obj.package

    @property
-    def input_filenames(self) -> List[str]:
+    def input_filenames(self) -> Iterable[str]:
        """Names of the input files used to build this output.

        Returns
        -------
-        List[str]
+        Iterable[str]
            Names of the input files used to build this output.
        """
-        return [f.name for f in self.input_files]
+        return sorted(f.name for f in self.input_files)

    @property
    def python_module_imports(self) -> Set[str]:
@@ -253,6 +275,7 @@ class OutputTemplate:
 class MessageCompiler(ProtoContentBase):
    """Representation of a protobuf message."""

+    source_file: FileDescriptorProto
    parent: Union["MessageCompiler", OutputTemplate] = PLACEHOLDER
    proto_obj: DescriptorProto = PLACEHOLDER
    path: List[int] = PLACEHOLDER
@@ -291,12 +314,16 @@ class MessageCompiler(ProtoContentBase):
            if f.deprecated:
                yield f.py_name

+    @property
+    def has_deprecated_fields(self) -> bool:
+        return any(self.deprecated_fields)
+

 def is_map(
    proto_field_obj: FieldDescriptorProto, parent_message: DescriptorProto
 ) -> bool:
    """True if proto_field_obj is a map, otherwise False."""
-    if proto_field_obj.type == FieldDescriptorProto.TYPE_MESSAGE:
+    if proto_field_obj.type == FieldDescriptorProtoType.TYPE_MESSAGE:
        # This might be a map...
        message_type = proto_field_obj.type_name.split(".").pop().lower()
        map_entry = f"{proto_field_obj.name.replace('_', '').lower()}entry"
@@ -311,8 +338,20 @@ def is_map(


 def is_oneof(proto_field_obj: FieldDescriptorProto) -> bool:
-    """True if proto_field_obj is a OneOf, otherwise False."""
-    return proto_field_obj.HasField("oneof_index")
+    """
+    True if proto_field_obj is a OneOf, otherwise False.
+
+    .. warning::
+        Becuase the message from protoc is defined in proto2, and betterproto works with
+        proto3, and interpreting the FieldDescriptorProto.oneof_index field requires
+        distinguishing between default and unset values (which proto3 doesn't support),
+        we have to hack the generated FieldDescriptorProto class for this to work.
+        The hack consists of setting group="oneof_index" in the field metadata,
+        essentially making oneof_index the sole member of a one_of group, which allows
+        us to tell whether it was set, via the which_one_of interface.
+    """
+
+    return which_one_of(proto_field_obj, "oneof_index")[0] == "oneof_index"


@dataclass
@@ -324,17 +363,7 @@ class FieldCompiler(MessageCompiler):
        # Add field to message
        self.parent.fields.append(self)
        # Check for new imports
-        annotation = self.annotation
-        if "Optional[" in annotation:
-            self.output_file.typing_imports.add("Optional")
-        if "List[" in annotation:
-            self.output_file.typing_imports.add("List")
-        if "Dict[" in annotation:
-            self.output_file.typing_imports.add("Dict")
-        if "timedelta" in annotation:
-            self.output_file.datetime_imports.add("timedelta")
-        if "datetime" in annotation:
-            self.output_file.datetime_imports.add("datetime")
+        self.add_imports_to(self.output_file)
        super().__post_init__()  # call FieldCompiler-> MessageCompiler __post_init__

    def get_field_string(self, indent: int = 4) -> str:
@@ -356,11 +385,38 @@ class FieldCompiler(MessageCompiler):
            args.append(f"wraps={self.field_wraps}")
        return args

+    @property
+    def datetime_imports(self) -> Set[str]:
+        imports = set()
+        annotation = self.annotation
+        # FIXME: false positives - e.g. `MyDatetimedelta`
+        if "timedelta" in annotation:
+            imports.add("timedelta")
+        if "datetime" in annotation:
+            imports.add("datetime")
+        return imports
+
+    @property
+    def typing_imports(self) -> Set[str]:
+        imports = set()
+        annotation = self.annotation
+        if "Optional[" in annotation:
+            imports.add("Optional")
+        if "List[" in annotation:
+            imports.add("List")
+        if "Dict[" in annotation:
+            imports.add("Dict")
+        return imports
+
+    def add_imports_to(self, output_file: OutputTemplate) -> None:
+        output_file.datetime_imports.update(self.datetime_imports)
+        output_file.typing_imports.update(self.typing_imports)
+
    @property
    def field_wraps(self) -> Optional[str]:
        """Returns betterproto wrapped field type or None."""
        match_wrapper = re.match(
-            r"\.google\.protobuf\.(.+)Value", self.proto_obj.type_name
+            r"\.google\.protobuf\.(.+)Value$", self.proto_obj.type_name
        )
        if match_wrapper:
            wrapped_type = "TYPE_" + match_wrapper.group(1).upper()
@@ -371,7 +427,7 @@ class FieldCompiler(MessageCompiler):
    @property
    def repeated(self) -> bool:
        return (
-            self.proto_obj.label == FieldDescriptorProto.LABEL_REPEATED
+            self.proto_obj.label == FieldDescriptorProtoLabel.LABEL_REPEATED
            and not is_map(self.proto_obj, self.parent)
        )

@@ -384,7 +440,9 @@ class FieldCompiler(MessageCompiler):
    def field_type(self) -> str:
        """String representation of proto field type."""
        return (
-            self.proto_obj.Type.Name(self.proto_obj.type).lower().replace("type_", "")
+            FieldDescriptorProtoType(self.proto_obj.type)
+            .name.lower()
+            .replace("type_", "")
        )

    @property
@@ -478,14 +536,19 @@ class MapEntryCompiler(FieldCompiler):
            ):
                # Get Python types
                self.py_k_type = FieldCompiler(
-                    parent=self, proto_obj=nested.field[0]  # key
+                    source_file=self.source_file,
+                    parent=self,
+                    proto_obj=nested.field[0],  # key
                ).py_type
                self.py_v_type = FieldCompiler(
-                    parent=self, proto_obj=nested.field[1]  # value
+                    source_file=self.source_file,
+                    parent=self,
+                    proto_obj=nested.field[1],  # value
                ).py_type
+
                # Get proto types
-                self.proto_k_type = self.proto_obj.Type.Name(nested.field[0].type)
-                self.proto_v_type = self.proto_obj.Type.Name(nested.field[1].type)
+                self.proto_k_type = FieldDescriptorProtoType(nested.field[0].type).name
+                self.proto_v_type = FieldDescriptorProtoType(nested.field[1].type).name
        super().__post_init__()  # call FieldCompiler-> MessageCompiler __post_init__

    @property
@@ -527,7 +590,7 @@ class EnumDefinitionCompiler(MessageCompiler):
                name=sanitize_name(entry_proto_value.name),
                value=entry_proto_value.number,
                comment=get_comment(
-                    proto_file=self.proto_file, path=self.path + [2, entry_number]
+                    proto_file=self.source_file, path=self.path + [2, entry_number]
                ),
            )
            for entry_number, entry_proto_value in enumerate(self.proto_obj.value)
@@ -553,6 +616,7 @@ class ServiceCompiler(ProtoContentBase):
    def __post_init__(self) -> None:
        # Add service to output file
        self.output_file.services.append(self)
+        self.output_file.typing_imports.add("Dict")
        super().__post_init__()  # check for unset fields

    @property
@@ -576,11 +640,10 @@ class ServiceMethodCompiler(ProtoContentBase):
        # Add method to service
        self.parent.methods.append(self)

-        # Check for Optional import
+        # Check for imports
        if self.py_input_message:
            for f in self.py_input_message.fields:
-                if f.default_value_string == "None":
-                    self.output_file.typing_imports.add("Optional")
+                f.add_imports_to(self.output_file)
        if "Optional" in self.py_output_message_type:
            self.output_file.typing_imports.add("Optional")
        self.mutable_default_args  # ensure this is called before rendering
@@ -590,7 +653,9 @@ class ServiceMethodCompiler(ProtoContentBase):
            self.output_file.typing_imports.add("AsyncIterable")
            self.output_file.typing_imports.add("Iterable")
            self.output_file.typing_imports.add("Union")
-        if self.server_streaming:
+
+        # Required by both client and server
+        if self.client_streaming or self.server_streaming:
            self.output_file.typing_imports.add("AsyncIterator")

        super().__post_init__()  # check for unset fields
@@ -638,7 +703,10 @@ class ServiceMethodCompiler(ProtoContentBase):

    @property
    def route(self) -> str:
-        return f"/{self.output_file.package}.{self.parent.proto_name}/{self.proto_name}"
+        package_part = (
+            f"{self.output_file.package}." if self.output_file.package else ""
+        )
+        return f"/{package_part}{self.parent.proto_name}/{self.proto_name}"

    @property
    def py_input_message(self) -> Optional[MessageCompiler]:
--- a/src/betterproto/plugin/parser.py
+++ b/src/betterproto/plugin/parser.py
@@ -1,28 +1,19 @@
+from betterproto.lib.google.protobuf import (
+    DescriptorProto,
+    EnumDescriptorProto,
+    FieldDescriptorProto,
+    FileDescriptorProto,
+    ServiceDescriptorProto,
+)
+from betterproto.lib.google.protobuf.compiler import (
+    CodeGeneratorRequest,
+    CodeGeneratorResponse,
+    CodeGeneratorResponseFile,
+)
 import itertools
 import pathlib
 import sys
-from typing import TYPE_CHECKING, Iterator, List, Tuple, Union, Set
-
-try:
-    # betterproto[compiler] specific dependencies
-    from google.protobuf.compiler import plugin_pb2 as plugin
-    from google.protobuf.descriptor_pb2 import (
-        DescriptorProto,
-        EnumDescriptorProto,
-        FieldDescriptorProto,
-        ServiceDescriptorProto,
-    )
-except ImportError as err:
-    print(
-        "\033[31m"
-        f"Unable to import `{err.name}` from betterproto plugin! "
-        "Please ensure that you've installed betterproto as "
-        '`pip install "betterproto[compiler]"` so that compiler dependencies '
-        "are included."
-        "\033[0m"
-    )
-    raise SystemExit(1)
-
+from typing import Iterator, List, Set, Tuple, TYPE_CHECKING, Union
 from .compiler import outputfile_compiler
 from .models import (
    EnumDefinitionCompiler,
@@ -47,7 +38,7 @@ def traverse(
 ) -> "itertools.chain[Tuple[Union[str, EnumDescriptorProto], List[int]]]":
    # Todo: Keep information about nested hierarchy
    def _traverse(
-        path: List[int], items: List["Descriptor"], prefix=""
+        path: List[int], items: List["EnumDescriptorProto"], prefix=""
    ) -> Iterator[Tuple[Union[str, EnumDescriptorProto], List[int]]]:
        for i, item in enumerate(items):
            # Adjust the name since we flatten the hierarchy.
@@ -70,7 +61,7 @@ def traverse(


 def generate_code(
-    request: plugin.CodeGeneratorRequest, response: plugin.CodeGeneratorResponse
+    request: CodeGeneratorRequest, response: CodeGeneratorResponse
 ) -> None:
    plugin_options = request.parameter.split(",") if request.parameter else []

@@ -100,7 +91,12 @@ def generate_code(
    for output_package_name, output_package in request_data.output_packages.items():
        for proto_input_file in output_package.input_files:
            for item, path in traverse(proto_input_file):
-                read_protobuf_type(item=item, path=path, output_package=output_package)
+                read_protobuf_type(
+                    source_file=proto_input_file,
+                    item=item,
+                    path=path,
+                    output_package=output_package,
+                )

    # Read Services
    for output_package_name, output_package in request_data.output_packages.items():
@@ -116,11 +112,13 @@ def generate_code(
        output_path = pathlib.Path(*output_package_name.split("."), "__init__.py")
        output_paths.add(output_path)

-        f: response.File = response.file.add()
-        f.name = str(output_path)
-
-        # Render and then format the output file
-        f.content = outputfile_compiler(output_file=output_package)
+        response.file.append(
+            CodeGeneratorResponseFile(
+                name=str(output_path),
+                # Render and then format the output file
+                content=outputfile_compiler(output_file=output_package),
+            )
+        )

    # Make each output directory a package with __init__ file
    init_files = {
@@ -130,38 +128,53 @@ def generate_code(
    } - output_paths

    for init_file in init_files:
-        init = response.file.add()
-        init.name = str(init_file)
+        response.file.append(CodeGeneratorResponseFile(name=str(init_file)))

    for output_package_name in sorted(output_paths.union(init_files)):
        print(f"Writing {output_package_name}", file=sys.stderr)


 def read_protobuf_type(
-    item: DescriptorProto, path: List[int], output_package: OutputTemplate
+    item: DescriptorProto,
+    path: List[int],
+    source_file: "FileDescriptorProto",
+    output_package: OutputTemplate,
 ) -> None:
    if isinstance(item, DescriptorProto):
        if item.options.map_entry:
            # Skip generated map entry messages since we just use dicts
            return
        # Process Message
-        message_data = MessageCompiler(parent=output_package, proto_obj=item, path=path)
+        message_data = MessageCompiler(
+            source_file=source_file, parent=output_package, proto_obj=item, path=path
+        )
        for index, field in enumerate(item.field):
            if is_map(field, item):
                MapEntryCompiler(
-                    parent=message_data, proto_obj=field, path=path + [2, index]
+                    source_file=source_file,
+                    parent=message_data,
+                    proto_obj=field,
+                    path=path + [2, index],
                )
            elif is_oneof(field):
                OneOfFieldCompiler(
-                    parent=message_data, proto_obj=field, path=path + [2, index]
+                    source_file=source_file,
+                    parent=message_data,
+                    proto_obj=field,
+                    path=path + [2, index],
                )
            else:
                FieldCompiler(
-                    parent=message_data, proto_obj=field, path=path + [2, index]
+                    source_file=source_file,
+                    parent=message_data,
+                    proto_obj=field,
+                    path=path + [2, index],
                )
    elif isinstance(item, EnumDescriptorProto):
        # Enum
-        EnumDefinitionCompiler(parent=output_package, proto_obj=item, path=path)
+        EnumDefinitionCompiler(
+            source_file=source_file, parent=output_package, proto_obj=item, path=path
+        )


 def read_protobuf_service(
--- a/src/betterproto/templates/template.py.j2
+++ b/src/betterproto/templates/template.py.j2
@@ -15,6 +15,7 @@ from typing import {% for i in output_file.typing_imports|sort %}{{ i }}{% if no
 {% endif %}

 import betterproto
+from betterproto.grpc.grpclib_server import ServiceBase
 {% if output_file.services %}
 import grpclib
 {% endif %}
@@ -53,7 +54,7 @@ class {{ message.py_name }}(betterproto.Message):
    pass
    {% endif %}

-    {% if message.deprecated or message.deprecated_fields %}
+    {% if message.deprecated or message.has_deprecated_fields %}
    def __post_init__(self) -> None:
        {% if message.deprecated %}
        warnings.warn("{{ message.py_name }} is deprecated", DeprecationWarning)
@@ -72,6 +73,8 @@ class {{ service.py_name }}Stub(betterproto.ServiceStub):
    {% if service.comment %}
 {{ service.comment }}

+    {% elif not service.methods %}
+    pass
    {% endif %}
    {% for method in service.methods %}
    async def {{ method.py_name }}(self
@@ -82,7 +85,7 @@ class {{ service.py_name }}Stub(betterproto.ServiceStub):
                                            Optional[{{ field.annotation }}]
                                         {%- else -%}
                                            {{ field.annotation }}
-                                         {%- endif -%} = 
+                                         {%- endif -%} =
                                            {%- if field.py_name not in method.mutable_default_args -%}
                                                {{ field.default_value_string }}
                                            {%- else -%}
@@ -154,6 +157,89 @@ class {{ service.py_name }}Stub(betterproto.ServiceStub):
    {% endfor %}
 {% endfor %}

+{% for service in output_file.services %}
+class {{ service.py_name }}Base(ServiceBase):
+    {% if service.comment %}
+{{ service.comment }}
+
+    {% endif %}
+
+    {% for method in service.methods %}
+    async def {{ method.py_name }}(self
+        {%- if not method.client_streaming -%}
+            {%- if method.py_input_message and method.py_input_message.fields -%},
+                {%- for field in method.py_input_message.fields -%}
+                    {{ field.py_name }}: {% if field.py_name in method.mutable_default_args and not field.annotation.startswith("Optional[") -%}
+                                            Optional[{{ field.annotation }}]
+                                         {%- else -%}
+                                            {{ field.annotation }}
+                                         {%- endif -%}
+                    {%- if not loop.last %}, {% endif -%}
+                {%- endfor -%}
+            {%- endif -%}
+        {%- else -%}
+            {# Client streaming: need a request iterator instead #}
+            , request_iterator: AsyncIterator["{{ method.py_input_message_type }}"]
+        {%- endif -%}
+            ) -> {% if method.server_streaming %}AsyncIterator["{{ method.py_output_message_type }}"]{% else %}"{{ method.py_output_message_type }}"{% endif %}:
+        {% if method.comment %}
+{{ method.comment }}
+
+        {% endif %}
+        raise grpclib.GRPCError(grpclib.const.Status.UNIMPLEMENTED)
+
+    {% endfor %}
+
+    {% for method in service.methods %}
+    async def __rpc_{{ method.py_name }}(self, stream: grpclib.server.Stream) -> None:
+        {% if not method.client_streaming %}
+        request = await stream.recv_message()
+
+        request_kwargs = {
+        {% for field in method.py_input_message.fields %}
+            "{{ field.py_name }}": request.{{ field.py_name }},
+        {% endfor %}
+        }
+
+        {% else %}
+        request_kwargs = {"request_iterator": stream.__aiter__()}
+        {% endif %}
+
+        {% if not method.server_streaming %}
+        response = await self.{{ method.py_name }}(**request_kwargs)
+        await stream.send_message(response)
+        {% else %}
+        await self._call_rpc_handler_server_stream(
+            self.{{ method.py_name }},
+            stream,
+            request_kwargs,
+        )
+        {% endif %}
+
+    {% endfor %}
+
+    def __mapping__(self) -> Dict[str, grpclib.const.Handler]:
+        return {
+        {% for method in service.methods %}
+        "{{ method.route }}": grpclib.const.Handler(
+            self.__rpc_{{ method.py_name }},
+            {% if not method.client_streaming and not method.server_streaming %}
+            grpclib.const.Cardinality.UNARY_UNARY,
+            {% elif not method.client_streaming and method.server_streaming %}
+            grpclib.const.Cardinality.UNARY_STREAM,
+            {% elif method.client_streaming and not method.server_streaming %}
+            grpclib.const.Cardinality.STREAM_UNARY,
+            {% else %}
+            grpclib.const.Cardinality.STREAM_STREAM,
+            {% endif %}
+            {{ method.py_input_message_type }},
+            {{ method.py_output_message_type }},
+        ),
+        {% endfor %}
+        }
+
+{% endfor %}
+
 {% for i in output_file.imports|sort %}
 {{ i }}
 {% endfor %}
--- a/tests/inputs/config.py
+++ b/tests/inputs/config.py
@@ -1,9 +1,7 @@
 # Test cases that are expected to fail, e.g. unimplemented features or bug-fixes.
 # Remove from list when fixed.
 xfail = {
-    "oneof_enum",  # 63
    "namespace_keywords",  # 70
-    "namespace_builtin_types",  # 53
    "googletypes_struct",  # 9
    "googletypes_value",  # 9
    "import_capitalized_package",
@@ -14,7 +12,17 @@ services = {
    "googletypes_response",
    "googletypes_response_embedded",
    "service",
+    "service_separate_packages",
    "import_service_input_message",
    "googletypes_service_returns_empty",
    "googletypes_service_returns_googletype",
+    "example_service",
+    "empty_service",
 }
+
+
+# Indicate json sample messages to skip when testing that json (de)serialization
+# is symmetrical becuase some cases legitimately are not symmetrical.
+# Each key references the name of the test scenario and the values in the tuple
+# Are the names of the json files.
+non_symmetrical_json = {"empty_repeated": ("empty_repeated",)}
--- a/tests/inputs/empty_repeated/empty_repeated.json
+++ b/tests/inputs/empty_repeated/empty_repeated.json
@@ -0,0 +1,3 @@
+{
+  "msg": [{"values":[]}]
+}
--- a/tests/inputs/empty_repeated/empty_repeated.proto
+++ b/tests/inputs/empty_repeated/empty_repeated.proto
@@ -0,0 +1,9 @@
+syntax = "proto3";
+
+message MessageA {
+  repeated float values = 1;
+}
+
+message Test {
+  repeated MessageA msg = 1;
+}
--- a/tests/inputs/empty_service/empty_service.proto
+++ b/tests/inputs/empty_service/empty_service.proto
@@ -0,0 +1,7 @@
+/* Empty service without comments */
+syntax = "proto3";
+
+package empty_service;
+
+service Test {
+}
--- a/tests/inputs/example/example.proto
+++ b/tests/inputs/example/example.proto
@@ -1,8 +1,909 @@
-syntax = "proto3";
+// Protocol Buffers - Google's data interchange format
+// Copyright 2008 Google Inc.  All rights reserved.
+// https://developers.google.com/protocol-buffers/
+//
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+// notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+// copyright notice, this list of conditions and the following disclaimer
+// in the documentation and/or other materials provided with the
+// distribution.
+//     * Neither the name of Google Inc. nor the names of its
+// contributors may be used to endorse or promote products derived from
+// this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

-package hello;
+// Author: kenton@google.com (Kenton Varda)
+//  Based on original Protocol Buffers design by
+//  Sanjay Ghemawat, Jeff Dean, and others.
+//
+// The messages in this file describe the definitions found in .proto files.
+// A valid .proto file can be translated directly to a FileDescriptorProto
+// without any other information (e.g. without reading its imports).

-// Greeting represents a message you can tell a user.
-message Greeting {
-  string message = 1;
+
+syntax = "proto2";
+
+// package google.protobuf;
+
+option go_package = "google.golang.org/protobuf/types/descriptorpb";
+option java_package = "com.google.protobuf";
+option java_outer_classname = "DescriptorProtos";
+option csharp_namespace = "Google.Protobuf.Reflection";
+option objc_class_prefix = "GPB";
+option cc_enable_arenas = true;
+
+// descriptor.proto must be optimized for speed because reflection-based
+// algorithms don't work during bootstrapping.
+option optimize_for = SPEED;
+
+// The protocol compiler can output a FileDescriptorSet containing the .proto
+// files it parses.
+message FileDescriptorSet {
+  repeated FileDescriptorProto file = 1;
+}
+
+// Describes a complete .proto file.
+message FileDescriptorProto {
+  optional string name = 1;     // file name, relative to root of source tree
+  optional string package = 2;  // e.g. "foo", "foo.bar", etc.
+
+  // Names of files imported by this file.
+  repeated string dependency = 3;
+  // Indexes of the public imported files in the dependency list above.
+  repeated int32 public_dependency = 10;
+  // Indexes of the weak imported files in the dependency list.
+  // For Google-internal migration only. Do not use.
+  repeated int32 weak_dependency = 11;
+
+  // All top-level definitions in this file.
+  repeated DescriptorProto message_type = 4;
+  repeated EnumDescriptorProto enum_type = 5;
+  repeated ServiceDescriptorProto service = 6;
+  repeated FieldDescriptorProto extension = 7;
+
+  optional FileOptions options = 8;
+
+  // This field contains optional information about the original source code.
+  // You may safely remove this entire field without harming runtime
+  // functionality of the descriptors -- the information is needed only by
+  // development tools.
+  optional SourceCodeInfo source_code_info = 9;
+
+  // The syntax of the proto file.
+  // The supported values are "proto2" and "proto3".
+  optional string syntax = 12;
+}
+
+// Describes a message type.
+message DescriptorProto {
+  optional string name = 1;
+
+  repeated FieldDescriptorProto field = 2;
+  repeated FieldDescriptorProto extension = 6;
+
+  repeated DescriptorProto nested_type = 3;
+  repeated EnumDescriptorProto enum_type = 4;
+
+  message ExtensionRange {
+    optional int32 start = 1;  // Inclusive.
+    optional int32 end = 2;    // Exclusive.
+
+    optional ExtensionRangeOptions options = 3;
+  }
+  repeated ExtensionRange extension_range = 5;
+
+  repeated OneofDescriptorProto oneof_decl = 8;
+
+  optional MessageOptions options = 7;
+
+  // Range of reserved tag numbers. Reserved tag numbers may not be used by
+  // fields or extension ranges in the same message. Reserved ranges may
+  // not overlap.
+  message ReservedRange {
+    optional int32 start = 1;  // Inclusive.
+    optional int32 end = 2;    // Exclusive.
+  }
+  repeated ReservedRange reserved_range = 9;
+  // Reserved field names, which may not be used by fields in the same message.
+  // A given name may only be reserved once.
+  repeated string reserved_name = 10;
+}
+
+message ExtensionRangeOptions {
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+// Describes a field within a message.
+message FieldDescriptorProto {
+  enum Type {
+    // 0 is reserved for errors.
+    // Order is weird for historical reasons.
+    TYPE_DOUBLE = 1;
+    TYPE_FLOAT = 2;
+    // Not ZigZag encoded.  Negative numbers take 10 bytes.  Use TYPE_SINT64 if
+    // negative values are likely.
+    TYPE_INT64 = 3;
+    TYPE_UINT64 = 4;
+    // Not ZigZag encoded.  Negative numbers take 10 bytes.  Use TYPE_SINT32 if
+    // negative values are likely.
+    TYPE_INT32 = 5;
+    TYPE_FIXED64 = 6;
+    TYPE_FIXED32 = 7;
+    TYPE_BOOL = 8;
+    TYPE_STRING = 9;
+    // Tag-delimited aggregate.
+    // Group type is deprecated and not supported in proto3. However, Proto3
+    // implementations should still be able to parse the group wire format and
+    // treat group fields as unknown fields.
+    TYPE_GROUP = 10;
+    TYPE_MESSAGE = 11;  // Length-delimited aggregate.
+
+    // New in version 2.
+    TYPE_BYTES = 12;
+    TYPE_UINT32 = 13;
+    TYPE_ENUM = 14;
+    TYPE_SFIXED32 = 15;
+    TYPE_SFIXED64 = 16;
+    TYPE_SINT32 = 17;  // Uses ZigZag encoding.
+    TYPE_SINT64 = 18;  // Uses ZigZag encoding.
+  }
+
+  enum Label {
+    // 0 is reserved for errors
+    LABEL_OPTIONAL = 1;
+    LABEL_REQUIRED = 2;
+    LABEL_REPEATED = 3;
+  }
+
+  optional string name = 1;
+  optional int32 number = 3;
+  optional Label label = 4;
+
+  // If type_name is set, this need not be set.  If both this and type_name
+  // are set, this must be one of TYPE_ENUM, TYPE_MESSAGE or TYPE_GROUP.
+  optional Type type = 5;
+
+  // For message and enum types, this is the name of the type.  If the name
+  // starts with a '.', it is fully-qualified.  Otherwise, C++-like scoping
+  // rules are used to find the type (i.e. first the nested types within this
+  // message are searched, then within the parent, on up to the root
+  // namespace).
+  optional string type_name = 6;
+
+  // For extensions, this is the name of the type being extended.  It is
+  // resolved in the same manner as type_name.
+  optional string extendee = 2;
+
+  // For numeric types, contains the original text representation of the value.
+  // For booleans, "true" or "false".
+  // For strings, contains the default text contents (not escaped in any way).
+  // For bytes, contains the C escaped value.  All bytes >= 128 are escaped.
+  // TODO(kenton):  Base-64 encode?
+  optional string default_value = 7;
+
+  // If set, gives the index of a oneof in the containing type's oneof_decl
+  // list.  This field is a member of that oneof.
+  optional int32 oneof_index = 9;
+
+  // JSON name of this field. The value is set by protocol compiler. If the
+  // user has set a "json_name" option on this field, that option's value
+  // will be used. Otherwise, it's deduced from the field's name by converting
+  // it to camelCase.
+  optional string json_name = 10;
+
+  optional FieldOptions options = 8;
+
+  // If true, this is a proto3 "optional". When a proto3 field is optional, it
+  // tracks presence regardless of field type.
+  //
+  // When proto3_optional is true, this field must be belong to a oneof to
+  // signal to old proto3 clients that presence is tracked for this field. This
+  // oneof is known as a "synthetic" oneof, and this field must be its sole
+  // member (each proto3 optional field gets its own synthetic oneof). Synthetic
+  // oneofs exist in the descriptor only, and do not generate any API. Synthetic
+  // oneofs must be ordered after all "real" oneofs.
+  //
+  // For message fields, proto3_optional doesn't create any semantic change,
+  // since non-repeated message fields always track presence. However it still
+  // indicates the semantic detail of whether the user wrote "optional" or not.
+  // This can be useful for round-tripping the .proto file. For consistency we
+  // give message fields a synthetic oneof also, even though it is not required
+  // to track presence. This is especially important because the parser can't
+  // tell if a field is a message or an enum, so it must always create a
+  // synthetic oneof.
+  //
+  // Proto2 optional fields do not set this flag, because they already indicate
+  // optional with `LABEL_OPTIONAL`.
+  optional bool proto3_optional = 17;
+}
+
+// Describes a oneof.
+message OneofDescriptorProto {
+  optional string name = 1;
+  optional OneofOptions options = 2;
+}
+
+// Describes an enum type.
+message EnumDescriptorProto {
+  optional string name = 1;
+
+  repeated EnumValueDescriptorProto value = 2;
+
+  optional EnumOptions options = 3;
+
+  // Range of reserved numeric values. Reserved values may not be used by
+  // entries in the same enum. Reserved ranges may not overlap.
+  //
+  // Note that this is distinct from DescriptorProto.ReservedRange in that it
+  // is inclusive such that it can appropriately represent the entire int32
+  // domain.
+  message EnumReservedRange {
+    optional int32 start = 1;  // Inclusive.
+    optional int32 end = 2;    // Inclusive.
+  }
+
+  // Range of reserved numeric values. Reserved numeric values may not be used
+  // by enum values in the same enum declaration. Reserved ranges may not
+  // overlap.
+  repeated EnumReservedRange reserved_range = 4;
+
+  // Reserved enum value names, which may not be reused. A given name may only
+  // be reserved once.
+  repeated string reserved_name = 5;
+}
+
+// Describes a value within an enum.
+message EnumValueDescriptorProto {
+  optional string name = 1;
+  optional int32 number = 2;
+
+  optional EnumValueOptions options = 3;
+}
+
+// Describes a service.
+message ServiceDescriptorProto {
+  optional string name = 1;
+  repeated MethodDescriptorProto method = 2;
+
+  optional ServiceOptions options = 3;
+}
+
+// Describes a method of a service.
+message MethodDescriptorProto {
+  optional string name = 1;
+
+  // Input and output type names.  These are resolved in the same way as
+  // FieldDescriptorProto.type_name, but must refer to a message type.
+  optional string input_type = 2;
+  optional string output_type = 3;
+
+  optional MethodOptions options = 4;
+
+  // Identifies if client streams multiple client messages
+  optional bool client_streaming = 5 [default = false];
+  // Identifies if server streams multiple server messages
+  optional bool server_streaming = 6 [default = false];
+}
+
+
+// ===================================================================
+// Options
+
+// Each of the definitions above may have "options" attached.  These are
+// just annotations which may cause code to be generated slightly differently
+// or may contain hints for code that manipulates protocol messages.
+//
+// Clients may define custom options as extensions of the *Options messages.
+// These extensions may not yet be known at parsing time, so the parser cannot
+// store the values in them.  Instead it stores them in a field in the *Options
+// message called uninterpreted_option. This field must have the same name
+// across all *Options messages. We then use this field to populate the
+// extensions when we build a descriptor, at which point all protos have been
+// parsed and so all extensions are known.
+//
+// Extension numbers for custom options may be chosen as follows:
+// * For options which will only be used within a single application or
+//   organization, or for experimental options, use field numbers 50000
+//   through 99999.  It is up to you to ensure that you do not use the
+//   same number for multiple options.
+// * For options which will be published and used publicly by multiple
+//   independent entities, e-mail protobuf-global-extension-registry@google.com
+//   to reserve extension numbers. Simply provide your project name (e.g.
+//   Objective-C plugin) and your project website (if available) -- there's no
+//   need to explain how you intend to use them. Usually you only need one
+//   extension number. You can declare multiple options with only one extension
+//   number by putting them in a sub-message. See the Custom Options section of
+//   the docs for examples:
+//   https://developers.google.com/protocol-buffers/docs/proto#options
+//   If this turns out to be popular, a web service will be set up
+//   to automatically assign option numbers.
+
+message FileOptions {
+
+  // Sets the Java package where classes generated from this .proto will be
+  // placed.  By default, the proto package is used, but this is often
+  // inappropriate because proto packages do not normally start with backwards
+  // domain names.
+  optional string java_package = 1;
+
+
+  // If set, all the classes from the .proto file are wrapped in a single
+  // outer class with the given name.  This applies to both Proto1
+  // (equivalent to the old "--one_java_file" option) and Proto2 (where
+  // a .proto always translates to a single class, but you may want to
+  // explicitly choose the class name).
+  optional string java_outer_classname = 8;
+
+  // If set true, then the Java code generator will generate a separate .java
+  // file for each top-level message, enum, and service defined in the .proto
+  // file.  Thus, these types will *not* be nested inside the outer class
+  // named by java_outer_classname.  However, the outer class will still be
+  // generated to contain the file's getDescriptor() method as well as any
+  // top-level extensions defined in the file.
+  optional bool java_multiple_files = 10 [default = false];
+
+  // This option does nothing.
+  optional bool java_generate_equals_and_hash = 20 [deprecated=true];
+
+  // If set true, then the Java2 code generator will generate code that
+  // throws an exception whenever an attempt is made to assign a non-UTF-8
+  // byte sequence to a string field.
+  // Message reflection will do the same.
+  // However, an extension field still accepts non-UTF-8 byte sequences.
+  // This option has no effect on when used with the lite runtime.
+  optional bool java_string_check_utf8 = 27 [default = false];
+
+
+  // Generated classes can be optimized for speed or code size.
+  enum OptimizeMode {
+    SPEED = 1;         // Generate complete code for parsing, serialization,
+                       // etc.
+    CODE_SIZE = 2;     // Use ReflectionOps to implement these methods.
+    LITE_RUNTIME = 3;  // Generate code using MessageLite and the lite runtime.
+  }
+  optional OptimizeMode optimize_for = 9 [default = SPEED];
+
+  // Sets the Go package where structs generated from this .proto will be
+  // placed. If omitted, the Go package will be derived from the following:
+  //   - The basename of the package import path, if provided.
+  //   - Otherwise, the package statement in the .proto file, if present.
+  //   - Otherwise, the basename of the .proto file, without extension.
+  optional string go_package = 11;
+
+
+
+
+  // Should generic services be generated in each language?  "Generic" services
+  // are not specific to any particular RPC system.  They are generated by the
+  // main code generators in each language (without additional plugins).
+  // Generic services were the only kind of service generation supported by
+  // early versions of google.protobuf.
+  //
+  // Generic services are now considered deprecated in favor of using plugins
+  // that generate code specific to your particular RPC system.  Therefore,
+  // these default to false.  Old code which depends on generic services should
+  // explicitly set them to true.
+  optional bool cc_generic_services = 16 [default = false];
+  optional bool java_generic_services = 17 [default = false];
+  optional bool py_generic_services = 18 [default = false];
+  optional bool php_generic_services = 42 [default = false];
+
+  // Is this file deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for everything in the file, or it will be completely ignored; in the very
+  // least, this is a formalization for deprecating files.
+  optional bool deprecated = 23 [default = false];
+
+  // Enables the use of arenas for the proto messages in this file. This applies
+  // only to generated classes for C++.
+  optional bool cc_enable_arenas = 31 [default = true];
+
+
+  // Sets the objective c class prefix which is prepended to all objective c
+  // generated classes from this .proto. There is no default.
+  optional string objc_class_prefix = 36;
+
+  // Namespace for generated classes; defaults to the package.
+  optional string csharp_namespace = 37;
+
+  // By default Swift generators will take the proto package and CamelCase it
+  // replacing '.' with underscore and use that to prefix the types/symbols
+  // defined. When this options is provided, they will use this value instead
+  // to prefix the types/symbols defined.
+  optional string swift_prefix = 39;
+
+  // Sets the php class prefix which is prepended to all php generated classes
+  // from this .proto. Default is empty.
+  optional string php_class_prefix = 40;
+
+  // Use this option to change the namespace of php generated classes. Default
+  // is empty. When this option is empty, the package name will be used for
+  // determining the namespace.
+  optional string php_namespace = 41;
+
+  // Use this option to change the namespace of php generated metadata classes.
+  // Default is empty. When this option is empty, the proto file name will be
+  // used for determining the namespace.
+  optional string php_metadata_namespace = 44;
+
+  // Use this option to change the package of ruby generated classes. Default
+  // is empty. When this option is not set, the package name will be used for
+  // determining the ruby package.
+  optional string ruby_package = 45;
+
+
+  // The parser stores options it doesn't recognize here.
+  // See the documentation for the "Options" section above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message.
+  // See the documentation for the "Options" section above.
+  extensions 1000 to max;
+
+  reserved 38;
+}
+
+message MessageOptions {
+  // Set true to use the old proto1 MessageSet wire format for extensions.
+  // This is provided for backwards-compatibility with the MessageSet wire
+  // format.  You should not use this for any other reason:  It's less
+  // efficient, has fewer features, and is more complicated.
+  //
+  // The message must be defined exactly as follows:
+  //   message Foo {
+  //     option message_set_wire_format = true;
+  //     extensions 4 to max;
+  //   }
+  // Note that the message cannot have any defined fields; MessageSets only
+  // have extensions.
+  //
+  // All extensions of your type must be singular messages; e.g. they cannot
+  // be int32s, enums, or repeated messages.
+  //
+  // Because this is an option, the above two restrictions are not enforced by
+  // the protocol compiler.
+  optional bool message_set_wire_format = 1 [default = false];
+
+  // Disables the generation of the standard "descriptor()" accessor, which can
+  // conflict with a field of the same name.  This is meant to make migration
+  // from proto1 easier; new code should avoid fields named "descriptor".
+  optional bool no_standard_descriptor_accessor = 2 [default = false];
+
+  // Is this message deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for the message, or it will be completely ignored; in the very least,
+  // this is a formalization for deprecating messages.
+  optional bool deprecated = 3 [default = false];
+
+  // Whether the message is an automatically generated map entry type for the
+  // maps field.
+  //
+  // For maps fields:
+  //     map<KeyType, ValueType> map_field = 1;
+  // The parsed descriptor looks like:
+  //     message MapFieldEntry {
+  //         option map_entry = true;
+  //         optional KeyType key = 1;
+  //         optional ValueType value = 2;
+  //     }
+  //     repeated MapFieldEntry map_field = 1;
+  //
+  // Implementations may choose not to generate the map_entry=true message, but
+  // use a native map in the target language to hold the keys and values.
+  // The reflection APIs in such implementations still need to work as
+  // if the field is a repeated message field.
+  //
+  // NOTE: Do not set the option in .proto files. Always use the maps syntax
+  // instead. The option should only be implicitly set by the proto compiler
+  // parser.
+  optional bool map_entry = 7;
+
+  reserved 8;  // javalite_serializable
+  reserved 9;  // javanano_as_lite
+
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+message FieldOptions {
+  // The ctype option instructs the C++ code generator to use a different
+  // representation of the field than it normally would.  See the specific
+  // options below.  This option is not yet implemented in the open source
+  // release -- sorry, we'll try to include it in a future version!
+  optional CType ctype = 1 [default = STRING];
+  enum CType {
+    // Default mode.
+    STRING = 0;
+
+    CORD = 1;
+
+    STRING_PIECE = 2;
+  }
+  // The packed option can be enabled for repeated primitive fields to enable
+  // a more efficient representation on the wire. Rather than repeatedly
+  // writing the tag and type for each element, the entire array is encoded as
+  // a single length-delimited blob. In proto3, only explicit setting it to
+  // false will avoid using packed encoding.
+  optional bool packed = 2;
+
+  // The jstype option determines the JavaScript type used for values of the
+  // field.  The option is permitted only for 64 bit integral and fixed types
+  // (int64, uint64, sint64, fixed64, sfixed64).  A field with jstype JS_STRING
+  // is represented as JavaScript string, which avoids loss of precision that
+  // can happen when a large value is converted to a floating point JavaScript.
+  // Specifying JS_NUMBER for the jstype causes the generated JavaScript code to
+  // use the JavaScript "number" type.  The behavior of the default option
+  // JS_NORMAL is implementation dependent.
+  //
+  // This option is an enum to permit additional types to be added, e.g.
+  // goog.math.Integer.
+  optional JSType jstype = 6 [default = JS_NORMAL];
+  enum JSType {
+    // Use the default type.
+    JS_NORMAL = 0;
+
+    // Use JavaScript strings.
+    JS_STRING = 1;
+
+    // Use JavaScript numbers.
+    JS_NUMBER = 2;
+  }
+
+  // Should this field be parsed lazily?  Lazy applies only to message-type
+  // fields.  It means that when the outer message is initially parsed, the
+  // inner message's contents will not be parsed but instead stored in encoded
+  // form.  The inner message will actually be parsed when it is first accessed.
+  //
+  // This is only a hint.  Implementations are free to choose whether to use
+  // eager or lazy parsing regardless of the value of this option.  However,
+  // setting this option true suggests that the protocol author believes that
+  // using lazy parsing on this field is worth the additional bookkeeping
+  // overhead typically needed to implement it.
+  //
+  // This option does not affect the public interface of any generated code;
+  // all method signatures remain the same.  Furthermore, thread-safety of the
+  // interface is not affected by this option; const methods remain safe to
+  // call from multiple threads concurrently, while non-const methods continue
+  // to require exclusive access.
+  //
+  //
+  // Note that implementations may choose not to check required fields within
+  // a lazy sub-message.  That is, calling IsInitialized() on the outer message
+  // may return true even if the inner message has missing required fields.
+  // This is necessary because otherwise the inner message would have to be
+  // parsed in order to perform the check, defeating the purpose of lazy
+  // parsing.  An implementation which chooses not to check required fields
+  // must be consistent about it.  That is, for any particular sub-message, the
+  // implementation must either *always* check its required fields, or *never*
+  // check its required fields, regardless of whether or not the message has
+  // been parsed.
+  optional bool lazy = 5 [default = false];
+
+  // Is this field deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for accessors, or it will be completely ignored; in the very least, this
+  // is a formalization for deprecating fields.
+  optional bool deprecated = 3 [default = false];
+
+  // For Google-internal migration only. Do not use.
+  optional bool weak = 10 [default = false];
+
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+
+  reserved 4;  // removed jtype
+}
+
+message OneofOptions {
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+message EnumOptions {
+
+  // Set this option to true to allow mapping different tag names to the same
+  // value.
+  optional bool allow_alias = 2;
+
+  // Is this enum deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for the enum, or it will be completely ignored; in the very least, this
+  // is a formalization for deprecating enums.
+  optional bool deprecated = 3 [default = false];
+
+  reserved 5;  // javanano_as_lite
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+message EnumValueOptions {
+  // Is this enum value deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for the enum value, or it will be completely ignored; in the very least,
+  // this is a formalization for deprecating enum values.
+  optional bool deprecated = 1 [default = false];
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+message ServiceOptions {
+
+  // Note:  Field numbers 1 through 32 are reserved for Google's internal RPC
+  //   framework.  We apologize for hoarding these numbers to ourselves, but
+  //   we were already using them long before we decided to release Protocol
+  //   Buffers.
+
+  // Is this service deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for the service, or it will be completely ignored; in the very least,
+  // this is a formalization for deprecating services.
+  optional bool deprecated = 33 [default = false];
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+message MethodOptions {
+
+  // Note:  Field numbers 1 through 32 are reserved for Google's internal RPC
+  //   framework.  We apologize for hoarding these numbers to ourselves, but
+  //   we were already using them long before we decided to release Protocol
+  //   Buffers.
+
+  // Is this method deprecated?
+  // Depending on the target platform, this can emit Deprecated annotations
+  // for the method, or it will be completely ignored; in the very least,
+  // this is a formalization for deprecating methods.
+  optional bool deprecated = 33 [default = false];
+
+  // Is this method side-effect-free (or safe in HTTP parlance), or idempotent,
+  // or neither? HTTP based RPC implementation may choose GET verb for safe
+  // methods, and PUT verb for idempotent methods instead of the default POST.
+  enum IdempotencyLevel {
+    IDEMPOTENCY_UNKNOWN = 0;
+    NO_SIDE_EFFECTS = 1;  // implies idempotent
+    IDEMPOTENT = 2;       // idempotent, but may have side effects
+  }
+  optional IdempotencyLevel idempotency_level = 34
+      [default = IDEMPOTENCY_UNKNOWN];
+
+  // The parser stores options it doesn't recognize here. See above.
+  repeated UninterpretedOption uninterpreted_option = 999;
+
+  // Clients can define custom options in extensions of this message. See above.
+  extensions 1000 to max;
+}
+
+
+// A message representing a option the parser does not recognize. This only
+// appears in options protos created by the compiler::Parser class.
+// DescriptorPool resolves these when building Descriptor objects. Therefore,
+// options protos in descriptor objects (e.g. returned by Descriptor::options(),
+// or produced by Descriptor::CopyTo()) will never have UninterpretedOptions
+// in them.
+message UninterpretedOption {
+  // The name of the uninterpreted option.  Each string represents a segment in
+  // a dot-separated name.  is_extension is true iff a segment represents an
+  // extension (denoted with parentheses in options specs in .proto files).
+  // E.g.,{ ["foo", false], ["bar.baz", true], ["qux", false] } represents
+  // "foo.(bar.baz).qux".
+  message NamePart {
+    required string name_part = 1;
+    required bool is_extension = 2;
+  }
+  repeated NamePart name = 2;
+
+  // The value of the uninterpreted option, in whatever type the tokenizer
+  // identified it as during parsing. Exactly one of these should be set.
+  optional string identifier_value = 3;
+  optional uint64 positive_int_value = 4;
+  optional int64 negative_int_value = 5;
+  optional double double_value = 6;
+  optional bytes string_value = 7;
+  optional string aggregate_value = 8;
+}
+
+// ===================================================================
+// Optional source code info
+
+// Encapsulates information about the original source file from which a
+// FileDescriptorProto was generated.
+message SourceCodeInfo {
+  // A Location identifies a piece of source code in a .proto file which
+  // corresponds to a particular definition.  This information is intended
+  // to be useful to IDEs, code indexers, documentation generators, and similar
+  // tools.
+  //
+  // For example, say we have a file like:
+  //   message Foo {
+  //     optional string foo = 1;
+  //   }
+  // Let's look at just the field definition:
+  //   optional string foo = 1;
+  //   ^       ^^     ^^  ^  ^^^
+  //   a       bc     de  f  ghi
+  // We have the following locations:
+  //   span   path               represents
+  //   [a,i)  [ 4, 0, 2, 0 ]     The whole field definition.
+  //   [a,b)  [ 4, 0, 2, 0, 4 ]  The label (optional).
+  //   [c,d)  [ 4, 0, 2, 0, 5 ]  The type (string).
+  //   [e,f)  [ 4, 0, 2, 0, 1 ]  The name (foo).
+  //   [g,h)  [ 4, 0, 2, 0, 3 ]  The number (1).
+  //
+  // Notes:
+  // - A location may refer to a repeated field itself (i.e. not to any
+  //   particular index within it).  This is used whenever a set of elements are
+  //   logically enclosed in a single code segment.  For example, an entire
+  //   extend block (possibly containing multiple extension definitions) will
+  //   have an outer location whose path refers to the "extensions" repeated
+  //   field without an index.
+  // - Multiple locations may have the same path.  This happens when a single
+  //   logical declaration is spread out across multiple places.  The most
+  //   obvious example is the "extend" block again -- there may be multiple
+  //   extend blocks in the same scope, each of which will have the same path.
+  // - A location's span is not always a subset of its parent's span.  For
+  //   example, the "extendee" of an extension declaration appears at the
+  //   beginning of the "extend" block and is shared by all extensions within
+  //   the block.
+  // - Just because a location's span is a subset of some other location's span
+  //   does not mean that it is a descendant.  For example, a "group" defines
+  //   both a type and a field in a single declaration.  Thus, the locations
+  //   corresponding to the type and field and their components will overlap.
+  // - Code which tries to interpret locations should probably be designed to
+  //   ignore those that it doesn't understand, as more types of locations could
+  //   be recorded in the future.
+  repeated Location location = 1;
+  message Location {
+    // Identifies which part of the FileDescriptorProto was defined at this
+    // location.
+    //
+    // Each element is a field number or an index.  They form a path from
+    // the root FileDescriptorProto to the place where the definition.  For
+    // example, this path:
+    //   [ 4, 3, 2, 7, 1 ]
+    // refers to:
+    //   file.message_type(3)  // 4, 3
+    //       .field(7)         // 2, 7
+    //       .name()           // 1
+    // This is because FileDescriptorProto.message_type has field number 4:
+    //   repeated DescriptorProto message_type = 4;
+    // and DescriptorProto.field has field number 2:
+    //   repeated FieldDescriptorProto field = 2;
+    // and FieldDescriptorProto.name has field number 1:
+    //   optional string name = 1;
+    //
+    // Thus, the above path gives the location of a field name.  If we removed
+    // the last element:
+    //   [ 4, 3, 2, 7 ]
+    // this path refers to the whole field declaration (from the beginning
+    // of the label to the terminating semicolon).
+    repeated int32 path = 1 [packed = true];
+
+    // Always has exactly three or four elements: start line, start column,
+    // end line (optional, otherwise assumed same as start line), end column.
+    // These are packed into a single field for efficiency.  Note that line
+    // and column numbers are zero-based -- typically you will want to add
+    // 1 to each before displaying to a user.
+    repeated int32 span = 2 [packed = true];
+
+    // If this SourceCodeInfo represents a complete declaration, these are any
+    // comments appearing before and after the declaration which appear to be
+    // attached to the declaration.
+    //
+    // A series of line comments appearing on consecutive lines, with no other
+    // tokens appearing on those lines, will be treated as a single comment.
+    //
+    // leading_detached_comments will keep paragraphs of comments that appear
+    // before (but not connected to) the current element. Each paragraph,
+    // separated by empty lines, will be one comment element in the repeated
+    // field.
+    //
+    // Only the comment content is provided; comment markers (e.g. //) are
+    // stripped out.  For block comments, leading whitespace and an asterisk
+    // will be stripped from the beginning of each line other than the first.
+    // Newlines are included in the output.
+    //
+    // Examples:
+    //
+    //   optional int32 foo = 1;  // Comment attached to foo.
+    //   // Comment attached to bar.
+    //   optional int32 bar = 2;
+    //
+    //   optional string baz = 3;
+    //   // Comment attached to baz.
+    //   // Another line attached to baz.
+    //
+    //   // Comment attached to qux.
+    //   //
+    //   // Another line attached to qux.
+    //   optional double qux = 4;
+    //
+    //   // Detached comment for corge. This is not leading or trailing comments
+    //   // to qux or corge because there are blank lines separating it from
+    //   // both.
+    //
+    //   // Detached comment for corge paragraph 2.
+    //
+    //   optional string corge = 5;
+    //   /* Block comment attached
+    //    * to corge.  Leading asterisks
+    //    * will be removed. */
+    //   /* Block comment attached to
+    //    * grault. */
+    //   optional int32 grault = 6;
+    //
+    //   // ignored detached comments.
+    optional string leading_comments = 3;
+    optional string trailing_comments = 4;
+    repeated string leading_detached_comments = 6;
+  }
+}
+
+// Describes the relationship between generated code and its original source
+// file. A GeneratedCodeInfo message is associated with only one generated
+// source file, but may contain references to different source .proto files.
+message GeneratedCodeInfo {
+  // An Annotation connects some span of text in generated code to an element
+  // of its generating .proto file.
+  repeated Annotation annotation = 1;
+  message Annotation {
+    // Identifies the element in the original source .proto file. This field
+    // is formatted the same as SourceCodeInfo.Location.path.
+    repeated int32 path = 1 [packed = true];
+
+    // Identifies the filesystem path to the original source .proto.
+    optional string source_file = 2;
+
+    // Identifies the starting offset in bytes in the generated code
+    // that relates to the identified object.
+    optional int32 begin = 3;
+
+    // Identifies the ending offset in bytes in the generated code that
+    // relates to the identified offset. The end offset should be one past
+    // the last relevant byte (so the length of the text = end - begin).
+    optional int32 end = 4;
+  }
 }
--- a/tests/inputs/example_service/example_service.proto
+++ b/tests/inputs/example_service/example_service.proto
@@ -0,0 +1,20 @@
+syntax = "proto3";
+
+package example_service;
+
+service Test {
+  rpc ExampleUnaryUnary(ExampleRequest) returns (ExampleResponse);
+  rpc ExampleUnaryStream(ExampleRequest) returns (stream ExampleResponse);
+  rpc ExampleStreamUnary(stream ExampleRequest) returns (ExampleResponse);
+  rpc ExampleStreamStream(stream ExampleRequest) returns (stream ExampleResponse);
+}
+
+message ExampleRequest {
+  string example_string = 1;
+  int64  example_integer = 2;
+}
+
+message ExampleResponse {
+  string example_string = 1;
+  int64  example_integer = 2;
+}
--- a/tests/inputs/example_service/test_example_service.py
+++ b/tests/inputs/example_service/test_example_service.py
@@ -0,0 +1,95 @@
+from typing import AsyncIterator, AsyncIterable
+
+import pytest
+from grpclib.testing import ChannelFor
+
+from tests.output_betterproto.example_service.example_service import (
+    TestBase,
+    TestStub,
+    ExampleRequest,
+    ExampleResponse,
+)
+
+
+class ExampleService(TestBase):
+    async def example_unary_unary(
+        self, example_string: str, example_integer: int
+    ) -> "ExampleResponse":
+        return ExampleResponse(
+            example_string=example_string,
+            example_integer=example_integer,
+        )
+
+    async def example_unary_stream(
+        self, example_string: str, example_integer: int
+    ) -> AsyncIterator["ExampleResponse"]:
+        response = ExampleResponse(
+            example_string=example_string,
+            example_integer=example_integer,
+        )
+        yield response
+        yield response
+        yield response
+
+    async def example_stream_unary(
+        self, request_iterator: AsyncIterator["ExampleRequest"]
+    ) -> "ExampleResponse":
+        async for example_request in request_iterator:
+            return ExampleResponse(
+                example_string=example_request.example_string,
+                example_integer=example_request.example_integer,
+            )
+
+    async def example_stream_stream(
+        self, request_iterator: AsyncIterator["ExampleRequest"]
+    ) -> AsyncIterator["ExampleResponse"]:
+        async for example_request in request_iterator:
+            yield ExampleResponse(
+                example_string=example_request.example_string,
+                example_integer=example_request.example_integer,
+            )
+
+
+@pytest.mark.asyncio
+async def test_calls_with_different_cardinalities():
+    test_string = "test string"
+    test_int = 42
+
+    async with ChannelFor([ExampleService()]) as channel:
+        stub = TestStub(channel)
+
+        # unary unary
+        response = await stub.example_unary_unary(
+            example_string="test string",
+            example_integer=42,
+        )
+        assert response.example_string == test_string
+        assert response.example_integer == test_int
+
+        # unary stream
+        async for response in stub.example_unary_stream(
+            example_string="test string",
+            example_integer=42,
+        ):
+            assert response.example_string == test_string
+            assert response.example_integer == test_int
+
+        # stream unary
+        request = ExampleRequest(
+            example_string=test_string,
+            example_integer=42,
+        )
+
+        async def request_iterator():
+            yield request
+            yield request
+            yield request
+
+        response = await stub.example_stream_unary(request_iterator())
+        assert response.example_string == test_string
+        assert response.example_integer == test_int
+
+        # stream stream
+        async for response in stub.example_stream_stream(request_iterator()):
+            assert response.example_string == test_string
+            assert response.example_integer == test_int
--- a/tests/inputs/float/float.json
+++ b/tests/inputs/float/float.json
@@ -0,0 +1,9 @@
+{
+    "positive": "Infinity",
+    "negative": "-Infinity",
+    "nan": "NaN",
+    "three": 3.0,
+    "threePointOneFour": 3.14,
+    "negThree": -3.0,
+    "negThreePointOneFour": -3.14
+  }
--- a/tests/inputs/float/float.proto
+++ b/tests/inputs/float/float.proto
@@ -0,0 +1,12 @@
+syntax = "proto3";
+
+// Some documentation about the Test message.
+message Test {
+    double positive = 1;
+    double negative = 2;
+    double nan = 3;
+    double three = 4;
+    double three_point_one_four = 5;
+    double neg_three = 6;
+    double neg_three_point_one_four = 7;
+}
--- a/tests/inputs/oneof/oneof-name.json
+++ b/tests/inputs/oneof/oneof-name.json
@@ -1,3 +1,3 @@
 {
-  "name": "foobar"
+  "pitier": "Mr. T"
 }
--- a/tests/inputs/oneof/oneof.json
+++ b/tests/inputs/oneof/oneof.json
@@ -1,3 +1,3 @@
 {
-  "count": 100
+  "pitied": 100
 }
--- a/tests/inputs/oneof/oneof.proto
+++ b/tests/inputs/oneof/oneof.proto
@@ -2,7 +2,15 @@ syntax = "proto3";

 message Test {
  oneof foo {
-    int32 count = 1;
-    string name = 2;
+    int32 pitied = 1;
+    string pitier = 2;
+  }
+
+  int32 just_a_regular_field = 3;
+
+  oneof bar {
+    int32 drinks = 11;
+    string bar_name = 12;
  }
 }
+
--- a/tests/inputs/oneof/oneof_name.json
+++ b/tests/inputs/oneof/oneof_name.json
@@ -0,0 +1,3 @@
+{
+  "pitier": "Mr. T"
+}
--- a/tests/inputs/oneof/test_oneof.py
+++ b/tests/inputs/oneof/test_oneof.py
@@ -5,11 +5,11 @@ from tests.util import get_test_case_json_data

 def test_which_count():
    message = Test()
-    message.from_json(get_test_case_json_data("oneof"))
-    assert betterproto.which_one_of(message, "foo") == ("count", 100)
+    message.from_json(get_test_case_json_data("oneof")[0].json)
+    assert betterproto.which_one_of(message, "foo") == ("pitied", 100)


 def test_which_name():
    message = Test()
-    message.from_json(get_test_case_json_data("oneof", "oneof-name.json"))
-    assert betterproto.which_one_of(message, "foo") == ("name", "foobar")
+    message.from_json(get_test_case_json_data("oneof", "oneof_name.json")[0].json)
+    assert betterproto.which_one_of(message, "foo") == ("pitier", "Mr. T")
--- a/tests/inputs/oneof_empty/oneof_empty.json
+++ b/tests/inputs/oneof_empty/oneof_empty.json
@@ -0,0 +1,3 @@
+{
+  "nothing": {}
+}
--- a/tests/inputs/oneof_empty/oneof_empty.proto
+++ b/tests/inputs/oneof_empty/oneof_empty.proto
@@ -0,0 +1,15 @@
+syntax = "proto3";
+
+message Nothing {}
+
+message MaybeNothing {
+  string sometimes = 42;
+}
+
+message Test {
+  oneof empty {
+    Nothing nothing = 1;
+    MaybeNothing maybe1 = 2;
+    MaybeNothing maybe2 = 3;
+  }
+}
--- a/tests/inputs/oneof_empty/oneof_empty_maybe1.json
+++ b/tests/inputs/oneof_empty/oneof_empty_maybe1.json
@@ -0,0 +1,3 @@
+{
+  "maybe1": {}
+}
--- a/tests/inputs/oneof_empty/oneof_empty_maybe2.json
+++ b/tests/inputs/oneof_empty/oneof_empty_maybe2.json
@@ -0,0 +1,5 @@
+{
+  "maybe2": {
+    "sometimes": "now"
+  }
+}
--- a/tests/inputs/oneof_empty/test_oneof_empty.py
+++ b/tests/inputs/oneof_empty/test_oneof_empty.py
--- a/tests/inputs/oneof_enum/test_oneof_enum.py
+++ b/tests/inputs/oneof_enum/test_oneof_enum.py
@@ -14,7 +14,9 @@ def test_which_one_of_returns_enum_with_default_value():
    returns first field when it is enum and set with default value
    """
    message = Test()
-    message.from_json(get_test_case_json_data("oneof_enum", "oneof_enum-enum-0.json"))
+    message.from_json(
+        get_test_case_json_data("oneof_enum", "oneof_enum-enum-0.json")[0].json
+    )

    assert message.move == Move(
        x=0, y=0
@@ -28,7 +30,9 @@ def test_which_one_of_returns_enum_with_non_default_value():
    returns first field when it is enum and set with non default value
    """
    message = Test()
-    message.from_json(get_test_case_json_data("oneof_enum", "oneof_enum-enum-1.json"))
+    message.from_json(
+        get_test_case_json_data("oneof_enum", "oneof_enum-enum-1.json")[0].json
+    )
    assert message.move == Move(
        x=0, y=0
    )  # Proto3 will default this as there is no null
@@ -38,7 +42,7 @@ def test_which_one_of_returns_enum_with_non_default_value():

 def test_which_one_of_returns_second_field_when_set():
    message = Test()
-    message.from_json(get_test_case_json_data("oneof_enum"))
+    message.from_json(get_test_case_json_data("oneof_enum")[0].json)
    assert message.move == Move(x=2, y=3)
    assert message.signal == Signal.PASS
    assert betterproto.which_one_of(message, "action") == ("move", Move(x=2, y=3))
--- a/tests/inputs/repeated_duration_timestamp/repeated_duration_timestamp.json
+++ b/tests/inputs/repeated_duration_timestamp/repeated_duration_timestamp.json
@@ -0,0 +1,4 @@
+{
+    "times": ["1972-01-01T10:00:20.021Z", "1972-01-01T10:00:20.021Z"],
+    "durations": ["1.200s", "1.200s"]
+}
--- a/tests/inputs/repeated_duration_timestamp/repeated_duration_timestamp.proto
+++ b/tests/inputs/repeated_duration_timestamp/repeated_duration_timestamp.proto
@@ -0,0 +1,10 @@
+syntax = "proto3";
+
+import "google/protobuf/duration.proto";
+import "google/protobuf/timestamp.proto";
+
+
+message Test {
+    repeated google.protobuf.Timestamp times = 1;
+    repeated google.protobuf.Duration durations = 2;
+}
--- a/tests/inputs/repeated_duration_timestamp/test_repeated_duration_timestamp.py
+++ b/tests/inputs/repeated_duration_timestamp/test_repeated_duration_timestamp.py
@@ -0,0 +1,9 @@
+from datetime import datetime, timedelta
+
+from tests.output_betterproto.repeated_duration_timestamp import Test
+
+
+def test_roundtrip():
+    message = Test()
+    message.times = [datetime.now(), datetime.now()]
+    message.durations = [timedelta(), timedelta()]
--- a/tests/inputs/service_separate_packages/messages.proto
+++ b/tests/inputs/service_separate_packages/messages.proto
@@ -0,0 +1,31 @@
+syntax = "proto3";
+
+import "google/protobuf/duration.proto";
+import "google/protobuf/timestamp.proto";
+
+package things.messages;
+
+message DoThingRequest {
+  string name = 1;
+
+  // use `repeated` so we can check if `List` is correctly imported
+  repeated string comments = 2;
+
+  // use google types `timestamp` and `duration` so we can check
+  // if everything from `datetime` is correctly imported
+  google.protobuf.Timestamp when = 3;
+  google.protobuf.Duration duration = 4;
+}
+
+message DoThingResponse {
+  repeated string names = 1;
+}
+
+message GetThingRequest {
+  string name = 1;
+}
+
+message GetThingResponse {
+  string name = 1;
+  int32 version = 2;
+}
--- a/tests/inputs/service_separate_packages/service.proto
+++ b/tests/inputs/service_separate_packages/service.proto
@@ -0,0 +1,12 @@
+syntax = "proto3";
+
+import "messages.proto";
+
+package things.service;
+
+service Test {
+  rpc DoThing (things.messages.DoThingRequest) returns (things.messages.DoThingResponse);
+  rpc DoManyThings (stream things.messages.DoThingRequest) returns (things.messages.DoThingResponse);
+  rpc GetThingVersions (things.messages.GetThingRequest) returns (stream things.messages.GetThingResponse);
+  rpc GetDifferentThings (stream things.messages.GetThingRequest) returns (stream things.messages.GetThingResponse);
+}
--- a/tests/test_features.py
+++ b/tests/test_features.py
@@ -1,6 +1,7 @@
 import betterproto
 from dataclasses import dataclass
 from typing import Optional, List, Dict
+from datetime import datetime, timedelta


 def test_has_field():
@@ -285,17 +286,23 @@ def test_to_dict_default_values():


 def test_oneof_default_value_set_causes_writes_wire():
+    @dataclass
+    class Empty(betterproto.Message):
+        pass
+
    @dataclass
    class Foo(betterproto.Message):
        bar: int = betterproto.int32_field(1, group="group1")
        baz: str = betterproto.string_field(2, group="group1")
+        qux: Empty = betterproto.message_field(3, group="group1")

    def _round_trip_serialization(foo: Foo) -> Foo:
        return Foo().parse(bytes(foo))

    foo1 = Foo(bar=0)
    foo2 = Foo(baz="")
-    foo3 = Foo()
+    foo3 = Foo(qux=Empty())
+    foo4 = Foo()

    assert bytes(foo1) == b"\x08\x00"
    assert (
@@ -311,10 +318,17 @@ def test_oneof_default_value_set_causes_writes_wire():
        == ("baz", "")
    )

-    assert bytes(foo3) == b""
+    assert bytes(foo3) == b"\x1a\x00"
    assert (
        betterproto.which_one_of(foo3, "group1")
        == betterproto.which_one_of(_round_trip_serialization(foo3), "group1")
+        == ("qux", Empty())
+    )
+
+    assert bytes(foo4) == b""
+    assert (
+        betterproto.which_one_of(foo4, "group1")
+        == betterproto.which_one_of(_round_trip_serialization(foo4), "group1")
        == ("", None)
    )

@@ -395,3 +409,70 @@ def test_bool():
    assert t
    t.bar = 0
    assert not t
+
+
+# valid ISO datetimes according to https://www.myintervals.com/blog/2009/05/20/iso-8601-date-validation-that-doesnt-suck/
+iso_candidates = """2009-12-12T12:34
+2009
+2009-05-19
+2009-05-19
+20090519
+2009123
+2009-05
+2009-123
+2009-222
+2009-001
+2009-W01-1
+2009-W51-1
+2009-W33
+2009W511
+2009-05-19
+2009-05-19 00:00
+2009-05-19 14
+2009-05-19 14:31
+2009-05-19 14:39:22
+2009-05-19T14:39Z
+2009-W21-2
+2009-W21-2T01:22
+2009-139
+2009-05-19 14:39:22-06:00
+2009-05-19 14:39:22+0600
+2009-05-19 14:39:22-01
+20090621T0545Z
+2007-04-06T00:00
+2007-04-05T24:00
+2010-02-18T16:23:48.5
+2010-02-18T16:23:48,444
+2010-02-18T16:23:48,3-06:00
+2010-02-18T16:23:00.4
+2010-02-18T16:23:00,25
+2010-02-18T16:23:00.33+0600
+2010-02-18T16:00:00.23334444
+2010-02-18T16:00:00,2283
+2009-05-19 143922
+2009-05-19 1439""".split(
+    "\n"
+)
+
+
+def test_iso_datetime():
+    @dataclass
+    class Envelope(betterproto.Message):
+        ts: datetime = betterproto.message_field(1)
+
+    msg = Envelope()
+
+    for _, candidate in enumerate(iso_candidates):
+        msg.from_dict({"ts": candidate})
+        assert isinstance(msg.ts, datetime)
+
+
+def test_iso_datetime_list():
+    @dataclass
+    class Envelope(betterproto.Message):
+        timestamps: List[datetime] = betterproto.message_field(1)
+
+    msg = Envelope()
+
+    msg.from_dict({"timestamps": iso_candidates})
+    assert all([isinstance(item, datetime) for item in msg.timestamps])
--- a/tests/test_inputs.py
+++ b/tests/test_inputs.py
@@ -1,10 +1,11 @@
 import importlib
 import json
+import math
 import os
 import sys
 from collections import namedtuple
 from types import ModuleType
-from typing import Set
+from typing import Any, Dict, List, Set, Tuple

 import pytest

@@ -28,7 +29,12 @@ from google.protobuf.json_format import Parse


 class TestCases:
-    def __init__(self, path, services: Set[str], xfail: Set[str]):
+    def __init__(
+        self,
+        path,
+        services: Set[str],
+        xfail: Set[str],
+    ):
        _all = set(get_directories(path)) - {"__pycache__"}
        _services = services
        _messages = (_all - services) - {"__pycache__"}
@@ -69,6 +75,55 @@ def module_has_entry_point(module: ModuleType):
    return any(hasattr(module, attr) for attr in ["Test", "TestStub"])


+def list_replace_nans(items: List) -> List[Any]:
+    """Replace float("nan") in a list with the string "NaN"
+
+    Parameters
+    ----------
+    items : List
+            List to update
+
+    Returns
+    -------
+    List[Any]
+        Updated list
+    """
+    result = []
+    for item in items:
+        if isinstance(item, list):
+            result.append(list_replace_nans(item))
+        elif isinstance(item, dict):
+            result.append(dict_replace_nans(item))
+        elif isinstance(item, float) and math.isnan(item):
+            result.append(betterproto.NAN)
+    return result
+
+
+def dict_replace_nans(input_dict: Dict[Any, Any]) -> Dict[Any, Any]:
+    """Replace float("nan") in a dictionary with the string "NaN"
+
+    Parameters
+    ----------
+    input_dict : Dict[Any, Any]
+            Dictionary to update
+
+    Returns
+    -------
+    Dict[Any, Any]
+        Updated dictionary
+    """
+    result = {}
+    for key, value in input_dict.items():
+        if isinstance(value, dict):
+            value = dict_replace_nans(value)
+        elif isinstance(value, list):
+            value = list_replace_nans(value)
+        elif isinstance(value, float) and math.isnan(value):
+            value = betterproto.NAN
+        result[key] = value
+    return result
+
+
@pytest.fixture
 def test_data(request):
    test_case_name = request.param
@@ -81,7 +136,6 @@ def test_data(request):
    reference_module_root = os.path.join(
        *reference_output_package.split("."), test_case_name
    )
-
    sys.path.append(reference_module_root)

    plugin_module = importlib.import_module(f"{plugin_output_package}.{test_case_name}")
@@ -126,42 +180,48 @@ def test_message_json(repeat, test_data: TestData) -> None:
    plugin_module, _, json_data = test_data

    for _ in range(repeat):
-        message: betterproto.Message = plugin_module.Test()
+        for sample in json_data:
+            if sample.belongs_to(test_input_config.non_symmetrical_json):
+                continue

-        message.from_json(json_data)
-        message_json = message.to_json(0)
+            message: betterproto.Message = plugin_module.Test()

-        assert json.loads(message_json) == json.loads(json_data)
+            message.from_json(sample.json)
+            message_json = message.to_json(0)
+
+            assert dict_replace_nans(json.loads(message_json)) == dict_replace_nans(
+                json.loads(sample.json)
+            )


@pytest.mark.parametrize("test_data", test_cases.services, indirect=True)
 def test_service_can_be_instantiated(test_data: TestData) -> None:
-    plugin_module, _, json_data = test_data
-    plugin_module.TestStub(MockChannel())
+    test_data.plugin_module.TestStub(MockChannel())


@pytest.mark.parametrize("test_data", test_cases.messages_with_json, indirect=True)
 def test_binary_compatibility(repeat, test_data: TestData) -> None:
    plugin_module, reference_module, json_data = test_data

-    reference_instance = Parse(json_data, reference_module().Test())
-    reference_binary_output = reference_instance.SerializeToString()
+    for sample in json_data:
+        reference_instance = Parse(sample.json, reference_module().Test())
+        reference_binary_output = reference_instance.SerializeToString()

-    for _ in range(repeat):
-        plugin_instance_from_json: betterproto.Message = plugin_module.Test().from_json(
-            json_data
-        )
-        plugin_instance_from_binary = plugin_module.Test.FromString(
-            reference_binary_output
-        )
+        for _ in range(repeat):
+            plugin_instance_from_json: betterproto.Message = (
+                plugin_module.Test().from_json(sample.json)
+            )
+            plugin_instance_from_binary = plugin_module.Test.FromString(
+                reference_binary_output
+            )

-        # # Generally this can't be relied on, but here we are aiming to match the
-        # # existing Python implementation and aren't doing anything tricky.
-        # # https://developers.google.com/protocol-buffers/docs/encoding#implications
-        assert bytes(plugin_instance_from_json) == reference_binary_output
-        assert bytes(plugin_instance_from_binary) == reference_binary_output
+            # Generally this can't be relied on, but here we are aiming to match the
+            # existing Python implementation and aren't doing anything tricky.
+            # https://developers.google.com/protocol-buffers/docs/encoding#implications
+            assert bytes(plugin_instance_from_json) == reference_binary_output
+            assert bytes(plugin_instance_from_binary) == reference_binary_output

-        assert plugin_instance_from_json == plugin_instance_from_binary
-        assert (
-            plugin_instance_from_json.to_dict() == plugin_instance_from_binary.to_dict()
-        )
+            assert plugin_instance_from_json == plugin_instance_from_binary
+            assert dict_replace_nans(
+                plugin_instance_from_json.to_dict()
+            ) == dict_replace_nans(plugin_instance_from_binary.to_dict())
--- a/tests/test_version.py
+++ b/tests/test_version.py
@@ -0,0 +1,13 @@
+from betterproto import __version__
+from pathlib import Path
+import tomlkit
+
+PROJECT_TOML = Path(__file__).joinpath("..", "..", "pyproject.toml").resolve()
+
+
+def test_version():
+    with PROJECT_TOML.open() as toml_file:
+        project_config = tomlkit.loads(toml_file.read())
+    assert (
+        __version__ == project_config["tool"]["poetry"]["version"]
+    ), "Project version should match in package and package config"
--- a/tests/util.py
+++ b/tests/util.py
@@ -1,11 +1,11 @@
 import asyncio
+from dataclasses import dataclass
 import importlib
 import os
-import pathlib
-import sys
 from pathlib import Path
+import sys
 from types import ModuleType
-from typing import Callable, Generator, Optional, Union
+from typing import Callable, Dict, Generator, List, Optional, Tuple, Union

 os.environ["PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION"] = "python"

@@ -47,15 +47,44 @@ async def protoc(
    return stdout, stderr, proc.returncode


-def get_test_case_json_data(test_case_name: str, json_file_name: Optional[str] = None):
-    test_data_file_name = json_file_name or f"{test_case_name}.json"
-    test_data_file_path = inputs_path.joinpath(test_case_name, test_data_file_name)
+@dataclass
+class TestCaseJsonFile:
+    json: str
+    test_name: str
+    file_name: str

-    if not test_data_file_path.exists():
-        return None
+    def belongs_to(self, non_symmetrical_json: Dict[str, Tuple[str, ...]]):
+        return self.file_name in non_symmetrical_json.get(self.test_name, tuple())

-    with test_data_file_path.open("r") as fh:
-        return fh.read()
+
+def get_test_case_json_data(
+    test_case_name: str, *json_file_names: str
+) -> List[TestCaseJsonFile]:
+    """
+    :return:
+        A list of all files found in "{inputs_path}/test_case_name" with names matching
+        f"{test_case_name}.json" or f"{test_case_name}_*.json", OR given by
+        json_file_names
+    """
+    test_case_dir = inputs_path.joinpath(test_case_name)
+    possible_file_paths = [
+        *(test_case_dir.joinpath(json_file_name) for json_file_name in json_file_names),
+        test_case_dir.joinpath(f"{test_case_name}.json"),
+        *test_case_dir.glob(f"{test_case_name}_*.json"),
+    ]
+
+    result = []
+    for test_data_file_path in possible_file_paths:
+        if not test_data_file_path.exists():
+            continue
+        with test_data_file_path.open("r") as fh:
+            result.append(
+                TestCaseJsonFile(
+                    fh.read(), test_case_name, test_data_file_path.name.split(".")[0]
+                )
+            )
+
+    return result


 def find_module(
@@ -74,7 +103,7 @@ def find_module(
    if predicate(module):
        return module

-    module_path = pathlib.Path(*module.__path__)
+    module_path = Path(*module.__path__)

    for sub in [sub.parent for sub in module_path.glob("**/__init__.py")]:
        if sub == module_path:
--- a/yarn.lock
+++ b/yarn.lock
Author	SHA1	Message	Date
kalzoo	ee4265492d	docs: Add semantic release information and issue template	2021-11-16 22:42:31 -08:00
kalzoo	32faddf322	chore: Add Semantic Release GitHub action	2021-11-16 22:42:03 -08:00
kalzoo	89d1f47fac	Install Semantic Release	2021-11-16 22:30:16 -08:00
lazytype	c424b6f8db	Include AsyncIterator import for both clients and servers (#264 ) Co-authored-by: Robin Lambertz <github@roblab.la>	2021-11-05 14:22:15 +00:00
James Hilton-Balfe	421fdba309	Allow parsing of messages from ByteStrings #266	2021-10-26 00:34:33 +01:00
Robin Lambertz	fb2793e0b6	Allow parsing messages from byteslike Byteslike objects (like memoryview) do not have a decode function defined. Instead, a string may be created from them by passing them to the str constructor along with an encoding.	2021-08-25 12:53:02 +02:00
PIGNOSE	ad8b91766a	Add benchmarking cases for nested, repeat and deserialize (#241 )	2021-06-21 23:38:22 +02:00
Bekhzod Tillakhanov	a33126544b	Fix readme docs 'Async gRPC Support' (#249 )	2021-06-21 23:29:59 +02:00
nat	02e41afd09	Release v2.0.0b3 (#182 ) Updated change log to include all new features and fixes from master.	2021-04-07 12:57:44 +02:00
nat	7368299a70	Fix serialization of repeated fields with empty messages (#180 ) Extend test config and utils to support exclusion of certain json samples from testing for symetry.	2021-04-06 10:50:45 +10:00
nat	deb623ed14	Fix compilation of fields named 'bytes' or 'str' (#226 ) * if you have a field named "bytes" using the bytes type, it doesn't work. * Enable existing use-case & generalize solution to cover it Co-authored-by: Spencer <spencer@sf-n.com>	2021-04-06 10:45:57 +10:00
nat	95339bf74d	Misc cleanup, see commit body (#227 ) - Enable oneof_enum test case that passes now (removed the xfail) - Switch from toml to tomlkit as a dev dep for better toml support - upgrade poethepoet to latest stable release - use full table format for poe tasks to avoid long lines in pyproject.toml - remove redundant _WrappedMessage class - fix various Mypy warnings - reformat some comments for consistent line length	2021-04-06 10:43:09 +10:00
nat	5b639c82b2	Micro-optimization: use tuples instead of lists for conditions (#228 ) This should give a small speed boost to some critical code paths.	2021-04-06 10:40:45 +10:00
Matthew Badger	7c5ee47e68	Added support for infinite and nan floats/doubles (#215 ) - Added support for the custom double values from the protobuf json spec: "Infinity", "-Infinity", and "NaN" - Added `infinite_floats` test data - Updated Message.__eq__ to consider nan values equal - Updated `test_message_json` and `test_binary_compatibility` to replace NaN float values in dictionaries before comparison (because two NaN values are not equal)	2021-04-02 15:15:28 +02:00
Nat Noordanus	bb646fe26f	Fix template bug resulting in empty __post_init__ methods	2021-04-02 10:13:08 +11:00
Nat Noordanus	fc90653ab1	Sort the list of sources in generated file headers	2021-04-02 10:13:00 +11:00
Nat Noordanus	2a73dbac98	Make plugin use betterproto generated classes internally This means the betterproto plugin no longer needs to depend durectly on protobuf. This requires a small runtime hack to monkey patch some google types to get around the fact that the compiler uses proto2, but betterproto expects proto3. Also: - regenerate google.protobuf package - fix a regex bug in the logic for determining whether to use a google wrapper type. - fix a bug causing comments to get mixed up when multiple proto files generate code into a single python module	2021-04-02 10:13:00 +11:00
nat	891c9e5d6c	Update readme to avoid confusion about unreleased features. (#223 )	2021-04-01 20:40:02 +02:00
Nat Noordanus	a890514b5c	Update deps & add generate_lib task - Remove plugin dependency on protobuf since it's no longer required. - Update poethepoet to for better pyproject toml syntax support - Add handy generate_lib poe task for maintaining generated libs	2021-04-01 09:49:22 +11:00
Nat Noordanus	fe1e712fdb	Make plugin use betterproto generated classes internally This means the betterproto plugin no longer needs to depend durectly on protobuf. This requires a small runtime hack to monkey patch some google types to get around the fact that the compiler uses proto2, but betterproto expects proto3. Also: - regenerate google.protobuf package - fix a regex bug in the logic for determining whether to use a google wrapper type. - fix a bug causing comments to get mixed up when multiple proto files generate code into a single python module	2021-04-01 09:49:22 +11:00
Vasili Syrakis	7a358a63cf	Add __version__ attribute to package	2021-03-31 11:44:32 +11:00
nat	342e6559dc	Properly serialize zero-value messages in a oneof group (#176 ) Also improve test utils to make it easier to have multiple json examples. Co-authored-by: Christopher Chambers <chris@peanutcode.com>	2021-03-15 13:52:35 +01:00
Vladimir Solomatin	2f62189346	Fix `typing` and `datetime` imports not being present for service method type annotations (#183 )	2021-03-12 22:15:15 +01:00
MinJune Kim	8a215367ad	Allow empty services (#222 ) Fixes issue #220	2021-03-12 21:49:58 +01:00
robinaly	6c1c41e9cc	Use dateutil parser (#213 ) Switch to using `isoparse` from `dateutil.parser` instead of `datetime.fromisoformat` for more robust parsing of dates in from_dict.	2021-02-24 22:18:05 +01:00
Matthew Badger	9e6881999e	Add support for repeated timestamps and durations to to_dict from_dict (#211 )	2021-02-16 19:54:50 +01:00
nat	59f5f88c0d	Rebuild poetry.lock to fix CI (#202 )	2021-01-25 20:28:30 +01:00
Tim Schmidt	8eea5fe256	added documentation for server-facing stubs (#186 )	2021-01-24 22:20:32 +01:00
Tim Schmidt	1d54ef8f99	Generate grpclib service stubs (#170 )	2020-12-04 22:22:11 +01:00
nat	73cea12e1f	Fix incorrect routes in generated client when service is not in a package (#177 )	2020-11-28 17:50:25 +01:00