chore: basedpyright -> ty, + flake.nix (#85)

Author: egorchakov

.gitignore

@@ -20,8 +20,8 @@ config/*
 # lockfiles
 uv.lock
 
-.basedpyright
-
 .ipynb_checkpoints
 
-.dprint.json
+# env
+.direnv
+.envrc

.pre-commit-config.yaml

@@ -11,19 +11,24 @@ repos:
       - id: pyupgrade
 
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.12.7
+    rev: v0.13.0
     hooks:
       - id: ruff-check
         args: [--fix]
       - id: ruff-format
 
-  - repo: https://github.com/DetachHead/basedpyright-pre-commit-mirror
-    rev: 1.31.1
+  - repo: local
+    # TODO: replace once available
     hooks:
-      - id: basedpyright
+      - id: ty
+        name: ty check
+        entry: uvx ty@latest check
+        language: system
+        always_run: true
+        pass_filenames: false
 
   - repo: https://github.com/tombi-toml/tombi-pre-commit
-    rev: v0.4.42
+    rev: v0.6.4
     hooks:
       - id: tombi-format
       - id: tombi-lint

flake.lock

@@ -0,0 +1,32 @@
+{
+  inputs.nixpkgs.url = "github:nixos/nixpkgs/nixos-unstable";
+  inputs.flake-utils.url = "github:numtide/flake-utils";
+
+  outputs = {
+    nixpkgs,
+    flake-utils,
+    ...
+  }:
+    flake-utils.lib.eachDefaultSystem (
+      system: let
+        pkgs = nixpkgs.legacyPackages.${system};
+      in {
+        devShells.default = with pkgs;
+          mkShell {
+            packages =
+              [
+                git
+                git-lfs
+                uv
+                ytt
+                just
+              ]
+              ++ lib.optional stdenv.isDarwin [ffmpeg];
+
+            shellHook = lib.optionalString stdenv.isDarwin ''
+              export DYLD_FALLBACK_LIBRARY_PATH=${pkgs.lib.makeLibraryPath [pkgs.ffmpeg]}
+            '';
+          };
+      }
+    );
+}

flake.nix

@@ -14,8 +14,8 @@
 
 
 @final
-class BuildYaakIdlProtosHook(BuildHookInterface):  # pyright: ignore[reportMissingTypeArgument]
-    PLUGIN_NAME = "build-yaak-idl-protos"  # pyright: ignore[reportIncompatibleUnannotatedOverride]
+class BuildYaakIdlProtosHook(BuildHookInterface):
+    PLUGIN_NAME = "build-yaak-idl-protos"
 
     YAAK_IDL_PROTO_PATH = (
         Path(__file__).resolve().parent
@@ -63,7 +63,7 @@ def _build_yaak_idl_protos(cls) -> None:
                 *cls.YAAK_IDL_PROTOS,
             ]
 
-            if protoc.main(protoc_cmd) != 0:  # pyright: ignore[reportUnknownMemberType]
+            if protoc.main(protoc_cmd) != 0:
                 msg = f"error: {protoc_cmd} failed"
                 raise RuntimeError(msg)
 

hatch_build.py

@@ -9,18 +9,13 @@ _default:
 sync:
     uv sync --all-extras --all-groups
 
-install-tools:
-    uv tool install --force --upgrade ruff
-    uv tool install --force --upgrade basedpyright
-    uv tool install --force --upgrade pre-commit --with pre-commit-uv
-
 install-duckdb-extensions:
     uv run python -c "import duckdb; duckdb.connect().install_extension('spatial')"
 
-setup: sync install-tools install-duckdb-extensions
+setup: sync install-duckdb-extensions
     git submodule update --init --recursive --force --remote
     git lfs pull
-    uvx pre-commit install --install-hooks
+    uvx --with=pre-commit-uv pre-commit install --install-hooks
 
 build:
     uv build
@@ -32,10 +27,10 @@ lint *ARGS:
     uvx ruff check {{ ARGS }}
 
 typecheck *ARGS:
-    uvx basedpyright {{ ARGS }}
+    uvx ty@latest check {{ ARGS }}
 
 pre-commit *ARGS: build
-    uvx pre-commit run --all-files --color=always {{ ARGS }}
+    uvx --with=pre-commit-uv pre-commit run --all-files --color=always {{ ARGS }}
 
 generate-config:
     ytt --ignore-unknown-comments \

justfile

@@ -74,18 +74,6 @@ test = [
 requires = ["hatchling>=1.27.0"]
 build-backend = "hatchling.build"
 
-[tool.basedpyright]
-typeCheckingMode = "all"
-enableTypeIgnoreComments = true
-reportMissingTypeStubs = "none"
-reportAny = "none"
-reportExplicitAny = "none"
-reportIgnoreCommentWithoutRule = "error"
-venvPath = "."
-venv = ".venv"
-include = ["src/"]
-ignore = ["src/**/*_pb2.py", "**/*.pyi"]
-
 [tool.hatch.metadata]
 allow-direct-references = true
 
@@ -132,7 +120,6 @@ ignore = [
   "CPY",
   "COM812",
   "F722",
-  "PD901",
   "ISC001",
   "TD",
   "TC006",

pyproject.toml

@@ -4,13 +4,13 @@
 from torch import Tensor
 
 
-class BatchMeta(TensorClass, autocast=True):  # pyright: ignore[reportGeneralTypeIssues, reportCallIssue]
+class BatchMeta(TensorClass, autocast=True):
     sample_idx: Tensor | None = None
     input_id: NonTensorData | None = None
 
 
-class Batch(TensorClass, autocast=True):  # pyright: ignore[reportGeneralTypeIssues, reportCallIssue]
-    data: TensorDict | None = None  # pyright: ignore[reportIncompatibleMethodOverride]
+class Batch(TensorClass, autocast=True):
+    data: TensorDict | None = None
     meta: BatchMeta | None = None
 
 

src/rbyte/batch.py

@@ -39,7 +39,7 @@
 
 
 class SourceConfig(BaseModel):
-    source: HydraConfig[TensorSource]  # pyright: ignore[reportMissingTypeArgument]
+    source: HydraConfig[TensorSource]
     index_column: str
 
 
@@ -83,7 +83,7 @@ class BasePipelineConfig(BaseModel):
     def _validate_inputs(
         cls, value: Sequence[dict[str, Any]]
     ) -> Sequence[dict[str, Any]]:
-        if not mit.all_equal(map(tree_structure, value)):  # pyright: ignore[reportArgumentType]
+        if not mit.all_equal(map(tree_structure, value)):  # ty: ignore[invalid-argument-type]
             msg = "inputs have different structures"
             raise ValueError(msg)
 
@@ -171,7 +171,7 @@ def get_batch(
         self,
         index: int | Sequence[int] | slice | range,
         *,
-        keys: BatchKeys = BATCH_KEYS_DEFAULT,
+        keys: BatchKeys = BATCH_KEYS_DEFAULT,  # ty: ignore[invalid-parameter-default]
     ) -> Batch:
         subkeys: dict[Literal["data", "meta"], set[_ALL_TYPE | str]] = {
             "data": set(),
@@ -217,7 +217,7 @@ def get_batch(
 
                 source_data = {
                     row[Column.source_id]: torch.stack([
-                        self._get_source(source)[idxs]  # pyright: ignore[reportUnknownMemberType]
+                        self._get_source(source)[idxs]
                         for (source, idxs) in zip(
                             row[Column.source_config],
                             row[Column.source_idxs],
@@ -232,7 +232,7 @@ def get_batch(
             sample_data_cols = (
                 pl.all()
                 if _ALL in subkeys_data
-                else pl.col(subkeys_data - source_data.keys())  # pyright: ignore[reportArgumentType]
+                else pl.col(subkeys_data - source_data.keys())
             ).exclude(Column.sample_idx, Column.input_id)
 
             samples_subset = samples.select(sample_data_cols.to_physical())
@@ -242,7 +242,7 @@ def get_batch(
             except TypeError:
                 sample_data = samples_subset.to_dict(as_series=False)
 
-            data = TensorDict(source_data | sample_data, batch_size=batch_size)  # pyright: ignore[reportArgumentType]
+            data = TensorDict(source_data | sample_data, batch_size=batch_size)
 
         else:
             data = None
@@ -277,8 +277,8 @@ def __len__(self) -> int:
         return len(self.samples)
 
     @cache  # noqa: B019
-    def _get_source(self, config: str) -> TensorSource:  # pyright: ignore[reportUnknownParameterType, reportMissingTypeArgument] # noqa: PLR6301
-        return HydraConfig[TensorSource].model_validate_json(config).instantiate()  # pyright: ignore[reportUnknownVariableType, reportMissingTypeArgument]
+    def _get_source(self, config: str) -> TensorSource:  # noqa: PLR6301
+        return HydraConfig[TensorSource].model_validate_json(config).instantiate()
 
     @classmethod
     def _build_samples(
@@ -293,33 +293,33 @@ def _build_samples(
 
             case PipelineHydraConfig():
                 pipeline = samples.pipeline.instantiate()
-                executor: Executor | dict[OUTPUT_TYPE, Executor] | None = tree_map(  # pyright: ignore[reportAssignmentType]
+                executor: Executor | dict[OUTPUT_TYPE, Executor] | None = tree_map(  # ty: ignore[invalid-assignment]
                     HydraConfig[Executor].instantiate,
-                    samples.executor,  # pyright: ignore[reportArgumentType]
+                    samples.executor,  # ty: ignore[invalid-argument-type]
                 )
 
         pipeline.print_documentation()
-        inputs = tree_transpose(  # pyright: ignore[reportUnknownVariableType]
-            tree_structure(list(range(len(samples.inputs)))),  # pyright: ignore[reportArgumentType]
-            tree_structure(samples.inputs[0]),  # pyright: ignore[reportArgumentType]
-            samples.inputs,  # pyright: ignore[reportArgumentType]
+        inputs = tree_transpose(
+            tree_structure(list(range(len(samples.inputs)))),  # ty: ignore[invalid-argument-type]
+            tree_structure(samples.inputs[0]),  # ty: ignore[invalid-argument-type]
+            samples.inputs,  # ty: ignore[invalid-argument-type]
         )
 
         results = pipeline.map(
-            inputs=inputs,  # pyright: ignore[reportArgumentType]
+            inputs=inputs,  # ty: ignore[invalid-argument-type]
             executor=executor,
             **samples.model_dump(exclude={"pipeline", "inputs", "executor"}),
         )
 
         if pipeline.profile:
             pipeline.print_profiling_stats()
 
-        output_name: str = pipeline.unique_leaf_node.output_name  # pyright: ignore[reportUnknownMemberType, reportAssignmentType]
+        output_name: str = pipeline.unique_leaf_node.output_name
 
         result: pl.DataFrame = (
             results[output_name].output
             if results
-            else load_outputs(output_name, run_folder=samples.run_folder)  # pyright: ignore[reportArgumentType]
+            else load_outputs(output_name, run_folder=samples.run_folder)  # ty: ignore[invalid-argument-type]
         )
 
         return (
@@ -346,7 +346,7 @@ def _build_sources(cls, sources: SourcesConfig) -> pl.DataFrame:
                             source_cfg.model_dump(exclude={"source"})
                             | {
                                 "id": source_id,
-                                "config": source_cfg.source.model_dump_json(  # pyright: ignore[reportUnknownMemberType]
+                                "config": source_cfg.source.model_dump_json(
                                     by_alias=True
                                 ),
                             }
@@ -357,8 +357,8 @@ def _build_sources(cls, sources: SourcesConfig) -> pl.DataFrame:
                 ],
                 schema_overrides={Column.input_id: input_id_enum},
             )
-            .explode(k)
-            .unnest(k)
-            .select(Column.input_id, pl.exclude(Column.input_id).name.prefix(f"{k}."))
+            .explode(k)  # ty: ignore[unresolved-reference]
+            .unnest(k)  # ty: ignore[unresolved-reference]
+            .select(Column.input_id, pl.exclude(Column.input_id).name.prefix(f"{k}."))  # ty: ignore[unresolved-reference]
             .rechunk()
         )

src/rbyte/dataset.py

@@ -8,7 +8,7 @@
 
 @final
 class DuckDbDataFrameBuilder:
-    __name__ = __qualname__
+    __name__ = __qualname__  # ty: ignore[unresolved-reference]
 
     @validate_call
     def __call__(self, *, query: str, path: PathLike[str]) -> pl.DataFrame: