Replace basedpyright with ty for Python type checking

.github/workflows/ci.yml

@@ -85,13 +85,13 @@ jobs:
         run: uvx ruff format --check .
       - name: Python Lint - Ruff
         run: uvx ruff check .
-      # PyRight needs the project for type information, so use uv run
-      - name: Python Lint - PyRight
-        env:
-          MATURIN_PEP517_ARGS: "--profile ci"
+      # ty needs the Python dependencies for type information, but resolves the
+      # vortex package itself from the in-tree sources and stubs, so skip the
+      # Rust extension build entirely.
+      - name: Python Lint - ty
         run: |
-          uv sync --all-packages
-          uv run basedpyright vortex-python
+          uv sync --all-packages --no-install-package vortex-data --no-install-package vortex-data-cuda
+          uv run --no-sync ty check vortex-python
 
   python-test:
     name: "Python (test)"

.github/workflows/cuda.yaml

@@ -134,9 +134,9 @@ jobs:
         env:
           MATURIN_PEP517_ARGS: "--profile ci"
         run: |
-          # --all-packages installs the shared dev tooling (basedpyright) from the
+          # --all-packages installs the shared dev tooling (ty) from the
           # root `dev` group; --extra cuda adds the vortex-data-cuda extension.
-          uv run --all-packages --extra cuda basedpyright vortex-python vortex-python-cuda
+          uv run --all-packages --extra cuda ty check vortex-python vortex-python-cuda
           uv run --all-packages --extra cuda pytest --benchmark-disable vortex-python/test/test_cuda.py vortex-python-cuda/test
 
   cuda-test-sanitizer:

pyproject.toml

@@ -17,7 +17,7 @@ packages = ["dummy"] # Required for workspace project
 # Shared dev tooling. Member-specific dev deps live in each member's pyproject.toml.
 # `uv sync --all-packages` picks up dev groups from all workspace members.
 dev = [
-    "basedpyright>=1.31",
+    "ty>=0.0.73",
     "ipython>=8.26.0",
     "pip>=23.3.2",
     "pytest>=7.4.0",
@@ -73,5 +73,16 @@ log_cli = true
 log_cli_level = "INFO"
 xfail_strict = true
 
-[tool.basedpyright]
-exclude = ["vortex-python/python/vortex/_lib/store/**.pyi"]
+[tool.ty.environment]
+# First-party source roots so `vortex` and `vortex_cuda` resolve without an
+# editable install of the Rust extensions.
+root = [
+    "vortex-python/python",
+    "vortex-python",
+    "vortex-python-cuda/python",
+    "vortex-python-cuda",
+]
+
+[tool.ty.src]
+# Vendored obstore stubs are not checked.
+exclude = ["vortex-python/python/vortex/_lib/store/**/*.pyi"]

vortex-python-cuda/python/vortex_cuda/__init__.py

@@ -1,8 +1,8 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright the Vortex contributors
-# pyright: reportAttributeAccessIssue=false, reportMissingModuleSource=false, reportPrivateUsage=false, reportUnknownMemberType=false, reportUnknownVariableType=false
 
 import importlib
+from types import ModuleType
 
 from . import _lib
 
@@ -39,7 +39,7 @@ def _install_vortex_array_methods() -> None:
         setattr(vortex.Array, "__arrow_c_device_array__", _Array___arrow_c_device_array__)
 
 
-def _import_cudf_modules() -> tuple[object, object]:
+def _import_cudf_modules() -> tuple[ModuleType, ModuleType]:
     try:
         cudf = importlib.import_module("cudf")
         pylibcudf = importlib.import_module("pylibcudf")

vortex-python-cuda/test/test_cuda.py

@@ -1,6 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright the Vortex contributors
-# pyright: reportAny=false, reportExplicitAny=false
 
 import gc
 import sys

vortex-python-cuda/test/test_native_bridge.py

@@ -1,6 +1,5 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright the Vortex contributors
-# pyright: reportPrivateUsage=false
 
 import gc
 
@@ -66,7 +65,7 @@ def test_metadata_bridge_struct_with_children():
 
     arrow_table = pa.table({"a": [1, 2, 3], "b": [4.0, 5.0, 6.0]})
     struct_array = vortex.Array.from_arrow(
-        pa.StructArray.from_arrays(  # pyright: ignore[reportUnknownMemberType]
+        pa.StructArray.from_arrays(
             [arrow_table.column("a").combine_chunks(), arrow_table.column("b").combine_chunks()],
             names=["a", "b"],
         )
@@ -128,7 +127,7 @@ def test_arrow_device_export_struct_array():
 
     arrow_table = pa.table({"a": [1, 2, 3], "b": [4.0, 5.0, 6.0]})
     struct_array = vortex.Array.from_arrow(
-        pa.StructArray.from_arrays(  # pyright: ignore[reportUnknownMemberType]
+        pa.StructArray.from_arrays(
             [arrow_table.column("a").combine_chunks(), arrow_table.column("b").combine_chunks()],
             names=["a", "b"],
         )

vortex-python/AGENTS.md

@@ -25,7 +25,7 @@ If Python docstrings, `docs/api/python/`, or Sphinx configuration change, also f
 ## Linting and Formatting
 
 ```bash
-uv run basedpyright vortex-python
+uv run ty check vortex-python
 uv run ruff format --check <changed-python-files>
 uv run ruff check <changed-python-files>
 ```

vortex-python/benchmark/conftest.py

@@ -27,15 +27,15 @@ def vxf(
         length = 100_000
 
         columns: dict[str, list[int] | list[float] | list[str]] = {}
-        assert "x" in request.param  # pyright: ignore[reportAny]
+        assert "x" in request.param
         columns["x"] = list(range(length))
 
-        if "y" in request.param:  # pyright: ignore[reportAny]
+        if "y" in request.param:
             columns["y"] = [hashlib.md5(x.to_bytes(length=4), usedforsecurity=False).hexdigest() for x in range(length)]
-        if "z" in request.param:  # pyright: ignore[reportAny]
+        if "z" in request.param:
             columns["z"] = [math.sqrt(x) for x in range(length)]
 
-        a = vx.array(pa.table(columns))  # pyright: ignore[reportCallIssue, reportUnknownArgumentType, reportArgumentType]
+        a = vx.array(pa.table(columns))  # ty: ignore[no-matching-overload]
         vx.io.write(a, str(fname))
     return vx.open(str(fname))
 

vortex-python/benchmark/test_aggregation.py

@@ -7,7 +7,7 @@
 import pyarrow as pa
 import pytest
 from pyarrow.types import is_floating, is_integer
-from pytest_benchmark.fixture import BenchmarkFixture  # pyright: ignore[reportMissingTypeStubs]
+from pytest_benchmark.fixture import BenchmarkFixture
 
 import vortex as vx
 
@@ -19,9 +19,7 @@ def _has_mean(t: pa.DataType) -> bool:
 @pytest.mark.benchmark(group="aggregation", disable_gc=True)
 def test_arrow_table_aggregation(benchmark: BenchmarkFixture, vxf: vx.VortexFile):
     aggregations: list[tuple[str, Literal["mean"]]] = [
-        (field.name, "mean")
-        for field in vxf.dtype.to_arrow_schema()  # pyright: ignore[reportUnknownVariableType]
-        if _has_mean(field.type)  # pyright: ignore[reportUnknownMemberType, reportUnknownArgumentType]
+        (field.name, "mean") for field in vxf.dtype.to_arrow_schema() if _has_mean(field.type)
     ]
     benchmark(lambda: pa.concat_tables(x.to_arrow_table() for x in vxf.scan()).group_by([]).aggregate(aggregations))
 
@@ -44,7 +42,7 @@ def test_duckdb_aggregation(benchmark: BenchmarkFixture, vxf: vx.VortexFile):
     ds = vxf.to_dataset()
     _ = conn.register("ds", ds)
     aggregations = ",".join(
-        [f"avg(ds.{field.name}) as {field.name}" for field in vxf.dtype.to_arrow_schema() if _has_mean(field.type)]  # pyright: ignore[reportUnknownVariableType, reportUnknownMemberType, reportUnknownArgumentType]
+        [f"avg(ds.{field.name}) as {field.name}" for field in vxf.dtype.to_arrow_schema() if _has_mean(field.type)]
     )
     print(aggregations)
     query = f"select {aggregations} from ds"

vortex-python/benchmark/test_filter.py

@@ -5,7 +5,7 @@
 import polars as pl
 import pyarrow as pa
 import pytest
-from pytest_benchmark.fixture import BenchmarkFixture  # pyright: ignore[reportMissingTypeStubs]
+from pytest_benchmark.fixture import BenchmarkFixture
 
 import vortex as vx
 from vortex.expr import column

vortex-python/benchmark/test_scalar_at.py

@@ -4,7 +4,7 @@
 import duckdb
 import pyarrow as pa
 import pytest
-from pytest_benchmark.fixture import BenchmarkFixture  # pyright: ignore[reportMissingTypeStubs]
+from pytest_benchmark.fixture import BenchmarkFixture
 
 import vortex as vx
 

vortex-python/benchmark/test_scan.py

@@ -4,7 +4,7 @@
 import duckdb
 import pyarrow as pa
 import pytest
-from pytest_benchmark.fixture import BenchmarkFixture  # pyright: ignore[reportMissingTypeStubs]
+from pytest_benchmark.fixture import BenchmarkFixture
 
 import vortex as vx
 

vortex-python/benchmark/test_serialization.py

@@ -5,12 +5,12 @@
 from typing import cast
 
 import pytest
-from pytest_benchmark.fixture import BenchmarkFixture  # pyright: ignore[reportMissingTypeStubs]
+from pytest_benchmark.fixture import BenchmarkFixture
 
 import vortex as vx
 
 
-@pytest.mark.parametrize("protocol", [4, 5], ids=lambda p: f"p{p}")  # pyright: ignore[reportAny]
+@pytest.mark.parametrize("protocol", [4, 5], ids=lambda p: f"p{p}")
 @pytest.mark.parametrize("operation", ["dumps", "loads", "roundtrip"])
 @pytest.mark.benchmark(disable_gc=True)
 def test_pickle(

vortex-python/check.sh

@@ -17,7 +17,7 @@ pushd $ROOT/vortex-python
 maturin develop
 ruff format --check
 ruff check
-basedpyright
+ty check
 popd
 
 pushd $ROOT/docs

vortex-python/clickbench.py

@@ -16,7 +16,7 @@
 import vortex as vx
 
 # 0: No., 1: SQL, 2: Polars
-queries: list[tuple[str, str, Callable[[pl.LazyFrame], Any]]] = [  # pyright: ignore[reportExplicitAny]
+queries: list[tuple[str, str, Callable[[pl.LazyFrame], Any]]] = [
     ("Q0", "SELECT COUNT(*) FROM hits;", lambda x: x.select(pl.len()).collect().height),
     (
         "Q1",
@@ -35,17 +35,17 @@
     (
         "Q3",
         "SELECT AVG(UserID) FROM hits;",
-        lambda x: x.select(pl.col("UserID").mean()).collect().item(),  # pyright: ignore[reportAny]
+        lambda x: x.select(pl.col("UserID").mean()).collect().item(),
     ),
     (
         "Q4",
         "SELECT COUNT(DISTINCT UserID) FROM hits;",
-        lambda x: x.select(pl.col("UserID").n_unique()).collect().item(),  # pyright: ignore[reportAny]
+        lambda x: x.select(pl.col("UserID").n_unique()).collect().item(),
     ),
     (
         "Q5",
         "SELECT COUNT(DISTINCT SearchPhrase) FROM hits;",
-        lambda x: x.select(pl.col("SearchPhrase").n_unique()).collect().item(),  # pyright: ignore[reportAny]
+        lambda x: x.select(pl.col("SearchPhrase").n_unique()).collect().item(),
     ),
     (
         "Q6",
@@ -187,7 +187,7 @@
     (
         "Q20",
         "SELECT COUNT(*) FROM hits WHERE URL LIKE '%google%';",
-        lambda x: x.filter(pl.col("URL").str.contains("google")).select(pl.len()).collect().item(),  # pyright: ignore[reportAny]
+        lambda x: x.filter(pl.col("URL").str.contains("google")).select(pl.len()).collect().item(),
     ),
     (
         "Q21",
@@ -511,7 +511,7 @@
 ]
 
 
-def run_timings(lf: pl.LazyFrame, name: str, src: str, load_time: int | None) -> dict[str, Any] | None:  # pyright: ignore[reportExplicitAny]
+def run_timings(lf: pl.LazyFrame, name: str, src: str, load_time: int | None) -> dict[str, Any] | None:
     queries_times: list[list[float | None]] = []
     for q in queries:
         # if q[0] == "Q19":
@@ -523,7 +523,7 @@ def run_timings(lf: pl.LazyFrame, name: str, src: str, load_time: int | None) ->
         for _ in range(3):
             start = timeit.default_timer()
             try:
-                result = q[2](lf)  # pyright: ignore[reportAny]
+                result = q[2](lf)
             except Exception as e:
                 print("Failed", e)
                 result = None
@@ -564,22 +564,22 @@ def run_timings(lf: pl.LazyFrame, name: str, src: str, load_time: int | None) ->
 
 PARSER = argparse.ArgumentParser()
 
-PARSER.add_argument(  # pyright: ignore[reportUnusedCallResult]
+PARSER.add_argument(
     "--path",
     type=str,
     default="hits.parquet",
     help="Path to the parquet file",
 )
 
-PARSER.add_argument(  # pyright: ignore[reportUnusedCallResult]
+PARSER.add_argument(
     "--formats",
     nargs="+",
     choices=("vortex", "parquet"),
     default=None,
     help="Formats to run",
 )
 
-PARSER.add_argument(  # pyright: ignore[reportUnusedCallResult]
+PARSER.add_argument(
     "-q",
     "--queries",
     nargs="+",
@@ -588,36 +588,36 @@ def run_timings(lf: pl.LazyFrame, name: str, src: str, load_time: int | None) ->
     help="Queries to run",
 )
 
-PARSER.add_argument("-i", "--iterations", type=int, default=3, help="Number of iterations to run")  # pyright: ignore[reportUnusedCallResult]
+PARSER.add_argument("-i", "--iterations", type=int, default=3, help="Number of iterations to run")
 
 
 def main(args: argparse.Namespace):
-    assert isinstance(args.path, str)  # pyright: ignore[reportAny]
-    assert isinstance(args.queries, list)  # pyright: ignore[reportAny]
-    assert isinstance(args.formats, list | None)  # pyright: ignore[reportAny]
+    assert isinstance(args.path, str)
+    assert isinstance(args.queries, list)
+    assert isinstance(args.formats, list | None)
 
     if not os.path.exists(args.path):
         raise ValueError(f"File {args.path} does not exist")
 
     results: defaultdict[str, list[float]] = defaultdict(list)
 
     def run_queries(format: str, lf: pl.LazyFrame):
-        for q in args.queries:  # pyright: ignore[reportAny]
+        for q in args.queries:
             assert isinstance(q, int)
 
             timings = []
-            for _ in range(args.iterations):  # pyright: ignore[reportAny]
+            for _ in range(args.iterations):
                 start = timeit.default_timer()
                 try:
-                    _result: Callable[[pl.LazyFrame], Any] = queries[q][2](lf)  # pyright: ignore[reportExplicitAny, reportAny]
+                    _result: Callable[[pl.LazyFrame], Any] = queries[q][2](lf)
                 except Exception as e:
                     print(f"Failed Q{q}", e)
-                timings.append(timeit.default_timer() - start)  # pyright: ignore[reportUnknownMemberType]
-            average = sum(timings) / len(timings)  # pyright: ignore[reportUnknownArgumentType]
+                timings.append(timeit.default_timer() - start)
+            average = sum(timings) / len(timings)
             results[format].append(average)
             print(f"{format} Q{q}", average)
 
-    if args.formats is None or "vortex" in args.formats:  # pyright: ignore[reportUnknownMemberType]
+    if args.formats is None or "vortex" in args.formats:
         vx_base, _ = os.path.splitext(args.path)
         vx_path = f"{vx_base}.vortex"
 
@@ -628,7 +628,7 @@ def run_queries(format: str, lf: pl.LazyFrame):
 
             def _iter():
                 for i in range(pf.num_row_groups):
-                    arr = pf.read_row_group(i).to_struct_array()  # pyright: ignore[reportUnknownMemberType]
+                    arr = pf.read_row_group(i).to_struct_array()
                     arr = vx.Array.from_arrow(arr)
                     yield arr
 
@@ -638,7 +638,7 @@ def _iter():
         lf = vx.open(vx_path).to_polars()
         run_queries("vortex", lf)
 
-    if args.formats is None or "parquet" in args.formats:  # pyright: ignore[reportUnknownMemberType]
+    if args.formats is None or "parquet" in args.formats:
         lf = pl.scan_parquet(args.path)
         run_queries("parquet", lf)
 

vortex-python/pyproject.toml

@@ -91,5 +91,9 @@ dev = [
     "protobuf>=6.33.5",
 ]
 
-[tool.basedpyright]
-exclude = ["python/vortex/_lib/store/**.pyi"]
+[tool.ty.environment]
+root = ["python", "."]
+
+[tool.ty.src]
+# Vendored obstore stubs are not checked.
+exclude = ["python/vortex/_lib/store/**/*.pyi"]