chore: Modernize pre-commit hooks

.github/workflows/ci_test.yml

@@ -47,30 +47,13 @@ jobs:
 
   lint:
     needs: [prepare]
-    if: >
-      needs.prepare.outputs.should_skip_ci_commit != 'true' &&
-      needs.prepare.outputs.should_skip_ci_docs_only != 'true'
-    name: Lint
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-        with:
-          submodules: recursive
-      - uses: actions/setup-python@v6
-        with:
-          python-version: '3.13'
-      - name: Install dependencies
-        run: |
-          pip install black pylint ruff
-          sudo apt-get install -y clang-format-15
-
-      - name: Lint
-        run: |
-          tests/scripts/task_lint.sh
-
+      - uses: pre-commit/action@2c7b3805fd2a0fd8c1884dcaebf91fc102a13ecd # v3.0.1
 
   test:
-    needs: [prepare]
+    needs: [lint, prepare]
     if: >
       needs.prepare.outputs.should_skip_ci_commit != 'true' &&
       needs.prepare.outputs.should_skip_ci_docs_only != 'true'

.pre-commit-config.yaml

@@ -15,8 +15,29 @@
 # specific language governing permissions and limitations
 # under the License.
 
+# TODO(@junrushao): adding a few extra hooks:
+# - Python type checking via mypy or ty
+# - CMake linters
+# - Conventional commits
+default_install_hook_types:
+  - pre-commit
 repos:
-  # Standard hooks
+  - repo: local
+    hooks:
+      - id: check-asf-header
+        name: check ASF Header
+        entry: python tests/lint/check_asf_header.py --check
+        language: system
+        pass_filenames: false
+        verbose: false
+  - repo: local
+    hooks:
+      - id: check-file-type
+        name: check file types
+        entry: python tests/lint/check_file_type.py
+        language: system
+        pass_filenames: false
+        verbose: false
   - repo: https://github.com/pre-commit/pre-commit-hooks
     rev: v5.0.0
     hooks:
@@ -28,3 +49,31 @@ repos:
       - id: mixed-line-ending
       - id: requirements-txt-fixer
       - id: trailing-whitespace
+      - id: check-yaml
+      - id: check-toml
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.12.3
+    hooks:
+      - id: ruff-check
+        types_or: [python, pyi, jupyter]
+        args: [--fix]
+      - id: ruff-format
+        types_or: [python, pyi, jupyter]
+  - repo: https://github.com/pre-commit/mirrors-clang-format
+    rev: "v20.1.8"
+    hooks:
+      - id: clang-format
+  - repo: https://github.com/MarcoGorelli/cython-lint
+    rev: v0.16.7
+    hooks:
+      - id: cython-lint
+        args: [--max-line-length=120]
+      - id: double-quote-cython-strings
+  - repo: https://github.com/scop/pre-commit-shfmt
+    rev: v3.12.0-2
+    hooks:
+      - id: shfmt
+  - repo: https://github.com/shellcheck-py/shellcheck-py
+    rev: v0.10.0.1
+    hooks:
+      - id: shellcheck

README.md

@@ -18,5 +18,3 @@
 # tvm ffi
 
 [![CI](https://github.com/apache/tvm-ffi/actions/workflows/ci_test.yml/badge.svg)](https://github.com/apache/tvm-ffi/actions/workflows/ci_test.yml)
-
-

docs/_static/custom.css

@@ -2,4 +2,4 @@
    See: https://github.com/executablebooks/sphinx-book-theme/issues/732 */
 #rtd-footer-container {
     margin: 0px !important;
-}
+}

docs/conf.py

@@ -16,7 +16,6 @@
 # under the License.
 # -*- coding: utf-8 -*-
 import os
-import sys
 
 import tomli
 
@@ -198,7 +197,7 @@ def footer_html():
               <div class="dropdown">
                   <button class="btn btn-link dropdown-toggle" type="button" id="footerDropdown" data-bs-toggle="dropdown"
                   aria-expanded="false" style="font-size: 0.9em; color: #6c757d; text-decoration: none; padding: 0; border: none; background: none;">
-                      {footer_dropdown['name']}
+                      {footer_dropdown["name"]}
                   </button>
                   <ul class="dropdown-menu" aria-labelledby="footerDropdown" style="font-size: 0.9em;">
 {dropdown_items}                  </ul>
@@ -226,6 +225,6 @@ def footer_html():
     "conf_py_path": "/docs/",
 }
 
-html_static_path = ['_static']
+html_static_path = ["_static"]
 
-html_css_files = ['custom.css']
+html_css_files = ["custom.css"]

docs/get_started/quick_start.md

@@ -279,4 +279,3 @@ The main takeaway points are:
 - **ffi::Tensor** is a universal tensor structure that enables zero-copy exchange of array data
 - **Module loading** is provided by tvm ffi APIs in multiple languages.
 - **C ABI** is provided for easy low-level integration
-

docs/guides/compiler_integration.md

@@ -112,14 +112,14 @@ with various kernel DSLs and libraries.
 
 ## Runtime and State Management for Compilers
 
-While TVM FFI provides a standard ABI for compiler-generated kernels, many compilers and domain-specific languages 
-(DSLs) require their own **runtime** to manage states like dynamic shapes, workspace memory, or other 
-application-specific data. This runtime can be a separate shared library accessible to all kernels from a specific 
+While TVM FFI provides a standard ABI for compiler-generated kernels, many compilers and domain-specific languages
+(DSLs) require their own **runtime** to manage states like dynamic shapes, workspace memory, or other
+application-specific data. This runtime can be a separate shared library accessible to all kernels from a specific
 compiler.
 
 ### Recommended Approach for State Management
 
-The recommended approach for managing compiler-specific state is to define the state within a **separate shared library**. 
+The recommended approach for managing compiler-specific state is to define the state within a **separate shared library**.
 This library exposes its functionality by registering functions as global `tvm::ffi::Function`s.
 
 Here's a breakdown of the process:
@@ -144,21 +144,21 @@ Here's a breakdown of the process:
     This method allows both C++ and Python to access the runtime state through a consistent API.
 3.  **Access State from Kernels**: Within your compiler-generated kernels, you can use
     `GetGlobalRequired("mylang.get_global_state")` in C++ or the C equivalent
-    `TVMFFIGetGlobalFunction("mylang.get_global_state", ...)` to get the function and then call it to retrieve the state 
+    `TVMFFIGetGlobalFunction("mylang.get_global_state", ...)` to get the function and then call it to retrieve the state
     pointer.
 
 ### Distributing the Runtime
 
-For a user to use a kernel from your compiler, they must have access to your runtime library. The preferred method is to 
-package the runtime shared library (e.g., `libmylang_runtime.so`) as part of a Python or C++ package. Users must install 
-and import this package before loading any kernels compiled by your system. 
+For a user to use a kernel from your compiler, they must have access to your runtime library. The preferred method is to
+package the runtime shared library (e.g., `libmylang_runtime.so`) as part of a Python or C++ package. Users must install
+and import this package before loading any kernels compiled by your system.
 This approach ensures the state is shared among different kernels.
 
 ### Common vs. Custom State
 
-It's important to distinguish between compiler-specific state and **common state** managed by TVM FFI. TVM FFI handles 
-common states like **streams** and **memory allocators** through environment functions (e.g., `TVMFFIEnvGetStream`), 
-allowing kernels to access these without managing their own. However, for any unique state required by your compiler, 
+It's important to distinguish between compiler-specific state and **common state** managed by TVM FFI. TVM FFI handles
+common states like **streams** and **memory allocators** through environment functions (e.g., `TVMFFIEnvGetStream`),
+allowing kernels to access these without managing their own. However, for any unique state required by your compiler,
 the global function registration approach is the most suitable method.
 
 ## Advanced: Custom Modules
@@ -196,4 +196,3 @@ the overall import relations from `<import_tree>` and return the final composed
 As long as the compiler generates the `__tvm_ffi__library_bin` in the above format, {py:func}`tvm_ffi.load_module` will correctly
 handle the loading and recover the original module. Note that we will need the custom module class definition to be available
 during loading, either by importing another runtime DLL, or embedding it in the generated library.
-

docs/guides/python_guide.md

@@ -178,7 +178,7 @@ torch.testing.assert_close(x + 1, y)
 ```
 
 The above code defines a C++ function `add_one_cpu` in Python script, compiles it on the fly and then loads the compiled
-{py:class}`tvm_ffi.Module` object via {py:func}`tvm_ffi.cpp.load_inline`. You can then call the function `add_one_cpu` 
+{py:class}`tvm_ffi.Module` object via {py:func}`tvm_ffi.cpp.load_inline`. You can then call the function `add_one_cpu`
 from the module as usual.
 
 ## Error Handling

docs/requirements.txt

@@ -1,6 +1,6 @@
 autodocsumm
-exhale
 breathe
+exhale
 linkify-it-py
 matplotlib
 myst-parser

examples/packaging/python/my_ffi_extension/_ffi_api.py

@@ -17,7 +17,6 @@
 import tvm_ffi
 
 # make sure lib is loaded first
-from .base import _LIB
 
 # this is a short cut to register all the global functions
 # prefixed by `my_ffi_extension.` to this module

examples/quick_start/run_example.py

@@ -21,7 +21,6 @@
 except ImportError:
     torch = None
 
-import ctypes
 
 import numpy
 

examples/quick_start/run_example.sh

@@ -1,3 +1,4 @@
+#!/bin/bash
 # Licensed to the Apache Software Foundation (ASF) under one
 # or more contributor license agreements.  See the NOTICE file
 # distributed with this work for additional information
@@ -14,7 +15,6 @@
 # KIND, either express or implied.  See the License for the
 # specific language governing permissions and limitations
 # under the License.
-#!/bin/bash
 set -ex
 
 cmake -B build -S .

include/tvm/ffi/c_api.h

@@ -272,16 +272,16 @@ typedef struct {
      */
     uint32_t small_str_len;
   };
-  union {                  // 8 bytes
-    int64_t v_int64;       // integers
-    double v_float64;      // floating-point numbers
-    void* v_ptr;           // typeless pointers
-    const char* v_c_str;   // raw C-string
-    TVMFFIObject* v_obj;   // ref counted objects
-    DLDataType v_dtype;    // data type
-    DLDevice v_device;     // device
-    char v_bytes[8];       // small string
-    uint64_t v_uint64;     // uint64 repr mainly used for hashing
+  union {                 // 8 bytes
+    int64_t v_int64;      // integers
+    double v_float64;     // floating-point numbers
+    void* v_ptr;          // typeless pointers
+    const char* v_c_str;  // raw C-string
+    TVMFFIObject* v_obj;  // ref counted objects
+    DLDataType v_dtype;   // data type
+    DLDevice v_device;    // device
+    char v_bytes[8];      // small string
+    uint64_t v_uint64;    // uint64 repr mainly used for hashing
   };
 } TVMFFIAny;
 

include/tvm/ffi/reflection/registry.h

@@ -113,7 +113,7 @@ class AttachFieldFlag : public FieldInfoTrait {
  * \returns The byteoffset
  */
 template <typename Class, typename T>
-TVM_FFI_INLINE int64_t GetFieldByteOffsetToObject(T Class::*field_ptr) {
+TVM_FFI_INLINE int64_t GetFieldByteOffsetToObject(T Class::* field_ptr) {
   int64_t field_offset_to_class =
       reinterpret_cast<int64_t>(&(static_cast<Class*>(nullptr)->*field_ptr));
   return field_offset_to_class - details::ObjectUnsafe::GetObjectOffsetToSubclass<Class>();
@@ -350,7 +350,7 @@ class ObjectDef : public ReflectionDefBase {
    * \return The reflection definition.
    */
   template <typename T, typename BaseClass, typename... Extra>
-  TVM_FFI_INLINE ObjectDef& def_ro(const char* name, T BaseClass::*field_ptr, Extra&&... extra) {
+  TVM_FFI_INLINE ObjectDef& def_ro(const char* name, T BaseClass::* field_ptr, Extra&&... extra) {
     RegisterField(name, field_ptr, false, std::forward<Extra>(extra)...);
     return *this;
   }
@@ -369,7 +369,7 @@ class ObjectDef : public ReflectionDefBase {
    * \return The reflection definition.
    */
   template <typename T, typename BaseClass, typename... Extra>
-  TVM_FFI_INLINE ObjectDef& def_rw(const char* name, T BaseClass::*field_ptr, Extra&&... extra) {
+  TVM_FFI_INLINE ObjectDef& def_rw(const char* name, T BaseClass::* field_ptr, Extra&&... extra) {
     static_assert(Class::_type_mutable, "Only mutable classes are supported for writable fields");
     RegisterField(name, field_ptr, true, std::forward<Extra>(extra)...);
     return *this;
@@ -430,7 +430,7 @@ class ObjectDef : public ReflectionDefBase {
   }
 
   template <typename T, typename BaseClass, typename... ExtraArgs>
-  void RegisterField(const char* name, T BaseClass::*field_ptr, bool writable,
+  void RegisterField(const char* name, T BaseClass::* field_ptr, bool writable,
                      ExtraArgs&&... extra_args) {
     static_assert(std::is_base_of_v<BaseClass, Class>, "BaseClass must be a base class of Class");
     TVMFFIFieldInfo info;

python/tvm_ffi/__init__.py

@@ -15,6 +15,7 @@
 # specific language governing permissions and limitations
 # under the License.
 """TVM FFI Python package."""
+
 # order matters here so we need to skip isort here
 # isort: skip_file
 # base always go first to load the libtvm_ffi

python/tvm_ffi/_convert.py

@@ -15,6 +15,7 @@
 # specific language governing permissions and limitations
 # under the License.
 """Conversion utilities to bring python objects into ffi values."""
+
 from numbers import Number
 from typing import Any
 

python/tvm_ffi/_dtype.py

@@ -15,6 +15,7 @@
 # specific language governing permissions and limitations
 # under the License.
 """dtype class."""
+
 # pylint: disable=invalid-name
 from enum import IntEnum
 
@@ -83,7 +84,10 @@ def with_lanes(self, lanes):
             The new dtype with the given number of lanes.
         """
         cdtype = core._create_dtype_from_tuple(
-            core.DataType, self.__tvm_ffi_dtype__.type_code, self.__tvm_ffi_dtype__.bits, lanes
+            core.DataType,
+            self.__tvm_ffi_dtype__.type_code,
+            self.__tvm_ffi_dtype__.bits,
+            lanes,
         )
         val = str.__new__(dtype, str(cdtype))
         val.__tvm_ffi_dtype__ = cdtype

python/tvm_ffi/_ffi_api.py

@@ -15,6 +15,7 @@
 # specific language governing permissions and limitations
 # under the License.
 """FFI API."""
+
 from . import registry
 
 registry.init_ffi_api("ffi", __name__)

python/tvm_ffi/_optional_torch_c_dlpack.py

@@ -29,6 +29,7 @@
 This module will load slowly at first time due to JITing,
 subsequent calls will be much faster.
 """
+
 import warnings
 
 from . import libinfo

python/tvm_ffi/base.py

@@ -16,6 +16,7 @@
 # under the License.
 # coding: utf-8
 """Base library for TVM FFI."""
+
 import ctypes
 import logging
 import os

python/tvm_ffi/config.py

@@ -37,16 +37,28 @@ def __main__():
         description="Get various configuration information needed to compile with tvm-ffi"
     )
 
-    parser.add_argument("--includedir", action="store_true", help="Print include directory")
     parser.add_argument(
-        "--dlpack-includedir", action="store_true", help="Print dlpack include directory"
+        "--includedir", action="store_true", help="Print include directory"
+    )
+    parser.add_argument(
+        "--dlpack-includedir",
+        action="store_true",
+        help="Print dlpack include directory",
+    )
+    parser.add_argument(
+        "--cmakedir", action="store_true", help="Print library directory"
+    )
+    parser.add_argument(
+        "--sourcedir", action="store_true", help="Print source directory"
+    )
+    parser.add_argument(
+        "--libfiles", action="store_true", help="Fully qualified library filenames"
     )
-    parser.add_argument("--cmakedir", action="store_true", help="Print library directory")
-    parser.add_argument("--sourcedir", action="store_true", help="Print source directory")
-    parser.add_argument("--libfiles", action="store_true", help="Fully qualified library filenames")
     parser.add_argument("--libdir", action="store_true", help="Print library directory")
     parser.add_argument("--libs", action="store_true", help="Libraries to be linked")
-    parser.add_argument("--cython-lib-path", action="store_true", help="Print cython path")
+    parser.add_argument(
+        "--cython-lib-path", action="store_true", help="Print cython path"
+    )
     parser.add_argument("--cxxflags", action="store_true", help="Print cxx flags")
     parser.add_argument("--cflags", action="store_true", help="Print c flags")
     parser.add_argument("--ldflags", action="store_true", help="Print ld flags")