pantab 4.0 with C++/nanobind and friends

.gitignore

@@ -2,6 +2,7 @@
 .mypy_cache
 *.hyper
 hyper_db*
+compile_commands.json
 
 #########################################
 # Editor temporary/working/backup files #

CMakeLists.txt

@@ -0,0 +1,35 @@
+project(pantab)
+
+cmake_minimum_required(VERSION 3.18)
+find_package(Python COMPONENTS Interpreter Development.Module NumPy REQUIRED)
+set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++20 -Wall")
+
+if (NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
+  set(CMAKE_BUILD_TYPE Release CACHE STRING "Choose the type of build." FORCE)
+  set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
+endif()
+
+# Detect the installed nanobind package and import it into CMake
+execute_process(
+  COMMAND "${Python_EXECUTABLE}" -m nanobind --cmake_dir
+  OUTPUT_STRIP_TRAILING_WHITESPACE OUTPUT_VARIABLE NB_DIR)
+list(APPEND CMAKE_PREFIX_PATH "${NB_DIR}")
+find_package(nanobind CONFIG REQUIRED)
+
+include(FetchContent)
+FetchContent_Declare(
+  tableauhyperapi-cxx
+  FIND_PACKAGE_ARGS NAMES Tableau
+  URL "https://downloads.tableau.com/tssoftware/tableauhyperapi-cxx-linux-x86_64-release-main.0.0.18369.r86e960ca.zip"
+)
+
+FetchContent_MakeAvailable(tableauhyperapi-cxx)
+list(APPEND CMAKE_PREFIX_PATH "${tableauhyperapi-cxx_SOURCE_DIR}/share/cmake")
+find_package(tableauhyperapi-cxx CONFIG REQUIRED)
+
+if (PANTAB_USE_SANITIZERS)
+    add_compile_options(-fsanitize=address -fsanitize=undefined)
+    add_link_options(-fsanitize=address -fsanitize=undefined)
+endif()
+
+add_subdirectory(pantab/src)

environment.yml

@@ -5,13 +5,14 @@ dependencies:
   - black
   - flake8
   - isort
-  - meson-python
   - mypy
+  - nanobind
   - pandas
   - pip
   - pyarrow
   - python
   - pytest
+  - scikit-build-core
   - sphinx
   - pre-commit
   - sphinx_rtd_theme

pantab/__init__.py

@@ -1,7 +1,5 @@
-__version__ = "3.0.3"
+__version__ = "4.0.0"
 
-import libpantab  # type: ignore
-from tableauhyperapi import __version__ as hyperapi_version
 
 from ._reader import frame_from_hyper, frame_from_hyper_query, frames_from_hyper
 from ._tester import test
@@ -16,112 +14,3 @@
     "frames_to_hyper",
     "test",
 ]
-
-# We link against HyperAPI in a fun way: In Python, we extract the function
-# pointers directly from the Python HyperAPI. We pass those function pointers
-# over to the C module which will then use those pointers to directly interact
-# with HyperAPI. Furthermore, we check the function signatures to guard
-# against API-breaking changes in HyperAPI.
-#
-# Directly using HyperAPI's C functions always was and still is discouraged and
-# unsupported by Tableu. In particular, Tableau will not be able to provide
-# official support for this hack.
-#
-# Because this is highly brittle, we try to make the error message as
-# actionable as possible and guide users in the right direction.
-
-api_incompatibility_msg = """
-pantab is incompatible with version {} of Tableau Hyper API. Please upgrade
-both `tableauhyperapi` and `pantab` to the latest version. See also
-https://pantab.readthedocs.io/en/latest/caveats.html#tableauhyperapi-compatability
-""".format(
-    hyperapi_version
-)
-
-try:
-    from tableauhyperapi.impl.dll import ffi, lib
-except ImportError as e:
-    raise NotImplementedError(api_incompatibility_msg) from e
-
-
-def _check_compatibility(check, message):
-    if not check:
-        raise NotImplementedError(message + "\n" + api_incompatibility_msg)
-
-
-def _get_hapi_function(name, sig):
-    _check_compatibility(hasattr(lib, name), f"function '{name}' missing")
-    f = getattr(lib, name)
-    func_type = ffi.typeof(f)
-    _check_compatibility(
-        func_type.kind == "function",
-        f"expected '{name}' to be a function, got {func_type.kind}",
-    )
-    _check_compatibility(
-        func_type.cname == sig,
-        f"expected '{name}' to have the signature '{sig}', got '{func_type.cname}'",
-    )
-    return f
-
-
-libpantab.load_hapi_functions(
-    _get_hapi_function("hyper_decode_date", "hyper_date_components_t(*)(uint32_t)"),
-    _get_hapi_function("hyper_encode_date", "uint32_t(*)(hyper_date_components_t)"),
-    _get_hapi_function("hyper_decode_time", "hyper_time_components_t(*)(uint64_t)"),
-    _get_hapi_function("hyper_encode_time", "uint64_t(*)(hyper_time_components_t)"),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_null",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_bool",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *, _Bool)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_int16",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *, int16_t)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_int32",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *, int32_t)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_int64",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *, int64_t)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_double",
-        "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *, double)",
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_binary",
-        (
-            "struct hyper_error_t *(*)"
-            "(struct hyper_inserter_buffer_t *, uint8_t *, size_t)"
-        ),
-    ),
-    _get_hapi_function(
-        "hyper_inserter_buffer_add_raw",
-        (
-            "struct hyper_error_t *(*)(struct hyper_inserter_buffer_t *"
-            ", uint8_t *, size_t)"
-        ),
-    ),
-    _get_hapi_function(
-        "hyper_rowset_get_next_chunk",
-        (
-            "struct hyper_error_t *(*)(struct hyper_rowset_t *"
-            ", struct hyper_rowset_chunk_t * *)"
-        ),
-    ),
-    _get_hapi_function(
-        "hyper_destroy_rowset_chunk", "void(*)(struct hyper_rowset_chunk_t *)"
-    ),
-    _get_hapi_function(
-        "hyper_rowset_chunk_field_values",
-        (
-            "void(*)(struct hyper_rowset_chunk_t *"
-            ", size_t *, size_t *, uint8_t * * *, size_t * *)"
-        ),
-    ),
-)