Explicit errors message when arrow version mismatch (#80)

python/.gitignore

@@ -6,4 +6,5 @@ __pycache__
 vcpkg_installed
 .vscode
 /palletjack/palletjack_cython.cpp
+/palletjack/package_metadata.py
 cython_debug

python/palletjack/__init__.py

@@ -1,3 +1,17 @@
 # Importing pyarrow is necessary to load the runtime libraries
+import pyarrow
 import pyarrow.parquet
-from .palletjack_cython import *
+
+from .package_metadata import (
+    __version__,
+    __dependencies__
+)
+
+try:
+    from .palletjack_cython import *
+except ImportError as e:
+    if any(x in str(e) for x in ['arrow', 'parquet']):
+        pyarrow_req = next((r for r in __dependencies__ if r.startswith('pyarrow')), '')
+        raise ImportError(f"This version of {__package__}={__version__} is built against {pyarrow_req}, please ensure you have it installed. Current pyarrow version is {pyarrow.__version__}. ({str(e)})") from None
+    else:
+        raise

python/pyproject.toml

@@ -11,7 +11,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "palletjack"
-version = "2.5.0"
+version = "2.5.1"
 description = "Faster parquet metadata reading"
 readme = "README.md"
 requires-python = ">=3.9"

python/setup.py

@@ -3,7 +3,8 @@
 import sys
 from codecs import open
 
-from setuptools import setup, find_packages
+from setuptools import setup
+from setuptools.command.build_py import build_py
 from distutils.extension import Extension
 from Cython.Build import cythonize
 import pyarrow
@@ -62,6 +63,37 @@ def no_cythonize(extensions, **_ignore):
 else:
     extensions = no_cythonize(extensions)
 
+# Custom build command to dynamically generate metadata file
+class GenerateMetadata(build_py):
+    def run(self):
+        # Call the original build_py command
+        super().run()
+
+        # Get the distribution object
+        dist = self.distribution
+
+        package_name = dist.get_name()
+        package_version = dist.get_version()
+        package_dependencies = dist.install_requires
+
+        print (f"package_name = {package_name}")
+        print (f"package_version = {package_version}")
+        print (f"package_dependencies = {package_dependencies}")
+
+        output_dir = os.path.join(package_name)
+        os.makedirs(output_dir, exist_ok=True)
+        metadata_file = os.path.join(output_dir, "package_metadata.py")
+
+        # Write metadata to the file
+        with open(metadata_file, "w") as f:
+            f.write("# Auto-generated package metadata\n")
+            f.write(f"__package__ = '{package_name}'\n")
+            f.write(f"__version__ = '{package_version}'\n")
+            f.write(f"__dependencies__ = {package_dependencies}\n")
+
+        print(f"Generated metadata file: {metadata_file}")
+        print (os.listdir(output_dir))
+
 # Make default named pyarrow shared libs available.
 pyarrow.create_library_symlinks()
 
@@ -74,4 +106,7 @@ def no_cythonize(extensions, **_ignore):
         "Documentation": "https://github.com/G-Research/PalletJack",
         "Source": "https://github.com/G-Research/PalletJack",
     },
+    cmdclass={
+        "build_py": GenerateMetadata,  # Use the custom build command
+    },
 )