Add validation for integer values to prevent overflow in dataset creation and writing

Author: kojix2

README.md

@@ -18,6 +18,9 @@ Unsupported at this stage:
 - string dataset read/write
 - multidimensional array write
 
+Integer datasets and integer attributes currently use native C `int` under the hood.
+Values outside that range are rejected with `HDF5::Error` to avoid silent overflow.
+
 ## Supported HDF5 Versions
 
 - HDF5 1.10

lib/hdf5/dataset.rb

@@ -46,6 +46,7 @@ def normalize_data(data)
 
       def datatype_id_for(data)
         if data.all? { |value| value.is_a?(Integer) }
+          validate_native_int_range!(data)
           HDF5::FFI.H5T_NATIVE_INT
         elsif data.all? { |value| value.is_a?(Numeric) }
           HDF5::FFI.H5T_NATIVE_DOUBLE
@@ -66,7 +67,23 @@ def buffer_for(data)
         buffer
       end
 
-      private :normalize_data, :datatype_id_for, :buffer_for
+      def native_int_bounds
+        bits = ::FFI.type_size(:int) * 8
+        max = (1 << (bits - 1)) - 1
+        min = -(1 << (bits - 1))
+        [min, max]
+      end
+
+      def validate_native_int_range!(values)
+        min, max = native_int_bounds
+        out_of_range = values.find { |value| value < min || value > max }
+        return unless out_of_range
+
+        raise HDF5::Error,
+              "Integer value #{out_of_range} is outside native int range (#{min}..#{max}). Use a smaller value."
+      end
+
+      private :normalize_data, :datatype_id_for, :buffer_for, :native_int_bounds, :validate_native_int_range!
 
       private
 
@@ -87,8 +104,8 @@ def attrs
 
     def write(data)
       values = self.class.send(:normalize_data, data)
-      buffer = self.class.send(:buffer_for, values)
       mem_type_id = self.class.send(:datatype_id_for, values)
+      buffer = self.class.send(:buffer_for, values)
       status = HDF5::FFI.H5Dwrite(@dataset_id, mem_type_id, HDF5::DEFAULT_PROPERTY_LIST, HDF5::DEFAULT_PROPERTY_LIST,
                                   HDF5::DEFAULT_PROPERTY_LIST, buffer)
       raise HDF5::Error, 'Failed to write dataset' if status < 0

test/hdf5_test.rb

@@ -74,6 +74,39 @@ class HDF5Test < Test::Unit::TestCase
     end
   end
 
+  test 'rejects integer values outside native int range on create' do
+    min = -(1 << (::FFI.type_size(:int) * 8 - 1))
+    max = (1 << (::FFI.type_size(:int) * 8 - 1)) - 1
+
+    Dir.mktmpdir do |dir|
+      path = File.join(dir, 'overflow-create.h5')
+
+      HDF5::File.create(path) do |file|
+        error = assert_raise(HDF5::Error) do
+          file.create_dataset('too_large', [max + 1])
+        end
+        assert_include(error.message, "#{min}..#{max}")
+      end
+    end
+  end
+
+  test 'rejects integer values outside native int range on write' do
+    min = -(1 << (::FFI.type_size(:int) * 8 - 1))
+    max = (1 << (::FFI.type_size(:int) * 8 - 1)) - 1
+
+    Dir.mktmpdir do |dir|
+      path = File.join(dir, 'overflow-write.h5')
+
+      HDF5::File.create(path) do |file|
+        dataset = file.create_dataset('ints', [1, 2, 3])
+        error = assert_raise(HDF5::Error) do
+          dataset.write([min - 1])
+        end
+        assert_include(error.message, "#{min}..#{max}")
+      end
+    end
+  end
+
   test 'block API closes resources automatically' do
     Dir.mktmpdir do |dir|
       path = File.join(dir, 'block.h5')