fix: make Avro reader buffer size configurable (#447)

Add ReaderProperties::kAvroBufferSize property to allow users to
configure the Avro input stream buffer size instead of using a hardcoded
1MB value.

Changes:
- Add kAvroBufferSize property to ReaderProperties (default: 1MB)
- Update AvroReader to use the configurable buffer size
- Remove TODO comment about making this configurable

Author: shangxinli

src/iceberg/avro/avro_reader.cc

@@ -89,10 +89,10 @@ class AvroReader::Impl {
     read_schema_ = options.projection;
 
     // Open the input stream and adapt to the avro interface.
-    // TODO(gangwu): make this configurable
-    constexpr int64_t kDefaultBufferSize = 1024 * 1024;
-    ICEBERG_ASSIGN_OR_RAISE(auto input_stream,
-                            CreateInputStream(options, kDefaultBufferSize));
+    ICEBERG_ASSIGN_OR_RAISE(
+        auto input_stream,
+        CreateInputStream(options,
+                          options.properties->Get(ReaderProperties::kAvroBufferSize)));
 
     ::avro::ValidSchema file_schema;
 

src/iceberg/file_reader.h

@@ -81,6 +81,9 @@ class ReaderProperties : public ConfigBase<ReaderProperties> {
   /// Default: true (skip GenericDatum for better performance).
   inline static Entry<bool> kAvroSkipDatum{"read.avro.skip-datum", true};
 
+  /// \brief The buffer size used by Avro input stream.
+  inline static Entry<int64_t> kAvroBufferSize{"read.avro.buffer-size", 1024 * 1024};
+
   /// \brief Create a default ReaderProperties instance.
   static std::unique_ptr<ReaderProperties> default_properties();
 

src/iceberg/test/avro_test.cc

@@ -503,4 +503,27 @@ INSTANTIATE_TEST_SUITE_P(DirectDecoderModes, AvroReaderParameterizedTest,
                            return info.param ? "DirectDecoder" : "GenericDatum";
                          });
 
+TEST_F(AvroReaderTest, BufferSizeConfiguration) {
+  // Test default buffer size
+  auto properties1 = ReaderProperties::default_properties();
+  ASSERT_EQ(properties1->Get(ReaderProperties::kAvroBufferSize), 1024 * 1024);
+
+  // Test setting custom buffer size
+  auto properties2 = ReaderProperties::default_properties();
+  constexpr int64_t kCustomBufferSize = 2 * 1024 * 1024;  // 2MB
+  properties2->Set(ReaderProperties::kAvroBufferSize, kCustomBufferSize);
+  ASSERT_EQ(properties2->Get(ReaderProperties::kAvroBufferSize), kCustomBufferSize);
+
+  // Test setting via FromMap
+  std::unordered_map<std::string, std::string> config_map = {
+      {"read.avro.buffer-size", "4194304"}  // 4MB
+  };
+  auto properties3 = ReaderProperties::FromMap(config_map);
+  ASSERT_EQ(properties3->Get(ReaderProperties::kAvroBufferSize), 4194304);
+
+  // Test that unset returns to default
+  properties2->Unset(ReaderProperties::kAvroBufferSize);
+  ASSERT_EQ(properties2->Get(ReaderProperties::kAvroBufferSize), 1024 * 1024);
+}
+
 }  // namespace iceberg::avro