Antalya 25.8 Backport of #87095 - Fix simple parquet reader v3 assert on bool column min/max

src/Processors/Formats/Impl/Parquet/Decoding.cpp

@@ -1096,6 +1096,8 @@ void IntConverter::convertField(std::span<const char> data, bool /*is_max*/, Fie
     UInt64 val = 0;
     switch (input_size)
     {
+        case 1: val = unalignedLoad<UInt8>(data.data()); break;
+        case 2: val = unalignedLoad<UInt16>(data.data()); break;
         case 4: val = unalignedLoad<UInt32>(data.data()); break;
         case 8: val = unalignedLoad<UInt64>(data.data()); break;
         default: chassert(false);

src/Processors/Formats/Impl/Parquet/Decoding.h

@@ -155,7 +155,8 @@ struct PageDecoderInfo
 };
 
 
-/// Input physical type: INT32 or INT64.
+/// Input physical type: BOOLEAN, INT32, or INT64.
+/// input_size in {1, 4, 8}.
 /// Output column type: [U]Int{8,16,32,64}.
 /// Output Field type: [U]Int64, IPv4, or Decimal{32,64}.
 struct IntConverter : public FixedSizeConverter

src/Processors/Formats/Impl/Parquet/SchemaConverter.cpp

@@ -1136,6 +1136,8 @@ void SchemaConverter::processPrimitiveColumn(
             out_inferred_type = DataTypeFactory::instance().get("Bool");
             auto converter = std::make_shared<IntConverter>();
             converter->input_size = 1;
+            converter->input_signed = false;
+            converter->field_signed = false;
             out_decoder.allow_stats = dispatch_int_stats_converter(/*allow_datetime_and_ipv4=*/ false, *converter);
             out_decoder.fixed_size_converter = std::move(converter);
             return;

tests/queries/0_stateless/03623_parquet_bool.reference

@@ -0,0 +1 @@
+true

tests/queries/0_stateless/03623_parquet_bool.sql

@@ -0,0 +1,4 @@
+-- Tags: no-parallel, no-fasttest
+
+insert into function file('03626_parquet_bool.parquet') select true as x settings engine_file_truncate_on_insert=1;
+select * from file('03626_parquet_bool.parquet') where x=1 settings input_format_parquet_use_native_reader_v3=1;