Merge pull request #1140 from Altinity/backports/antalya-25.8/87444

Antalya 25.8 Backport of ClickHouse#87444 - Make parquet reader v3 preserve decimal precision information

Author: Enmk

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

@@ -1045,7 +1045,7 @@ void SchemaConverter::processPrimitiveColumn(
         if (precision > max_precision)
             throw Exception(ErrorCodes::INCORRECT_DATA, "Parquet decimal type precision or scale is too big ({} digits) for physical type {}", precision, thriftToString(type));
 
-        out_inferred_type = createDecimal<DataTypeDecimal>(max_precision, scale);
+        out_inferred_type = createDecimal<DataTypeDecimal>(precision, scale);
         size_t output_size = out_inferred_type->getSizeOfValueInMemory();
         out_decoder.allow_stats = is_output_type_decimal(output_size, scale);
 

tests/queries/0_stateless/02845_parquet_odd_decimals.reference

@@ -1 +1,5 @@
 100
+col-1de12c05-5dd5-4fa7-9f93-33c43c9a4028	Decimal(20, 0)					
+col-5e1b97f1-dade-4c7d-b71b-e31d789e01a4	String					
+col-1de12c05-5dd5-4fa7-9f93-33c43c9a4028	Decimal(20, 0)					
+col-5e1b97f1-dade-4c7d-b71b-e31d789e01a4	String					

tests/queries/0_stateless/02845_parquet_odd_decimals.sh

@@ -11,3 +11,6 @@ ${CLICKHOUSE_CLIENT} --query="drop table if exists 02845_parquet_odd_decimals"
 ${CLICKHOUSE_CLIENT} --query="create table 02845_parquet_odd_decimals (\`col-1de12c05-5dd5-4fa7-9f93-33c43c9a4028\` Decimal(20, 0), \`col-5e1b97f1-dade-4c7d-b71b-e31d789e01a4\` String) engine Memory"
 ${CLICKHOUSE_CLIENT} --query="insert into 02845_parquet_odd_decimals from infile '$CUR_DIR/data_parquet/nine_byte_decimals_from_spark.parquet'"
 ${CLICKHOUSE_CLIENT} --query="select count() from 02845_parquet_odd_decimals"
+
+${CLICKHOUSE_LOCAL} --query="desc file('$CUR_DIR/data_parquet/nine_byte_decimals_from_spark.parquet') settings schema_inference_make_columns_nullable=0, input_format_parquet_use_native_reader_v3=0"
+${CLICKHOUSE_LOCAL} --query="desc file('$CUR_DIR/data_parquet/nine_byte_decimals_from_spark.parquet') settings schema_inference_make_columns_nullable=0, input_format_parquet_use_native_reader_v3=1"