diff --git a/src/io/parquet/read/deserialize/simple.rs b/src/io/parquet/read/deserialize/simple.rs index b4b614980e8..bac15f80357 100644 --- a/src/io/parquet/read/deserialize/simple.rs +++ b/src/io/parquet/read/deserialize/simple.rs @@ -8,7 +8,7 @@ use parquet2::{ use crate::types::i256; use crate::{ - array::{Array, DictionaryKey, MutablePrimitiveArray, PrimitiveArray}, + array::{Array, DictionaryKey, MutablePrimitiveArray, PrimitiveArray, BinaryArray}, datatypes::{DataType, IntervalUnit, TimeUnit}, error::{Error, Result}, types::{days_ms, NativeType}, @@ -230,6 +230,36 @@ pub fn page_iter_to_arrays<'a, I: Pages + 'a>( Box::new(arrays) as _ } + (PhysicalType::ByteArray, Decimal(_, _)) => { + let pages = binary::Iter::::new( + pages, + DataType::Binary, + chunk_size, + num_rows, + ); + + let pages = pages.map(move |maybe_array| { + let array = maybe_array?; + let array_len = array.len(); + let array = array.as_any().downcast_ref::>().unwrap(); + let values = (0..array_len).map(|i| { + let value = array.value(i); + let n = value.len(); + if n > 16 { + return Err(Error::Overflow); + } + Ok(super::super::convert_i128(value, n)) + }) + .collect::>>(); + let validity = array.validity().cloned(); + + PrimitiveArray::::try_new(data_type.clone(), values?.into(), validity) + }); + + let arrays = pages.map(|x| x.map(|x| x.boxed())); + + Box::new(arrays) as _ + } (PhysicalType::Int32, Decimal256(_, _)) => dyn_iter(iden(primitive::IntegerIter::new( pages, data_type, diff --git a/src/io/parquet/read/schema/convert.rs b/src/io/parquet/read/schema/convert.rs index 821d5107649..a8ea3dc3928 100644 --- a/src/io/parquet/read/schema/convert.rs +++ b/src/io/parquet/read/schema/convert.rs @@ -142,6 +142,8 @@ fn from_byte_array( (_, Some(PrimitiveConvertedType::Bson)) => DataType::Binary, (_, Some(PrimitiveConvertedType::Enum)) => DataType::Binary, (_, Some(PrimitiveConvertedType::Utf8)) => DataType::Utf8, + (Some(PrimitiveLogicalType::Decimal(precision, scale)), _) => DataType::Decimal(*precision, *scale), + (_, Some(PrimitiveConvertedType::Decimal(precision, scale))) => DataType::Decimal(*precision, *scale), (_, _) => DataType::Binary, } }