Github user rtreffer commented on a diff in the pull request:

    https://github.com/apache/spark/pull/6796#discussion_r33133591
  
    --- Diff: 
sql/core/src/main/scala/org/apache/spark/sql/parquet/ParquetTypes.scala ---
    @@ -229,11 +231,15 @@ private[parquet] object ParquetTypesConverter extends 
Logging {
         case LongType => Some(ParquetTypeInfo(ParquetPrimitiveTypeName.INT64))
         case TimestampType => 
Some(ParquetTypeInfo(ParquetPrimitiveTypeName.INT96))
         case DecimalType.Fixed(precision, scale) if precision <= 18 =>
    -      // TODO: for now, our writer only supports decimals that fit in a 
Long
           Some(ParquetTypeInfo(ParquetPrimitiveTypeName.FIXED_LEN_BYTE_ARRAY,
             Some(ParquetOriginalType.DECIMAL),
             Some(new DecimalMetadata(precision, scale)),
             Some(BYTES_FOR_PRECISION(precision))))
    +    case DecimalType.Fixed(precision, scale) =>
    +      Some(ParquetTypeInfo(ParquetPrimitiveTypeName.BINARY,
    --- End diff --
    
    Under the assumption that all values will use the full length, yes.
    
    But at some point the overhead of the length is low compared to the 
overhead if someone specifies just the upper bound of values.
    I have to check if it really uses 4 bytes for BINARY. I'd then raise the 
threshold to ~40 bytes length. (meaning <=10% worst case overhead before 
compression)
    
    It won't simplify the decoding/writing though, because the <=18 case  is 
used for long decoding.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org

Reply via email to