Dear Parquet Community,

We (Thomas, Oleksii, Jonas) would like to start a discussion about adding a
new numeric logical type, DECFLOAT, to the Parquet specification.

The DECFLOAT type is an 18-byte decimal floating-point numeric
representation with a maximum precision of 38 digits It addresses the value
range limitations inherent in fixed-point and loss of exactness in binary
floating-point types, making it especially suitable for financial
computations and scenarios requiring high numeric precision, range, and
exactness.

Introducing DECFLOAT also enhances compatibility and simplifies the data
interchange with many enterprise database systems (Oracle, DB2, MongoDB,
Snowflake) that already support or plan to support decimal floating-point
numeric types. Additionally, it aligns seamlessly with decimal
floating-point implementations in major programming languages, such as
Java's BigDecimal, Python's decimal module, and libraries like MPDecimal
for C/C++.

For further details about the design and rationale, please review the full
proposal:

https://docs.google.com/document/d/1j_Q6vnn6Nhy60K4o0tdC91kE5vKGNJaoDOAm71KLzNw/edit?usp=sharing

We look forward to your feedback and discussion.

Best regards,
Jonas

Reply via email to