kiszk commented on issue #6981:
URL: https://github.com/apache/arrow/pull/6981#issuecomment-616453579


   I understand your point. First, I implemented the approach ` entirely 
little-endian`.
   
   Then, I reconsidered it. Each primitive type should be represented in a 
little-endian as shown 
[here](https://github.com/apache/parquet-format/blob/master/Encodings.md) in 
**Parquet serialized form (e.g. file)**.  For example, int32 is kept using a 
native endian in a memory or register. It will be converted from/to a 
little-endian when it will be serialized. I extended this to Int96. This is a 
background of the current decision.
   
   Both approaches can work functionally correct. I understand that this is an 
implementation decision.
   
   It is ok with your approach. At that time, I have one question. How will we 
handle other primitive types (e.g. int32 or double)? Do we keep other 
primitives in entirely little-endian on both platforms (little-endian and 
big-endian CPUs)?
   
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


Reply via email to