asfimport commented on issue #406: URL: https://github.com/apache/parquet-format/issues/406#issuecomment-2184154147
[Xuwei Fu](https://issues.apache.org/jira/browse/PARQUET-2249?#comment-17691179) / @mapleFU: The problem seem to be that, float point is so widely used, but they are "partial order". Seems that iceberg provides NaN counts. And min-max is un-related to NaN. In Sorting, iceberg forces that: > Sorting floating-point numbers should produce the following behavior: `-NaN` < `-Infinity` < `-value` < `-0` < `0` < `value` < `Infinity` < `{}NaN{`}. This aligns with the implementation of Java floating-point types comparisons. I think (1) is bad, because NaN is never equal to NULL. IEEE754 (<https://ieeexplore.ieee.org/document/8766229>) and C++ standard support some "totalOrder", but I think regard it as totalOrder is strange, so min-max as `byte[0]` is wierd for me. I think iceberg style looks great If I'm wrong, please correct me -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
