wgtmac commented on PR #53671: URL: https://github.com/apache/arrow/pull/53671#issuecomment-6149234805
I didn't change the data samples since I don't have a good answer yet. AI has suggested me these cases for both `FLOAT` and `DOUBLE`: | Case | Suggested inputs | Why | |---|---|---| | Constants | `0.0`, `1.25`, `-1.25` | Exercise zero-bit-width packing with zero, positive and negative frame-of-reference values. | | Decimal-like data | Sequences at several decimal scales, with mixed signs | Exercise ALP’s typical workload and exponent/factor selection. | | Vector boundaries | `1`, `1023`, `1024`, `1025` non-null values | Cover short inputs, a full default vector and a partial final vector. | | Sparse exceptions | Decimal-like values with a few NaNs, infinities and `-0.0`, including exceptions at vector starts/ends | Exercise exception placeholders, positions and patching alongside normally encoded values. | | All exceptions | Only NaNs, infinities and `-0.0` | Exercise vectors with no normally encoded values and no packed-value bytes. | | Null/page combinations | Reuse the existing null patterns in both DataPage V1 and V2 files | Exercise non-null value counting and integration with definition levels and page layouts. | However, please correct me if I was wrong, these well-crafted cases will be altered in later fuzz runs at random so I don't think we need all the cases above. Do you have any suggestion? @pitrou -- 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]
