Split out arrow-ord (#2594)#3299
Conversation
bbb07b7 to
e844da8
Compare
There was a problem hiding this comment.
Parquet doesn't actually depend on any of these kernels, and so this can be removed
db961ea to
540af48
Compare
| - arrow-ipc/** | ||
| - arrow-schema/** | ||
| - arrow-select/** | ||
| - arrow-string/** |
There was a problem hiding this comment.
arrow-flight doesn't depend on this crate and so this can be removed
| gt_eq_utf8_scalar, | ||
| vec![false, false, true, true] | ||
| ); | ||
| test_flag_utf8!( |
There was a problem hiding this comment.
These tests are moved to arrow-string
|
|
||
| let array: ArrayRef = Arc::new(array); | ||
| let array = crate::compute::cast(&array, &DataType::Float64).unwrap(); | ||
| let array = array.unary::<_, Float64Type>(|x| x as f64); |
There was a problem hiding this comment.
It makes me happy that it is now really easy to define your own kernels. This avoids needing a dev dependency on arrow-cast
| run: cargo clippy -p arrow-schema --all-targets --all-features -- -D warnings | ||
| - name: Clippy arrow-array with all features | ||
| run: cargo clippy -p arrow-array --all-targets --all-features -- -D warnings | ||
| - name: Clippy arrow-array with all features except SIMD |
There was a problem hiding this comment.
I created rust-lang/cargo#11467 to track making this less error prone
alamb
left a comment
There was a problem hiding this comment.
Looks good to me -- again I think it is worth being overly conservative and running benchmarks again to ensure we didn't cause any performance regressions accidentally
| [package] | ||
| name = "arrow-ord" | ||
| version = "28.0.0" | ||
| description = "Ordering kernels for arrow arrays" |
There was a problem hiding this comment.
Would "comparison kernels" be a more accurate description?
There was a problem hiding this comment.
There are sort kernels in there, so it is kind of kernels that relate to the ordering of elements? Maybe? I don't really feel strongly, was just trying to justify why it is arrow-ord and not arrow-cmp or something 😅
|
Benchmarks just show noise |
|
Sort kernels also show just noise |
Which issue does this PR close?
Part of #2594
Rationale for this change
What changes are included in this PR?
Are there any user-facing changes?