move Log2, Log10, Ln to datafusion-functions #9869
Conversation
github-actions
bot
added
logical-expr
tinfoil-knight
changed the title
There was a problem hiding this comment.
Thank you so much @tinfoil-knight -- 🚀
Didn't migrate log because some parts (eg: simplifying expressions: see simpl_log, simpl_power) of log & power functions are dependent on each other & IMO these should be migrated together.
I agree -- thank you, this makes sense
Note I think we can implement simpl_log using ScalarUDFImpl::simplify (for example https://github.com/apache/arrow-datafusion/blob/9d47dcaab85294f517c899464c442f9a4518b6f2/datafusion/functions/src/datetime/current_date.rs#L76-L91)
It might be a fun little PR if you wanted to try
| make_math_unary_udf!(AcosFunc, ACOS, acos, acos); | ||
| make_math_unary_udf!(AsinFunc, ASIN, asin, asin); | ||
| make_math_unary_udf!(TanFunc, TAN, tan, tan); | ||
| make_math_unary_udf!(Log2Func, LOG2, log2, log2, Some(vec![Some(true)])); |
There was a problem hiding this comment.
I wonder if we could make a constant or something that would make this easier to read
For example
/// Single monotonic argument
const MONOTONIC_ONE: Option<Vec<bool>> = Some(vec![Some(true)])
const NON_MONOTONIC: Option<Vec<bool>> = None;And then
| make_math_unary_udf!(Log2Func, LOG2, log2, log2, Some(vec![Some(true)])); | |
| make_math_unary_udf!(Log2Func, LOG2, log2, log2, MONOTONIC_ONE); |
Perhaps as a follow on PR
There was a problem hiding this comment.
@alamb
I tried this readability fix but Rust doesn't allow allocation (which vec![] needs) in const or static items. Global let is also not allowed.
The compiler suggested this with static:
consider wrapping this expression in `Lazy::new(|| ...)` from the `once_cell` crate: https://crates.io/crates/once_cell
We use once_cell in the physical_plan crate but introducing it to datafusion functions just for this doesn't seem right to me.
There was a problem hiding this comment.
We use
once_cellin thephysical_plancrate but introducing it to datafusion functions just for this doesn't seem right to me.
That makes sense to me. We would probably have to throw in a clone() or something
I thought about this and wrote up another API that would be much nicer #9879 (both for this callsite and montonicity in general)
Yeah. I'll try this out. Edit: |
* move log2 * move log10, ln * refactor log_b functions to use macro * update proto
Which issue does this PR close?
Closes #9868.
Rationale for this change
As part of #9285,
log10,log2&lnfunctions should be migrated to the datafusion-functions crate.Sidenote:
Didn't migrate
logbecause some parts (eg: simplifying expressions: seesimpl_log,simpl_power) oflog&powerfunctions are dependent on each other & IMO these should be migrated together.What changes are included in this PR?
Extend the
make_math_unary_udfto accept a parameter for monotonicity & migrate log10, log2, ln from built-in functions to thedatafusion-functionscrate.Are these changes tested?
Covered by existing tests.
Are there any user-facing changes?
No.