Consolidate coercion code in datafusion_expr::type_coercion and submodules
#3728
Conversation
github-actions
bot
added
logical-expr
| use std::{fmt, str::FromStr}; | ||
|
|
||
| pub static STRINGS: &[DataType] = &[DataType::Utf8, DataType::LargeUtf8]; |
There was a problem hiding this comment.
Moved to type_coercion/aggregate.rs
| //! performed because all valid i32 values can be represented using an | ||
| //! i64. However, i64 -> i32 is never performed as there are i64 | ||
| //! values which can not be represented by i32 values. | ||
| //! | ||
|
|
||
| use crate::{Signature, TypeSignature}; |
There was a problem hiding this comment.
Moved this code into Moved to type_coercion/function.rsas it is related to function coercion, despite thedata_types` generic name
| Ok(()) | ||
| } | ||
| } | ||
| pub mod aggregates; |
There was a problem hiding this comment.
Now the code for various coercion lives in a module that helps identify what it is for
| @@ -421,30 +422,6 @@ fn coercion_decimal_mathematics_type( | |||
| } | |||
| } | |||
|
|
|||
| /// Determine if a DataType is signed numeric or not | |||
There was a problem hiding this comment.
moved into type_coercion.rs as this is not specific to binary expressions
| @@ -17,6 +17,7 @@ | |||
|
|
|||
| //! Coercion rules for matching argument types for binary operators | |||
There was a problem hiding this comment.
I renamed binary_rule.rs to type_coercion/binary.rs to better explain what it is for. I have always been confused about the name binary_rule
I think this module has a lot more code than for binary operators so it can be split / cleaned up more but I felt this is a good start
|
Thanks for your work @alamb |
|
I am going to merge this PR quickly as it is blocking some of my other work I think it has the potential to gather conflicts and I think @liukun4515 is ok with the plan (even though he hasn't yet explicitly approved it). I am happy to make any changes in follow on PRs -- please just let me know |
|
Benchmark runs are scheduled for baseline = 23682f6 and contender = 7c5c2e5. 7c5c2e5 is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
LGTM |
Which issue does this PR close?
re #3582
re #3419
Rationale for this change
I want to continue the good work from @liukun4515 to get type coercion into a single place as part of the logical planning phase, as well as to prepare for the additional work described in #3419 by @andygrove
This PR just moves coercion code around, no functional change is intended
Note that I plan to make a public API for coercion shortly -- see #3708
What changes are included in this PR?
Move coercion code from various places that it was scattered around the codebase into `datafusion_expr::type_coercion and submodules
Are there any user-facing changes?
If anyone calls these functions directly, they will need to change how to find them