alamb commented on code in PR #6662:
URL: https://github.com/apache/arrow-rs/pull/6662#discussion_r1826003382
##########
arrow-string/src/predicate.rs:
##########
@@ -116,10 +116,17 @@ impl<'a> Predicate<'a> {
}),
Predicate::Contains(finder) => {
if let Some(string_view_array) =
array.as_any().downcast_ref::<StringViewArray>() {
+ let nulls = string_view_array.logical_nulls();
BooleanArray::from(
string_view_array
.bytes_iter()
- .map(|haystack| finder.find(haystack).is_some() !=
negate)
+ .enumerate()
+ .map(|(idx, haystack)| {
+ if nulls.as_ref().map(|n|
n.is_null(idx)).unwrap_or_default() {
Review Comment:
my concern with this approach is that it adds code to the inner loop of the
kernel.
We should definitely run benchmarks on this PR before merging
I think another way to solve this would be to either:
1. Have two loops -- one when `nulls` was `Some` and the other when `nulls`
was `None` (so we aren't checking each row)
2. Leave the computation as is (don't check for nulls on the input) but
instead simply copy the `Nulls` array from the input to the output (so any
input that was `null` would also be `null` in the output)
--
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]
