Skip to content

Perf: Optimize substring_index via single-byte fast path and direct indexing #19590

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
alamb merged 5 commits into from
Jan 6, 2026
Merged

Perf: Optimize substring_index via single-byte fast path and direct indexing #19590

Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
4356605
optimize substrindex
lyne7-sc Jan 1, 2026
48a6f44
code fmt
lyne7-sc Jan 1, 2026
6e50d84
update bench/substrindex
lyne7-sc Jan 1, 2026
16a24d3
Merge branch 'main' into perf/substrindex
lyne7-sc Jan 2, 2026
0d40231
fix clippy warnings
lyne7-sc Jan 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
63 changes: 38 additions & 25 deletions datafusion/functions/src/unicode/substrindex.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,8 +19,7 @@ use std::any::Any;
use std::sync::Arc;

use arrow::array::{
ArrayAccessor, ArrayIter, ArrayRef, ArrowPrimitiveType, AsArray, OffsetSizeTrait,
PrimitiveArray, StringBuilder,
ArrayAccessor, ArrayIter, ArrayRef, ArrowPrimitiveType, AsArray, GenericStringBuilder, OffsetSizeTrait, PrimitiveArray, StringBuilder
};
use arrow::datatypes::{DataType, Int32Type, Int64Type};

Expand Down Expand Up @@ -182,7 +181,8 @@ fn substr_index_general<
where
T::Native: OffsetSizeTrait,
{
let mut builder = StringBuilder::new();
let num_rows = string_array.len();

@alamb alamb Jan 5, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If you wanted to make this really fast, you could also implement special case code for the common cases where delimiter and count were scalar values -- I suspect you could make it quite fast.

@lyne7-sc lyne7-sc Jan 6, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point! I’ll try adding a special-case implementation for that. Thanks!

let mut builder = GenericStringBuilder::<T::Native>::with_capacity(num_rows, 0);
let string_iter = ArrayIter::new(string_array);
let delimiter_array_iter = ArrayIter::new(delimiter_array);
let count_array_iter = ArrayIter::new(count_array);
Expand All @@ -198,31 +198,44 @@ where
}

let occurrences = usize::try_from(n.unsigned_abs()).unwrap_or(usize::MAX);
let length = if n > 0 {
let split = string.split(delimiter);
split
.take(occurrences)
.map(|s| s.len() + delimiter.len())
.sum::<usize>()
- delimiter.len()
} else {
let split = string.rsplit(delimiter);
split
.take(occurrences)
.map(|s| s.len() + delimiter.len())
.sum::<usize>()
- delimiter.len()
};
if n > 0 {
match string.get(..length) {
Some(substring) => builder.append_value(substring),
None => builder.append_null(),
let result_idx = if delimiter.len() == 1 {

@alamb alamb Jan 5, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I double checked that this checks for length in bytes

https://doc.rust-lang.org/std/string/struct.String.html#method.len

let d_byte = delimiter.as_bytes()[0];
let bytes = string.as_bytes();

if n > 0 {
bytes.iter()
.enumerate()
.filter(|&(_, &b)| b == d_byte)
.nth(occurrences - 1)
.map(|(idx, _)| idx)
} else {
bytes.iter()
.enumerate()
.rev()
.filter(|&(_, &b)| b == d_byte)
.nth(occurrences - 1)
.map(|(idx, _)| idx + 1)
}
} else {
match string.get(string.len().saturating_sub(length)..) {
Some(substring) => builder.append_value(substring),
None => builder.append_null(),
if n > 0 {

@uzqw uzqw Jan 1, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this else { if .. } block can be collapsed

string.match_indices(delimiter)
.nth(occurrences - 1)
.map(|(idx, _)| idx)
} else {
string.rmatch_indices(delimiter)
.nth(occurrences - 1)
.map(|(idx, _)| idx + delimiter.len())
}
};
match result_idx {
Some(idx) => {
if n > 0 {
builder.append_value(&string[..idx]);
} else {
builder.append_value(&string[idx..]);
}
}
None => builder.append_value(string),
}
}
_ => builder.append_null(),
Expand Down