metegenez commented on code in PR #21420:
URL: https://github.com/apache/datafusion/pull/21420#discussion_r3045224213
##########
datafusion/functions/src/string/split_part.rs:
##########
@@ -425,6 +424,116 @@ fn rsplit_nth_finder<'a>(
}
}
+/// Zero-copy scalar fast path for `StringViewArray` inputs.
+///
+/// Instead of copying substring bytes into a new buffer, constructs
+/// `StringView` entries that point back into the original array's data
+/// buffers.
+fn split_part_scalar_view(
+ string_view_array: &StringViewArray,
+ delimiter: &str,
+ position: i64,
+) -> Result<ArrayRef> {
+ let len = string_view_array.len();
+ let mut views_buf = Vec::with_capacity(len);
+ let views = string_view_array.views();
+
+ if delimiter.is_empty() {
+ // PostgreSQL: empty delimiter treats input as a single field.
+ let empty_view = make_view(b"", 0, 0);
+ let return_input = position == 1 || position == -1;
+ for i in 0..len {
+ if string_view_array.is_null(i) {
+ views_buf.push(0);
+ } else if return_input {
+ views_buf.push(views[i]);
+ } else {
+ views_buf.push(empty_view);
+ }
+ }
+ } else if position > 0 {
+ let idx: usize = (position - 1).try_into().map_err(|_| {
+ exec_datafusion_err!(
+ "split_part index {position} exceeds maximum supported value"
+ )
+ })?;
+ let finder = memmem::Finder::new(delimiter.as_bytes());
+ split_view_loop(string_view_array, views, &mut views_buf, |s| {
+ split_nth_finder(s, &finder, delimiter.len(), idx)
+ });
+ } else {
+ let idx: usize = (position.unsigned_abs() - 1).try_into().map_err(|_| {
Review Comment:
upstream guarantees position will not be zero, so this looks safe.
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
