martin-g commented on code in PR #20069:
URL: https://github.com/apache/datafusion/pull/20069#discussion_r2746466660
##########
datafusion/functions/src/unicode/right.rs:
##########
@@ -119,58 +119,140 @@ impl ScalarUDFImpl for RightFunc {
}
}
-/// Returns last n characters in the string, or when n is negative, returns
all but first |n| characters.
+/// Returns right n characters in the string, or when n is negative, returns
all but first |n| characters.
/// right('abcde', 2) = 'de'
+/// right('abcde', -2) = 'cde'
/// The implementation uses UTF-8 code points as characters
-fn right<T: OffsetSizeTrait>(args: &[ArrayRef]) -> Result<ArrayRef> {
+fn right(args: &[ArrayRef]) -> Result<ArrayRef> {
let n_array = as_int64_array(&args[1])?;
- if args[0].data_type() == &DataType::Utf8View {
- // string_view_right(args)
- let string_array = as_string_view_array(&args[0])?;
- right_impl::<T, _>(&mut string_array.iter(), n_array)
- } else {
- // string_right::<T>(args)
- let string_array = &as_generic_string_array::<T>(&args[0])?;
- right_impl::<T, _>(&mut string_array.iter(), n_array)
+
+ match args[0].data_type() {
+ DataType::Utf8 => {
+ let string_array = as_generic_string_array::<i32>(&args[0])?;
+ right_impl::<i32, _>(string_array, n_array)
+ }
+ DataType::LargeUtf8 => {
+ let string_array = as_generic_string_array::<i64>(&args[0])?;
+ right_impl::<i64, _>(string_array, n_array)
+ }
+ DataType::Utf8View => {
+ let string_view_array = as_string_view_array(&args[0])?;
+ right_impl_view(string_view_array, n_array)
+ }
+ _ => exec_err!("Not supported"),
}
}
-// Currently the return type can only be Utf8 or LargeUtf8, to reach fully
support, we need
-// to edit the `get_optimal_return_type` in utils.rs to make the udfs be able
to return Utf8View
-// See
https://github.com/apache/datafusion/issues/11790#issuecomment-2283777166
+/// `right` implementation for strings
fn right_impl<'a, T: OffsetSizeTrait, V: ArrayAccessor<Item = &'a str>>(
- string_array_iter: &mut ArrayIter<V>,
+ string_array: V,
n_array: &Int64Array,
) -> Result<ArrayRef> {
- let result = string_array_iter
+ let iter = ArrayIter::new(string_array);
+ let result = iter
.zip(n_array.iter())
.map(|(string, n)| match (string, n) {
- (Some(string), Some(n)) => match n.cmp(&0) {
- Ordering::Less => Some(
- string
- .chars()
- .skip(n.unsigned_abs() as usize)
- .collect::<String>(),
- ),
- Ordering::Equal => Some("".to_string()),
- Ordering::Greater => Some(
- string
- .chars()
- .skip(max(string.chars().count() as i64 - n, 0) as
usize)
- .collect::<String>(),
- ),
- },
+ (Some(string), Some(n)) => {
+ let byte_length = right_byte_length(string, n);
+ // println!(
+ // "Input string: {}, n: {} -> byte_length: {} -> {}",
+ // &string,
+ // n,
+ // byte_length,
+ // &string[byte_length..]
+ // );
+ // Extract starting from `byte_length` bytes from a
byte-indexed slice
+ Some(&string[byte_length..])
+ }
_ => None,
})
.collect::<GenericStringArray<T>>();
Ok(Arc::new(result) as ArrayRef)
}
+/// `right` implementation for StringViewArray
+fn right_impl_view(
+ string_view_array: &StringViewArray,
+ n_array: &Int64Array,
+) -> Result<ArrayRef> {
+ let len = n_array.len();
+
+ let views = string_view_array.views();
+ // Every string in StringViewArray has one corresponding view in `views`
+ debug_assert!(views.len() == string_view_array.len());
+
+ // Compose null buffer at once
+ let string_nulls = string_view_array.nulls();
+ let n_nulls = n_array.nulls();
+ let new_nulls = NullBuffer::union(string_nulls, n_nulls);
+
+ let new_views = (0..len)
+ .map(|idx| {
+ let view = views[idx];
+
+ let is_valid = match &new_nulls {
+ Some(nulls_buf) => nulls_buf.is_valid(idx),
+ None => true,
+ };
+
+ if is_valid {
+ let string: &str = string_view_array.value(idx);
+ let n = n_array.value(idx);
+
+ let new_offset = right_byte_length(string, n);
+ let result_bytes = &string.as_bytes()[new_offset..];
+
+ if result_bytes.len() > 12 {
+ let byte_view = ByteView::from(view);
+ // Reuse buffer, but adjust offset and length
+ make_view(
+ result_bytes,
+ byte_view.buffer_index,
+ byte_view.offset + new_offset as u32,
+ )
+ } else {
+ // inline value does not need block id or offset
+ make_view(result_bytes, 0, 0)
+ }
+ } else {
+ // For nulls, keep the original view
+ view
+ }
+ })
+ .collect::<Vec<u128>>();
+
+ // Buffers are unchanged
+ let result = StringViewArray::try_new(
+ ScalarBuffer::from(new_views),
+ Vec::from(string_view_array.data_buffers()),
+ new_nulls,
+ )?;
+ Ok(Arc::new(result) as ArrayRef)
+}
+
+/// Calculate the byte length of the substring of last `n` chars from string
`string`
+/// (or all but first `|n|` chars if n is negative)
+fn right_byte_length(string: &str, n: i64) -> usize {
+ match n.cmp(&0) {
+ Ordering::Less => string
+ .char_indices()
+ .nth(n.unsigned_abs() as usize)
+ .map(|(index, _)| index)
+ .unwrap_or(string.len()),
+ Ordering::Equal => string.len(),
+ Ordering::Greater => string
+ .char_indices()
+ .nth_back(n.unsigned_abs() as usize - 1)
Review Comment:
nit: This may truncate on 32-bit machines
##########
datafusion/functions/benches/right.rs:
##########
@@ -0,0 +1,140 @@
+// Licensed to the Apache Software Foundation (ASF) under one
+// or more contributor license agreements. See the NOTICE file
+// distributed with this work for additional information
+// regarding copyright ownership. The ASF licenses this file
+// to you under the Apache License, Version 2.0 (the
+// "License"); you may not use this file except in compliance
+// with the License. You may obtain a copy of the License at
+//
+// http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing,
+// software distributed under the License is distributed on an
+// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+// KIND, either express or implied. See the License for the
+// specific language governing permissions and limitations
+// under the License.
+
+extern crate criterion;
+
+use std::hint::black_box;
+use std::sync::Arc;
+
+use arrow::array::{ArrayRef, Int64Array};
+use arrow::datatypes::{DataType, Field};
+use arrow::util::bench_util::{
+ create_string_array_with_len, create_string_view_array_with_len,
+};
+use criterion::{BenchmarkId, Criterion, criterion_group, criterion_main};
+use datafusion_common::config::ConfigOptions;
+use datafusion_expr::{ColumnarValue, ScalarFunctionArgs};
+use datafusion_functions::unicode::right;
+
+fn create_args(
+ size: usize,
+ str_len: usize,
+ use_negative: bool,
+ is_string_view: bool,
+) -> Vec<ColumnarValue> {
+ let string_arg = if is_string_view {
+ ColumnarValue::Array(Arc::new(create_string_view_array_with_len(
+ size, 0.1, str_len, true,
+ )))
+ } else {
+ ColumnarValue::Array(Arc::new(create_string_array_with_len::<i32>(
+ size, 0.1, str_len,
+ )))
+ };
+
+ // For negative n, we want to trigger the double-iteration code path
+ let n_values: Vec<i64> = if use_negative {
+ (0..size).map(|i| -((i % 10 + 1) as i64)).collect()
+ } else {
+ (0..size).map(|i| (i % 10 + 1) as i64).collect()
+ };
+ let n_array = Arc::new(Int64Array::from(n_values));
+
+ vec![
+ string_arg,
+ ColumnarValue::Array(Arc::clone(&n_array) as ArrayRef),
+ ]
+}
+
+fn criterion_benchmark(c: &mut Criterion) {
+ for is_string_view in [false, true] {
+ for size in [1024, 4096] {
+ let mut group = c.benchmark_group(format!("right size={size}"));
+
+ // Benchmark with positive n (no optimization needed)
+ let mut function_name = if is_string_view {
+ "string_view_array positive n"
+ } else {
+ "string_array positive n"
+ };
+ let args = create_args(size, 32, false, is_string_view);
+ group.bench_function(BenchmarkId::new(function_name, size), |b| {
+ let arg_fields = args
+ .iter()
+ .enumerate()
+ .map(|(idx, arg)| {
+ Field::new(format!("arg_{idx}"), arg.data_type(),
true).into()
+ })
+ .collect::<Vec<_>>();
+ let config_options = Arc::new(ConfigOptions::default());
+
+ b.iter(|| {
+ black_box(
+ right()
+ .invoke_with_args(ScalarFunctionArgs {
+ args: args.clone(),
+ arg_fields: arg_fields.clone(),
+ number_rows: size,
+ return_field: Field::new("f", DataType::Utf8,
true)
Review Comment:
Shouldn't this use `Utf8View` when `is_string_view == true` ?
--
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]
