Skip to content

perf: Improve performance of CAST from string to int #3017

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
andygrove merged 7 commits into from
Jan 6, 2026
Merged

perf: Improve performance of CAST from string to int #3017

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
42a6d64
perf_string_to_int
coderfender Dec 31, 2025
768a017
perf_string_to_int
coderfender Dec 31, 2025
ca90587
perf_string_to_int
coderfender Jan 1, 2026
2b331f0
perf_string_to_int
coderfender Jan 1, 2026
5abce5f
improve_benchmark
coderfender Jan 4, 2026
94d0f32
improve_benchmark_remove_unwanted_branching
coderfender Jan 5, 2026
f1299eb
improve_benchmark_remove_unwanted_branching_per_eval_mode
coderfender Jan 6, 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
286 changes: 217 additions & 69 deletions native/spark-expr/src/conversion_funcs/cast.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -54,8 +54,8 @@ use datafusion::common::{
use datafusion::physical_expr::PhysicalExpr;
use datafusion::physical_plan::ColumnarValue;
use num::{
cast::AsPrimitive, integer::div_floor, traits::CheckedNeg, CheckedSub, Integer, Num,
ToPrimitive, Zero,
cast::AsPrimitive, integer::div_floor, traits::CheckedNeg, CheckedSub, Integer, ToPrimitive,
Zero,
};
use regex::Regex;
use std::str::FromStr;
Expand Down Expand Up @@ -389,13 +389,23 @@ macro_rules! cast_utf8_to_int {
($array:expr, $eval_mode:expr, $array_type:ty, $cast_method:ident) => {{
let len = $array.len();
let mut cast_array = PrimitiveArray::<$array_type>::builder(len);
for i in 0..len {
if $array.is_null(i) {
cast_array.append_null()
} else if let Some(cast_value) = $cast_method($array.value(i), $eval_mode)? {
cast_array.append_value(cast_value);
} else {
cast_array.append_null()
if $array.null_count() == 0 {
for i in 0..len {
if let Some(cast_value) = $cast_method($array.value(i), $eval_mode)? {
cast_array.append_value(cast_value);
} else {
cast_array.append_null()
}
}
} else {
for i in 0..len {
if $array.is_null(i) {
cast_array.append_null()
} else if let Some(cast_value) = $cast_method($array.value(i), $eval_mode)? {
cast_array.append_value(cast_value);
} else {
cast_array.append_null()
}
Copy link
Contributor Author

@coderfender coderfender Jan 3, 2026

Choose a reason for hiding this comment

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

made null check conditional to remove unwanted branching

}
}
let result: SparkResult<ArrayRef> = Ok(Arc::new(cast_array.finish()) as ArrayRef);
Expand Down Expand Up @@ -1954,100 +1964,247 @@ fn cast_string_to_int_with_range_check(
}
}

// Returns (start, end) indices after trimming whitespace
fn trim_whitespace(bytes: &[u8]) -> (usize, usize) {
let mut start = 0;
let mut end = bytes.len();

while start < end && bytes[start].is_ascii_whitespace() {
start += 1;
}
while end > start && bytes[end - 1].is_ascii_whitespace() {
end -= 1;
}

(start, end)
}

// Parses sign and returns (is_negative, start_idx after sign)
// Returns None if invalid (e.g., just "+" or "-")
fn parse_sign(trimmed_bytes: &[u8]) -> Option<(bool, usize)> {
let len = trimmed_bytes.len();
if len == 0 {
return None;
}

let first_char = trimmed_bytes[0];
let negative = first_char == b'-';

if negative || first_char == b'+' {
if len == 1 {
return None;
}
Some((negative, 1))
} else {
Some((false, 0))
}
}

/// Equivalent to
/// - org.apache.spark.unsafe.types.UTF8String.toInt(IntWrapper intWrapper, boolean allowDecimal)
/// - org.apache.spark.unsafe.types.UTF8String.toLong(LongWrapper longWrapper, boolean allowDecimal)
fn do_cast_string_to_int<
T: Num + PartialOrd + Integer + CheckedSub + CheckedNeg + From<i32> + Copy,
>(
fn do_parse_string_to_int_legacy<T: Integer + CheckedSub + CheckedNeg + From<u8> + Copy>(
str: &str,
eval_mode: EvalMode,
type_name: &str,
min_value: T,
) -> SparkResult<Option<T>> {
let trimmed_str = str.trim();
if trimmed_str.is_empty() {
return none_or_err(eval_mode, type_name, str);
let bytes = str.as_bytes();
let (start, end) = trim_whitespace(bytes);

if start == end {
return Ok(None);
}
let len = trimmed_str.len();
let trimmed_bytes = &bytes[start..end];

let (negative, idx) = match parse_sign(trimmed_bytes) {
Some(result) => result,
None => return Ok(None),
};

let mut result: T = T::zero();
let mut negative = false;
let radix = T::from(10);

let radix = T::from(10_u8);
let stop_value = min_value / radix;
let mut parse_sign_and_digits = true;

for (i, ch) in trimmed_str.char_indices() {
for &ch in &trimmed_bytes[idx..] {
Copy link
Contributor Author

@coderfender coderfender Jan 3, 2026

Choose a reason for hiding this comment

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

Cleaner and faster approach to access the chars directly

if parse_sign_and_digits {
if i == 0 {
negative = ch == '-';
let positive = ch == '+';
if negative || positive {
if i + 1 == len {
// input string is just "+" or "-"
return none_or_err(eval_mode, type_name, str);
}
// consume this char
continue;
}
if ch == b'.' {
// truncate decimal in legacy mode
parse_sign_and_digits = false;
continue;
}

if ch == '.' {
if eval_mode == EvalMode::Legacy {
// truncate decimal in legacy mode
parse_sign_and_digits = false;
continue;
} else {
return none_or_err(eval_mode, type_name, str);
}
if !ch.is_ascii_digit() {
return Ok(None);
}

let digit = if ch.is_ascii_digit() {
(ch as u32) - ('0' as u32)
} else {
return none_or_err(eval_mode, type_name, str);
};
let digit: T = T::from(ch - b'0');

// We are going to process the new digit and accumulate the result. However, before
// doing this, if the result is already smaller than the
// stopValue(Integer.MIN_VALUE / radix), then result * 10 will definitely be
// smaller than minValue, and we can stop
if result < stop_value {
return none_or_err(eval_mode, type_name, str);
return Ok(None);
}

// Since the previous result is greater than or equal to stopValue(Integer.MIN_VALUE /
// radix), we can just use `result > 0` to check overflow. If result
// overflows, we should stop
let v = result * radix;
let digit = (digit as i32).into();
match v.checked_sub(&digit) {
Some(x) if x <= T::zero() => result = x,
_ => {
return none_or_err(eval_mode, type_name, str);
return Ok(None);
}
}
} else {
// make sure fractional digits are valid digits but ignore them
// in legacy mode we still process chars after the dot and make sure the chars are digits
if !ch.is_ascii_digit() {
return none_or_err(eval_mode, type_name, str);
return Ok(None);
}
}
}

if !negative {
if let Some(neg) = result.checked_neg() {
if neg < T::zero() {
return Ok(None);
}
result = neg;
} else {
return Ok(None);
}
}

Ok(Some(result))
}

fn do_parse_string_to_int_ansi<T: Integer + CheckedSub + CheckedNeg + From<u8> + Copy>(
str: &str,
type_name: &str,
min_value: T,
) -> SparkResult<Option<T>> {
let bytes = str.as_bytes();
let (start, end) = trim_whitespace(bytes);

if start == end {
return Err(invalid_value(str, "STRING", type_name));
}
let trimmed_bytes = &bytes[start..end];

let (negative, idx) = match parse_sign(trimmed_bytes) {
Some(result) => result,
None => return Err(invalid_value(str, "STRING", type_name)),
};

let mut result: T = T::zero();

let radix = T::from(10_u8);
let stop_value = min_value / radix;

for &ch in &trimmed_bytes[idx..] {
if ch == b'.' {
return Err(invalid_value(str, "STRING", type_name));
}

if !ch.is_ascii_digit() {
return Err(invalid_value(str, "STRING", type_name));
}

let digit: T = T::from(ch - b'0');

if result < stop_value {
return Err(invalid_value(str, "STRING", type_name));
}
let v = result * radix;
match v.checked_sub(&digit) {
Some(x) if x <= T::zero() => result = x,
_ => {
return Err(invalid_value(str, "STRING", type_name));
}
}
}

if !negative {
if let Some(neg) = result.checked_neg() {
if neg < T::zero() {
return Err(invalid_value(str, "STRING", type_name));
}
result = neg;
} else {
return Err(invalid_value(str, "STRING", type_name));
}
}

Ok(Some(result))
}

fn do_parse_string_to_int_try<T: Integer + CheckedSub + CheckedNeg + From<u8> + Copy>(
str: &str,
min_value: T,
) -> SparkResult<Option<T>> {
let bytes = str.as_bytes();
let (start, end) = trim_whitespace(bytes);

if start == end {
return Ok(None);
}
let trimmed_bytes = &bytes[start..end];

let (negative, idx) = match parse_sign(trimmed_bytes) {
Some(result) => result,
None => return Ok(None),
};

let mut result: T = T::zero();

let radix = T::from(10_u8);
let stop_value = min_value / radix;

// we don't have to go beyond decimal point in try eval mode - early return NULL
for &ch in &trimmed_bytes[idx..] {
if ch == b'.' {
return Ok(None);
}

if !ch.is_ascii_digit() {
return Ok(None);
}

let digit: T = T::from(ch - b'0');

if result < stop_value {
return Ok(None);
}
let v = result * radix;
match v.checked_sub(&digit) {
Some(x) if x <= T::zero() => result = x,
_ => {
return Ok(None);
}
}
}

if !negative {
if let Some(neg) = result.checked_neg() {
if neg < T::zero() {
return none_or_err(eval_mode, type_name, str);
return Ok(None);
}
result = neg;
} else {
return none_or_err(eval_mode, type_name, str);
return Ok(None);
}
}

Ok(Some(result))
}

fn do_cast_string_to_int<T: Integer + CheckedSub + CheckedNeg + From<u8> + Copy>(
str: &str,
eval_mode: EvalMode,
type_name: &str,
min_value: T,
) -> SparkResult<Option<T>> {
match eval_mode {
Copy link
Member

@andygrove andygrove Jan 6, 2026

Choose a reason for hiding this comment

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

👍

EvalMode::Legacy => do_parse_string_to_int_legacy(str, min_value),
EvalMode::Ansi => do_parse_string_to_int_ansi(str, type_name, min_value),
EvalMode::Try => do_parse_string_to_int_try(str, min_value),
}
}

fn cast_string_to_decimal(
array: &ArrayRef,
to_type: &DataType,
Expand Down Expand Up @@ -2348,15 +2505,6 @@ fn parse_decimal_str(s: &str) -> Result<(i128, i32), String> {
Ok((final_mantissa, final_scale))
}

/// Either return Ok(None) or Err(SparkError::CastInvalidValue) depending on the evaluation mode
#[inline]
fn none_or_err<T>(eval_mode: EvalMode, type_name: &str, str: &str) -> SparkResult<Option<T>> {
match eval_mode {
EvalMode::Ansi => Err(invalid_value(str, "STRING", type_name)),
_ => Ok(None),
}
}

#[inline]
fn invalid_value(value: &str, from_type: &str, to_type: &str) -> SparkError {
SparkError::CastInvalidValue {
Expand Down
Loading