This is an automated email from the ASF dual-hosted git repository.
alamb pushed a commit to branch cherry_pick_ee1d1644
in repository https://gitbox.apache.org/repos/asf/arrow-rs.git
commit 1859aa0a381224e1c794b0ce8575b3eba293f472
Author: Jordan Deitch <jwdei...@users.noreply.github.com>
AuthorDate: Tue Nov 9 08:56:18 2021 -0500
add ilike comparitor (#874)
* add ilike comparitor
* add ilike comparitor
Co-authored-by: Jordan Deitch <jdei...@digitalocean.com>
---
arrow/benches/comparison_kernels.rs | 25 +++
arrow/src/compute/kernels/comparison.rs | 270 +++++++++++++++++++++++---------
2 files changed, 217 insertions(+), 78 deletions(-)
diff --git a/arrow/benches/comparison_kernels.rs
b/arrow/benches/comparison_kernels.rs
index bfee9b9..94ff7df 100644
--- a/arrow/benches/comparison_kernels.rs
+++ b/arrow/benches/comparison_kernels.rs
@@ -119,6 +119,11 @@ fn bench_nlike_utf8_scalar(arr_a: &StringArray, value_b:
&str) {
.unwrap();
}
+fn bench_ilike_utf8_scalar(arr_a: &StringArray, value_b: &str) {
+ ilike_utf8_scalar(criterion::black_box(arr_a),
criterion::black_box(value_b))
+ .unwrap();
+}
+
fn bench_regexp_is_match_utf8_scalar(arr_a: &StringArray, value_b: &str) {
regexp_is_match_utf8_scalar(
criterion::black_box(arr_a),
@@ -205,6 +210,26 @@ fn add_benchmark(c: &mut Criterion) {
b.iter(|| bench_nlike_utf8_scalar(&arr_string, "%xx_xx%xxx"))
});
+ c.bench_function("ilike_utf8 scalar equals", |b| {
+ b.iter(|| bench_ilike_utf8_scalar(&arr_string, "xxXX"))
+ });
+
+ c.bench_function("ilike_utf8 scalar contains", |b| {
+ b.iter(|| bench_ilike_utf8_scalar(&arr_string, "%xxXX%"))
+ });
+
+ c.bench_function("ilike_utf8 scalar ends with", |b| {
+ b.iter(|| bench_ilike_utf8_scalar(&arr_string, "xXXx%"))
+ });
+
+ c.bench_function("ilike_utf8 scalar starts with", |b| {
+ b.iter(|| bench_ilike_utf8_scalar(&arr_string, "%XXXx"))
+ });
+
+ c.bench_function("ilike_utf8 scalar complex", |b| {
+ b.iter(|| bench_ilike_utf8_scalar(&arr_string, "%xx_xX%xXX"))
+ });
+
c.bench_function("egexp_matches_utf8 scalar starts with", |b| {
b.iter(|| bench_regexp_is_match_utf8_scalar(&arr_string, "^xx"))
});
diff --git a/arrow/src/compute/kernels/comparison.rs
b/arrow/src/compute/kernels/comparison.rs
index 9d49e89..3b65f33 100644
--- a/arrow/src/compute/kernels/comparison.rs
+++ b/arrow/src/compute/kernels/comparison.rs
@@ -228,28 +228,23 @@ where
compare_op_scalar_primitive!(left, right, op)
}
-/// Perform SQL `left LIKE right` operation on [`StringArray`] /
[`LargeStringArray`].
-///
-/// There are two wildcards supported with the LIKE operator:
-///
-/// 1. `%` - The percent sign represents zero, one, or multiple characters
-/// 2. `_` - The underscore represents a single character
-///
-/// For example:
-/// ```
-/// use arrow::array::{StringArray, BooleanArray};
-/// use arrow::compute::like_utf8;
-///
-/// let strings = StringArray::from(vec!["Arrow", "Arrow", "Arrow", "Ar"]);
-/// let patterns = StringArray::from(vec!["A%", "B%", "A.", "A."]);
+fn is_like_pattern(c: char) -> bool {
+ c == '%' || c == '_'
+}
+
+/// Evaluate regex `op(left)` matching `right` on [`StringArray`] /
[`LargeStringArray`]
///
-/// let result = like_utf8(&strings, &patterns).unwrap();
-/// assert_eq!(result, BooleanArray::from(vec![true, false, false, true]));
-/// ```
-pub fn like_utf8<OffsetSize: StringOffsetSizeTrait>(
+/// If `negate_regex` is true, the regex expression will be negated. (for
example, with `not like`)
+fn regex_like<OffsetSize, F>(
left: &GenericStringArray<OffsetSize>,
right: &GenericStringArray<OffsetSize>,
-) -> Result<BooleanArray> {
+ negate_regex: bool,
+ op: F,
+) -> Result<BooleanArray>
+where
+ OffsetSize: StringOffsetSizeTrait,
+ F: Fn(&str) -> Result<Regex>,
+{
let mut map = HashMap::new();
if left.len() != right.len() {
return Err(ArrowError::ComputeError(
@@ -269,17 +264,16 @@ pub fn like_utf8<OffsetSize: StringOffsetSizeTrait>(
regex
} else {
let re_pattern = pat.replace("%", ".*").replace("_", ".");
- let re = Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
- ArrowError::ComputeError(format!(
- "Unable to build regex from LIKE pattern: {}",
- e
- ))
- })?;
+ let re = op(&re_pattern)?;
map.insert(pat, re);
map.get(pat).unwrap()
};
- result.append(re.is_match(haystack));
+ result.append(if negate_regex {
+ !re.is_match(haystack)
+ } else {
+ re.is_match(haystack)
+ });
}
let data = unsafe {
@@ -296,8 +290,36 @@ pub fn like_utf8<OffsetSize: StringOffsetSizeTrait>(
Ok(BooleanArray::from(data))
}
-fn is_like_pattern(c: char) -> bool {
- c == '%' || c == '_'
+/// Perform SQL `left LIKE right` operation on [`StringArray`] /
[`LargeStringArray`].
+///
+/// There are two wildcards supported with the LIKE operator:
+///
+/// 1. `%` - The percent sign represents zero, one, or multiple characters
+/// 2. `_` - The underscore represents a single character
+///
+/// For example:
+/// ```
+/// use arrow::array::{StringArray, BooleanArray};
+/// use arrow::compute::like_utf8;
+///
+/// let strings = StringArray::from(vec!["Arrow", "Arrow", "Arrow", "Ar"]);
+/// let patterns = StringArray::from(vec!["A%", "B%", "A.", "A."]);
+///
+/// let result = like_utf8(&strings, &patterns).unwrap();
+/// assert_eq!(result, BooleanArray::from(vec![true, false, false, true]));
+/// ```
+pub fn like_utf8<OffsetSize: StringOffsetSizeTrait>(
+ left: &GenericStringArray<OffsetSize>,
+ right: &GenericStringArray<OffsetSize>,
+) -> Result<BooleanArray> {
+ regex_like(left, right, false, |re_pattern| {
+ Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
+ ArrowError::ComputeError(format!(
+ "Unable to build regex from LIKE pattern: {}",
+ e
+ ))
+ })
+ })
}
/// Perform SQL `left LIKE right` operation on [`StringArray`] /
@@ -376,36 +398,55 @@ pub fn nlike_utf8<OffsetSize: StringOffsetSizeTrait>(
left: &GenericStringArray<OffsetSize>,
right: &GenericStringArray<OffsetSize>,
) -> Result<BooleanArray> {
- let mut map = HashMap::new();
- if left.len() != right.len() {
- return Err(ArrowError::ComputeError(
- "Cannot perform comparison operation on arrays of different length"
- .to_string(),
- ));
- }
-
- let null_bit_buffer =
- combine_option_bitmap(left.data_ref(), right.data_ref(), left.len())?;
+ regex_like(left, right, true, |re_pattern| {
+ Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
+ ArrowError::ComputeError(format!(
+ "Unable to build regex from LIKE pattern: {}",
+ e
+ ))
+ })
+ })
+}
+/// Perform SQL `left NOT LIKE right` operation on [`StringArray`] /
+/// [`LargeStringArray`] and a scalar.
+///
+/// See the documentation on [`like_utf8`] for more details.
+pub fn nlike_utf8_scalar<OffsetSize: StringOffsetSizeTrait>(
+ left: &GenericStringArray<OffsetSize>,
+ right: &str,
+) -> Result<BooleanArray> {
+ let null_bit_buffer = left.data().null_buffer().cloned();
let mut result = BooleanBufferBuilder::new(left.len());
- for i in 0..left.len() {
- let haystack = left.value(i);
- let pat = right.value(i);
- let re = if let Some(ref regex) = map.get(pat) {
- regex
- } else {
- let re_pattern = pat.replace("%", ".*").replace("_", ".");
- let re = Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
- ArrowError::ComputeError(format!(
- "Unable to build regex from LIKE pattern: {}",
- e
- ))
- })?;
- map.insert(pat, re);
- map.get(pat).unwrap()
- };
- result.append(!re.is_match(haystack));
+ if !right.contains(is_like_pattern) {
+ // fast path, can use equals
+ for i in 0..left.len() {
+ result.append(left.value(i) != right);
+ }
+ } else if right.ends_with('%') && !right[..right.len() -
1].contains(is_like_pattern)
+ {
+ // fast path, can use ends_with
+ for i in 0..left.len() {
+ result.append(!left.value(i).starts_with(&right[..right.len() -
1]));
+ }
+ } else if right.starts_with('%') && !right[1..].contains(is_like_pattern) {
+ // fast path, can use starts_with
+ for i in 0..left.len() {
+ result.append(!left.value(i).ends_with(&right[1..]));
+ }
+ } else {
+ let re_pattern = right.replace("%", ".*").replace("_", ".");
+ let re = Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
+ ArrowError::ComputeError(format!(
+ "Unable to build regex from LIKE pattern: {}",
+ e
+ ))
+ })?;
+ for i in 0..left.len() {
+ let haystack = left.value(i);
+ result.append(!re.is_match(haystack));
+ }
}
let data = unsafe {
@@ -422,11 +463,29 @@ pub fn nlike_utf8<OffsetSize: StringOffsetSizeTrait>(
Ok(BooleanArray::from(data))
}
-/// Perform SQL `left NOT LIKE right` operation on [`StringArray`] /
+/// Perform SQL `left ILIKE right` operation on [`StringArray`] /
+/// [`LargeStringArray`].
+///
+/// See the documentation on [`like_utf8`] for more details.
+pub fn ilike_utf8<OffsetSize: StringOffsetSizeTrait>(
+ left: &GenericStringArray<OffsetSize>,
+ right: &GenericStringArray<OffsetSize>,
+) -> Result<BooleanArray> {
+ regex_like(left, right, false, |re_pattern| {
+ Regex::new(&format!("(?i)^{}$", re_pattern)).map_err(|e| {
+ ArrowError::ComputeError(format!(
+ "Unable to build regex from ILIKE pattern: {}",
+ e
+ ))
+ })
+ })
+}
+
+/// Perform SQL `left ILIKE right` operation on [`StringArray`] /
/// [`LargeStringArray`] and a scalar.
///
/// See the documentation on [`like_utf8`] for more details.
-pub fn nlike_utf8_scalar<OffsetSize: StringOffsetSizeTrait>(
+pub fn ilike_utf8_scalar<OffsetSize: StringOffsetSizeTrait>(
left: &GenericStringArray<OffsetSize>,
right: &str,
) -> Result<BooleanArray> {
@@ -436,30 +495,38 @@ pub fn nlike_utf8_scalar<OffsetSize:
StringOffsetSizeTrait>(
if !right.contains(is_like_pattern) {
// fast path, can use equals
for i in 0..left.len() {
- result.append(left.value(i) != right);
+ result.append(left.value(i) == right);
}
} else if right.ends_with('%') && !right[..right.len() -
1].contains(is_like_pattern)
{
// fast path, can use ends_with
for i in 0..left.len() {
- result.append(!left.value(i).starts_with(&right[..right.len() -
1]));
+ result.append(
+ left.value(i)
+ .to_uppercase()
+ .starts_with(&right[..right.len() - 1].to_uppercase()),
+ );
}
} else if right.starts_with('%') && !right[1..].contains(is_like_pattern) {
// fast path, can use starts_with
for i in 0..left.len() {
- result.append(!left.value(i).ends_with(&right[1..]));
+ result.append(
+ left.value(i)
+ .to_uppercase()
+ .ends_with(&right[1..].to_uppercase()),
+ );
}
} else {
let re_pattern = right.replace("%", ".*").replace("_", ".");
- let re = Regex::new(&format!("^{}$", re_pattern)).map_err(|e| {
+ let re = Regex::new(&format!("(?i)^{}$", re_pattern)).map_err(|e| {
ArrowError::ComputeError(format!(
- "Unable to build regex from LIKE pattern: {}",
+ "Unable to build regex from ILIKE pattern: {}",
e
))
})?;
for i in 0..left.len() {
let haystack = left.value(i);
- result.append(!re.is_match(haystack));
+ result.append(re.is_match(haystack));
}
}
@@ -2128,21 +2195,6 @@ mod tests {
);
test_utf8!(
- test_utf8_array_nlike,
- vec!["arrow", "arrow", "arrow", "arrow", "arrow", "arrows", "arrow"],
- vec!["arrow", "ar%", "%ro%", "foo", "arr", "arrow_", "arrow_"],
- nlike_utf8,
- vec![false, false, false, true, true, false, true]
- );
- test_utf8_scalar!(
- test_utf8_array_nlike_scalar,
- vec!["arrow", "parquet", "datafusion", "flight"],
- "%ar%",
- nlike_utf8_scalar,
- vec![false, false, true, true]
- );
-
- test_utf8!(
test_utf8_array_eq,
vec!["arrow", "arrow", "arrow", "arrow"],
vec!["arrow", "parquet", "datafusion", "flight"],
@@ -2157,6 +2209,21 @@ mod tests {
vec![true, false, false, false]
);
+ test_utf8!(
+ test_utf8_array_nlike,
+ vec!["arrow", "arrow", "arrow", "arrow", "arrow", "arrows", "arrow"],
+ vec!["arrow", "ar%", "%ro%", "foo", "arr", "arrow_", "arrow_"],
+ nlike_utf8,
+ vec![false, false, false, true, true, false, true]
+ );
+ test_utf8_scalar!(
+ test_utf8_array_nlike_scalar,
+ vec!["arrow", "parquet", "datafusion", "flight"],
+ "%ar%",
+ nlike_utf8_scalar,
+ vec![false, false, true, true]
+ );
+
test_utf8_scalar!(
test_utf8_array_nlike_scalar_start,
vec!["arrow", "parrow", "arrows", "arr"],
@@ -2190,6 +2257,53 @@ mod tests {
);
test_utf8!(
+ test_utf8_array_ilike,
+ vec!["arrow", "arrow", "ARROW", "arrow", "ARROW", "ARROWS", "arROw"],
+ vec!["arrow", "ar%", "%ro%", "foo", "ar%r", "arrow_", "arrow_"],
+ ilike_utf8,
+ vec![true, true, true, false, false, true, false]
+ );
+ test_utf8_scalar!(
+ test_utf8_array_ilike_scalar,
+ vec!["arrow", "parquet", "datafusion", "flight"],
+ "%AR%",
+ ilike_utf8_scalar,
+ vec![true, true, false, false]
+ );
+
+ test_utf8_scalar!(
+ test_utf8_array_ilike_scalar_start,
+ vec!["arrow", "parrow", "arrows", "ARR"],
+ "aRRow%",
+ ilike_utf8_scalar,
+ vec![true, false, true, false]
+ );
+
+ test_utf8_scalar!(
+ test_utf8_array_ilike_scalar_end,
+ vec!["ArroW", "parrow", "ARRowS", "arr"],
+ "%arrow",
+ ilike_utf8_scalar,
+ vec![true, true, false, false]
+ );
+
+ test_utf8_scalar!(
+ test_utf8_array_ilike_scalar_equals,
+ vec!["arrow", "parrow", "arrows", "arr"],
+ "arrow",
+ ilike_utf8_scalar,
+ vec![true, false, false, false]
+ );
+
+ test_utf8_scalar!(
+ test_utf8_array_ilike_scalar_one,
+ vec!["arrow", "arrows", "parrow", "arr"],
+ "arrow_",
+ ilike_utf8_scalar,
+ vec![false, true, false, false]
+ );
+
+ test_utf8!(
test_utf8_array_neq,
vec!["arrow", "arrow", "arrow", "arrow"],
vec!["arrow", "parquet", "datafusion", "flight"],
