1use arrow::array::{Array, ArrayRef, AsArray, GenericStringArray, StringArrayType};
19use arrow::buffer::NullBuffer;
20use arrow::datatypes::DataType;
21use datafusion_common::HashMap;
22
23use super::common::try_as_scalar_str;
24use crate::strings::{
25 BulkNullStringArrayBuilder, GenericStringArrayBuilder, StringViewArrayBuilder,
26 StringWriter,
27};
28use crate::utils::make_scalar_function;
29use datafusion_common::{Result, exec_err};
30use datafusion_expr::TypeSignature::Exact;
31use datafusion_expr::{
32 ColumnarValue, Documentation, ScalarFunctionArgs, ScalarUDFImpl, Signature,
33 Volatility,
34};
35use datafusion_macros::user_doc;
36
37#[user_doc(
38 doc_section(label = "String Functions"),
39 description = "Performs character-wise substitution based on a mapping.",
40 syntax_example = "translate(str, from, to)",
41 sql_example = r#"```sql
42> select translate('twice', 'wic', 'her');
43+--------------------------------------------------+
44| translate(Utf8("twice"),Utf8("wic"),Utf8("her")) |
45+--------------------------------------------------+
46| there |
47+--------------------------------------------------+
48```"#,
49 standard_argument(name = "str", prefix = "String"),
50 argument(name = "from", description = "The characters to be replaced."),
51 argument(
52 name = "to",
53 description = "The characters to replace them with. Each character in **from** that is found in **str** is replaced by the character at the same index in **to**. Any characters in **from** that don't have a corresponding character in **to** are removed. If a character appears more than once in **from**, the first occurrence determines the mapping."
54 )
55)]
56#[derive(Debug, PartialEq, Eq, Hash)]
57pub struct TranslateFunc {
58 signature: Signature,
59}
60
61impl Default for TranslateFunc {
62 fn default() -> Self {
63 Self::new()
64 }
65}
66
67impl TranslateFunc {
68 pub fn new() -> Self {
69 use DataType::*;
70 Self {
71 signature: Signature::one_of(
72 vec![
73 Exact(vec![Utf8View, Utf8, Utf8]),
74 Exact(vec![Utf8, Utf8, Utf8]),
75 Exact(vec![LargeUtf8, Utf8, Utf8]),
76 ],
77 Volatility::Immutable,
78 ),
79 }
80 }
81}
82
83impl ScalarUDFImpl for TranslateFunc {
84 fn name(&self) -> &str {
85 "translate"
86 }
87
88 fn signature(&self) -> &Signature {
89 &self.signature
90 }
91
92 fn return_type(&self, arg_types: &[DataType]) -> Result<DataType> {
93 Ok(arg_types[0].clone())
94 }
95
96 fn invoke_with_args(&self, args: ScalarFunctionArgs) -> Result<ColumnarValue> {
97 if let (Some(from_str), Some(to_str)) = (
99 try_as_scalar_str(&args.args[1]),
100 try_as_scalar_str(&args.args[2]),
101 ) {
102 let table = build_translate_table(from_str, to_str);
103
104 let string_array = args.args[0].to_array_of_size(args.number_rows)?;
105 let len = string_array.len();
106
107 let result = match string_array.data_type() {
108 DataType::Utf8View => {
109 let arr = string_array.as_string_view();
110 let builder = StringViewArrayBuilder::with_capacity(len);
111 translate_with_table(&arr, &table, builder)
112 }
113 DataType::Utf8 => {
114 let arr = string_array.as_string::<i32>();
115 let builder = GenericStringArrayBuilder::<i32>::with_capacity(
116 len,
117 arr.value_data().len(),
118 );
119 translate_with_table(&arr, &table, builder)
120 }
121 DataType::LargeUtf8 => {
122 let arr = string_array.as_string::<i64>();
123 let builder = GenericStringArrayBuilder::<i64>::with_capacity(
124 len,
125 arr.value_data().len(),
126 );
127 translate_with_table(&arr, &table, builder)
128 }
129 other => {
130 return exec_err!(
131 "Unsupported data type {other:?} for function translate"
132 );
133 }
134 }?;
135
136 return Ok(ColumnarValue::Array(result));
137 }
138
139 make_scalar_function(invoke_translate, vec![])(&args.args)
140 }
141
142 fn documentation(&self) -> Option<&Documentation> {
143 self.doc()
144 }
145}
146
147fn invoke_translate(args: &[ArrayRef]) -> Result<ArrayRef> {
148 let len = args[0].len();
149 match args[0].data_type() {
150 DataType::Utf8View => {
151 let string_array = args[0].as_string_view();
152 let from_array = args[1].as_string::<i32>();
153 let to_array = args[2].as_string::<i32>();
154 let builder = StringViewArrayBuilder::with_capacity(len);
155 translate(&string_array, from_array, to_array, builder)
156 }
157 DataType::Utf8 => {
158 let string_array = args[0].as_string::<i32>();
159 let from_array = args[1].as_string::<i32>();
160 let to_array = args[2].as_string::<i32>();
161 let builder = GenericStringArrayBuilder::<i32>::with_capacity(
162 len,
163 string_array.value_data().len(),
164 );
165 translate(&string_array, from_array, to_array, builder)
166 }
167 DataType::LargeUtf8 => {
168 let string_array = args[0].as_string::<i64>();
169 let from_array = args[1].as_string::<i32>();
170 let to_array = args[2].as_string::<i32>();
171 let builder = GenericStringArrayBuilder::<i64>::with_capacity(
172 len,
173 string_array.value_data().len(),
174 );
175 translate(&string_array, from_array, to_array, builder)
176 }
177 other => {
178 exec_err!("Unsupported data type {other:?} for function translate")
179 }
180 }
181}
182
183fn translate<'a, S, O>(
189 string_array: &S,
190 from_array: &GenericStringArray<i32>,
191 to_array: &GenericStringArray<i32>,
192 mut builder: O,
193) -> Result<ArrayRef>
194where
195 S: StringArrayType<'a>,
196 O: BulkNullStringArrayBuilder,
197{
198 let mut from_map: HashMap<char, Option<char>> = HashMap::new();
199 let len = string_array.len();
200 let nulls = NullBuffer::union_many([
201 string_array.nulls(),
202 from_array.nulls(),
203 to_array.nulls(),
204 ]);
205
206 if let Some(nulls_ref) = nulls.as_ref() {
207 for i in 0..len {
208 if nulls_ref.is_null(i) {
209 builder.append_placeholder();
210 continue;
211 }
212
213 let string = unsafe { string_array.value_unchecked(i) };
215 let from = unsafe { from_array.value_unchecked(i) };
216 let to = unsafe { to_array.value_unchecked(i) };
217 append_translated_row(&mut builder, string, from, to, &mut from_map);
218 }
219 } else {
220 for i in 0..len {
221 let string = unsafe { string_array.value_unchecked(i) };
223 let from = unsafe { from_array.value_unchecked(i) };
224 let to = unsafe { to_array.value_unchecked(i) };
225 append_translated_row(&mut builder, string, from, to, &mut from_map);
226 }
227 }
228
229 builder.finish(nulls)
230}
231
232#[inline]
233fn append_translated_row<B: BulkNullStringArrayBuilder>(
234 builder: &mut B,
235 string: &str,
236 from: &str,
237 to: &str,
238 from_map: &mut HashMap<char, Option<char>>,
239) {
240 if let Some(ascii_table) = build_ascii_translate_table(from, to) {
241 append_translated_ascii(builder, string, &ascii_table);
242 return;
243 }
244
245 from_map.clear();
246 let mut to_iter = to.chars();
247 for c in from.chars() {
248 let replacement = to_iter.next();
249 from_map.entry(c).or_insert(replacement);
250 }
251
252 builder.append_with(|w| write_translated_chars(w, string, from_map));
253}
254
255#[inline]
256fn write_translated_chars<W: StringWriter>(
257 w: &mut W,
258 input: &str,
259 from_map: &HashMap<char, Option<char>>,
260) {
261 for c in input.chars() {
262 match from_map.get(&c) {
263 Some(Some(r)) => w.write_char(*r),
264 Some(None) => {} None => w.write_char(c),
266 }
267 }
268}
269
270const ASCII_DELETE: u8 = 0xFF;
274
275#[derive(Debug)]
280struct AsciiTranslateTable {
281 map: [u8; 256],
282 has_delete: bool,
283}
284
285#[expect(
288 clippy::large_enum_variant,
289 reason = "one instance per call, passed by reference"
290)]
291enum TranslateTable {
292 Byte(AsciiTranslateTable),
293 Char(HashMap<char, Option<char>>),
294}
295
296#[inline]
297fn build_translate_table(from: &str, to: &str) -> TranslateTable {
298 if let Some(ascii) = build_ascii_translate_table(from, to) {
299 return TranslateTable::Byte(ascii);
300 }
301 let mut from_map: HashMap<char, Option<char>> = HashMap::with_capacity(from.len());
302 let mut to_iter = to.chars();
303 for c in from.chars() {
304 let replacement = to_iter.next();
305 from_map.entry(c).or_insert(replacement);
306 }
307 TranslateTable::Char(from_map)
308}
309
310fn build_ascii_translate_table(from: &str, to: &str) -> Option<AsciiTranslateTable> {
312 if !from.is_ascii() || !to.is_ascii() {
313 return None;
314 }
315
316 let to_bytes = to.as_bytes();
317 let mut map = std::array::from_fn::<u8, 256, _>(|i| i as u8);
318 let mut seen = [false; 128];
319 let mut has_delete = false;
320
321 for (i, from_byte) in from.bytes().enumerate() {
322 let idx = from_byte as usize;
323 if !seen[idx] {
324 seen[idx] = true;
325 if i < to_bytes.len() {
326 map[idx] = to_bytes[i];
327 } else {
328 map[idx] = ASCII_DELETE;
329 has_delete = true;
330 }
331 }
332 }
333
334 Some(AsciiTranslateTable { map, has_delete })
335}
336
337#[inline]
338fn append_translated_ascii<B: BulkNullStringArrayBuilder>(
339 builder: &mut B,
340 input: &str,
341 table: &AsciiTranslateTable,
342) {
343 if !table.has_delete {
345 unsafe {
350 builder.append_byte_map(input.as_bytes(), |b| table.map[b as usize]);
351 }
352 } else {
353 builder.append_with(|w| write_translated_ascii(w, input, table));
354 }
355}
356
357#[inline]
358fn write_translated_ascii<W: StringWriter>(
359 w: &mut W,
360 input: &str,
361 table: &AsciiTranslateTable,
362) {
363 let bytes = input.as_bytes();
364 let mut copy_start = 0;
365
366 for (i, &b) in bytes.iter().enumerate() {
367 let mapped = table.map[b as usize];
368 if mapped == b {
369 continue;
370 }
371
372 if copy_start < i {
373 w.write_str(&input[copy_start..i]);
374 }
375 if mapped != ASCII_DELETE {
376 w.write_char(mapped as char);
377 }
378 copy_start = i + 1;
379 }
380
381 if copy_start < input.len() {
382 w.write_str(&input[copy_start..]);
383 }
384}
385
386fn translate_with_table<'a, S, O>(
387 string_array: &S,
388 table: &TranslateTable,
389 mut builder: O,
390) -> Result<ArrayRef>
391where
392 S: StringArrayType<'a>,
393 O: BulkNullStringArrayBuilder,
394{
395 let len = string_array.len();
396 let nulls = string_array.nulls().cloned();
397
398 if let Some(nulls_ref) = nulls.as_ref() {
399 for i in 0..len {
400 if nulls_ref.is_null(i) {
401 builder.append_placeholder();
402 continue;
403 }
404
405 let s = unsafe { string_array.value_unchecked(i) };
407 apply_translate_table(&mut builder, s, table);
408 }
409 } else {
410 for i in 0..len {
411 let s = unsafe { string_array.value_unchecked(i) };
413 apply_translate_table(&mut builder, s, table);
414 }
415 }
416
417 builder.finish(nulls)
418}
419
420#[inline]
421fn apply_translate_table<B: BulkNullStringArrayBuilder>(
422 builder: &mut B,
423 input: &str,
424 table: &TranslateTable,
425) {
426 match table {
427 TranslateTable::Byte(t) => append_translated_ascii(builder, input, t),
428 TranslateTable::Char(m) => {
429 builder.append_with(|w| write_translated_chars(w, input, m))
430 }
431 }
432}
433
434#[cfg(test)]
435mod tests {
436 use std::sync::Arc;
437
438 use arrow::array::{Array, ArrayRef, StringArray, StringViewArray};
439 use arrow::datatypes::DataType::{Utf8, Utf8View};
440
441 use datafusion_common::{Result, ScalarValue};
442 use datafusion_expr::{ColumnarValue, ScalarUDFImpl};
443
444 use crate::unicode::translate::TranslateFunc;
445 use crate::utils::test::test_function;
446
447 #[test]
448 fn test_functions() -> Result<()> {
449 test_function!(
450 TranslateFunc::new(),
451 vec![
452 ColumnarValue::Scalar(ScalarValue::from("12345")),
453 ColumnarValue::Scalar(ScalarValue::from("143")),
454 ColumnarValue::Scalar(ScalarValue::from("ax"))
455 ],
456 Ok(Some("a2x5")),
457 &str,
458 Utf8,
459 StringArray
460 );
461 test_function!(
462 TranslateFunc::new(),
463 vec![
464 ColumnarValue::Scalar(ScalarValue::Utf8(None)),
465 ColumnarValue::Scalar(ScalarValue::from("143")),
466 ColumnarValue::Scalar(ScalarValue::from("ax"))
467 ],
468 Ok(None),
469 &str,
470 Utf8,
471 StringArray
472 );
473 test_function!(
474 TranslateFunc::new(),
475 vec![
476 ColumnarValue::Scalar(ScalarValue::from("12345")),
477 ColumnarValue::Scalar(ScalarValue::Utf8(None)),
478 ColumnarValue::Scalar(ScalarValue::from("ax"))
479 ],
480 Ok(None),
481 &str,
482 Utf8,
483 StringArray
484 );
485 test_function!(
486 TranslateFunc::new(),
487 vec![
488 ColumnarValue::Scalar(ScalarValue::from("12345")),
489 ColumnarValue::Scalar(ScalarValue::from("143")),
490 ColumnarValue::Scalar(ScalarValue::Utf8(None))
491 ],
492 Ok(None),
493 &str,
494 Utf8,
495 StringArray
496 );
497 test_function!(
498 TranslateFunc::new(),
499 vec![
500 ColumnarValue::Scalar(ScalarValue::from("abcabc")),
501 ColumnarValue::Scalar(ScalarValue::from("aa")),
502 ColumnarValue::Scalar(ScalarValue::from("de"))
503 ],
504 Ok(Some("dbcdbc")),
505 &str,
506 Utf8,
507 StringArray
508 );
509 test_function!(
510 TranslateFunc::new(),
511 vec![
512 ColumnarValue::Scalar(ScalarValue::from("é2íñ5")),
513 ColumnarValue::Scalar(ScalarValue::from("éñí")),
514 ColumnarValue::Scalar(ScalarValue::from("óü")),
515 ],
516 Ok(Some("ó2ü5")),
517 &str,
518 Utf8,
519 StringArray
520 );
521 test_function!(
523 TranslateFunc::new(),
524 vec![
525 ColumnarValue::Scalar(ScalarValue::from("café")),
526 ColumnarValue::Scalar(ScalarValue::from("ae")),
527 ColumnarValue::Scalar(ScalarValue::from("AE"))
528 ],
529 Ok(Some("cAfé")),
530 &str,
531 Utf8,
532 StringArray
533 );
534 test_function!(
536 TranslateFunc::new(),
537 vec![
538 ColumnarValue::Scalar(ScalarValue::Utf8View(Some("12345".into()))),
539 ColumnarValue::Scalar(ScalarValue::from("143")),
540 ColumnarValue::Scalar(ScalarValue::from("ax"))
541 ],
542 Ok(Some("a2x5")),
543 &str,
544 Utf8View,
545 StringViewArray
546 );
547 test_function!(
549 TranslateFunc::new(),
550 vec![
551 ColumnarValue::Scalar(ScalarValue::Utf8View(None)),
552 ColumnarValue::Scalar(ScalarValue::from("143")),
553 ColumnarValue::Scalar(ScalarValue::from("ax"))
554 ],
555 Ok(None),
556 &str,
557 Utf8View,
558 StringViewArray
559 );
560 test_function!(
562 TranslateFunc::new(),
563 vec![
564 ColumnarValue::Scalar(ScalarValue::Utf8View(Some("é2íñ5".into()))),
565 ColumnarValue::Scalar(ScalarValue::from("éñí")),
566 ColumnarValue::Scalar(ScalarValue::from("óü"))
567 ],
568 Ok(Some("ó2ü5")),
569 &str,
570 Utf8View,
571 StringViewArray
572 );
573
574 #[cfg(not(feature = "unicode_expressions"))]
575 test_function!(
576 TranslateFunc::new(),
577 vec![
578 ColumnarValue::Scalar(ScalarValue::from("12345")),
579 ColumnarValue::Scalar(ScalarValue::from("143")),
580 ColumnarValue::Scalar(ScalarValue::from("ax")),
581 ],
582 internal_err!(
583 "function translate requires compilation with feature flag: unicode_expressions."
584 ),
585 &str,
586 Utf8,
587 StringArray
588 );
589
590 Ok(())
591 }
592
593 #[test]
594 fn test_array_args_with_nulls() -> Result<()> {
595 let string_array = Arc::new(StringArray::from(vec![
596 Some("café!"),
597 Some("abc"),
598 Some("abc"),
599 ])) as ArrayRef;
600 let from_array =
601 Arc::new(StringArray::from(vec![Some("!"), Some("a"), None])) as ArrayRef;
602 let to_array =
603 Arc::new(StringArray::from(vec![Some(""), Some("x"), Some("y")])) as ArrayRef;
604
605 let result = super::invoke_translate(&[string_array, from_array, to_array])?;
606 let result = result.as_any().downcast_ref::<StringArray>().unwrap();
607
608 assert_eq!(result.len(), 3);
609 assert_eq!(result.value(0), "café");
610 assert_eq!(result.value(1), "xbc");
611 assert!(result.is_null(2));
612
613 Ok(())
614 }
615}