1use super::{_MutableArrayData, ArrayData, Extend};
19use arrow_buffer::{ArrowNativeType, Buffer, ToByteSlice};
20use arrow_schema::{ArrowError, DataType};
21use num_traits::CheckedAdd;
22
23fn get_last_run_end<T: ArrowNativeType>(run_ends_data: &super::MutableArrayData) -> T {
25 if run_ends_data.data.len == 0 {
26 T::default()
27 } else {
28 let typed_slice: &[T] = run_ends_data.data.buffer1.typed_data();
29 if typed_slice.len() >= run_ends_data.data.len {
30 typed_slice[run_ends_data.data.len - 1]
31 } else {
32 T::default()
33 }
34 }
35}
36
37pub fn extend_nulls(mutable: &mut _MutableArrayData, len: usize) -> Result<(), ArrowError> {
42 if len == 0 {
43 return Ok(());
44 }
45
46 mutable.child_data[1].try_extend_nulls(1)?;
49
50 let run_end_type = if let DataType::RunEndEncoded(run_ends_field, _) = &mutable.data_type {
52 run_ends_field.data_type()
53 } else {
54 panic!("extend_nulls called on non-RunEndEncoded array");
55 };
56
57 macro_rules! extend_nulls_impl {
59 ($run_end_type:ty) => {{
60 let last_run_end = get_last_run_end::<$run_end_type>(&mutable.child_data[0]);
61 let new_value = last_run_end
62 .checked_add(<$run_end_type as ArrowNativeType>::usize_as(len))
63 .ok_or_else(|| {
64 ArrowError::InvalidArgumentError(
65 "run end overflow when extending RunEndEncoded array: \
66 use a larger run-end type (e.g. Int64 instead of Int32)"
67 .to_string(),
68 )
69 })?;
70 mutable.child_data[0]
71 .data
72 .buffer1
73 .try_extend_from_slice(new_value.to_byte_slice())
74 .map_err(|e| ArrowError::MemoryError(e.to_string()))?;
75 }};
76 }
77
78 match run_end_type {
80 DataType::Int16 => extend_nulls_impl!(i16),
81 DataType::Int32 => extend_nulls_impl!(i32),
82 DataType::Int64 => extend_nulls_impl!(i64),
83 _ => panic!("Invalid run end type for RunEndEncoded array: {run_end_type}"),
84 };
85
86 mutable.child_data[0].data.len += 1;
87 Ok(())
88}
89
90type ExtendArrays = (Vec<u8>, Option<(usize, usize)>);
92
93fn build_extend_arrays<T: ArrowNativeType + std::ops::Add<Output = T> + CheckedAdd>(
95 buffer: &Buffer,
96 length: usize,
97 start: usize,
98 len: usize,
99 dest_last_run_end: T,
100) -> Result<ExtendArrays, ArrowError> {
101 let mut run_ends_bytes = Vec::new();
102 let mut values_range: Option<(usize, usize)> = None;
103 let end = start + len;
104 let mut prev_end = 0;
105 let mut current_run_end = dest_last_run_end;
106
107 let typed_slice: &[T] = buffer.typed_data();
109
110 for i in 0..length {
111 if i < typed_slice.len() {
112 let run_end = typed_slice[i].to_usize().unwrap();
113
114 if prev_end <= start && run_end > start {
115 let start_offset = start - prev_end;
116 let end_offset = if run_end >= end {
117 end - prev_end
118 } else {
119 run_end - prev_end
120 };
121 current_run_end = current_run_end
122 .checked_add(&T::usize_as(end_offset - start_offset))
123 .ok_or_else(|| {
124 ArrowError::InvalidArgumentError(
125 "run end overflow when extending RunEndEncoded array: \
126 use a larger run-end type (e.g. Int64 instead of Int32)"
127 .to_string(),
128 )
129 })?;
130 run_ends_bytes.extend_from_slice(current_run_end.to_byte_slice());
131
132 values_range = Some((i, i + 1));
134 } else if prev_end >= start && run_end <= end {
135 current_run_end = current_run_end
136 .checked_add(&T::usize_as(run_end - prev_end))
137 .ok_or_else(|| {
138 ArrowError::InvalidArgumentError(
139 "run end overflow when extending RunEndEncoded array: \
140 use a larger run-end type (e.g. Int64 instead of Int32)"
141 .to_string(),
142 )
143 })?;
144 run_ends_bytes.extend_from_slice(current_run_end.to_byte_slice());
145
146 values_range = Some((values_range.expect("Unreachable: values_range cannot be None when prev_end >= start && run_end <= end. \
148 If prev_end >= start and run_end > prev_end (required for valid runs), then run_end > start, \
149 which means the first condition (prev_end <= start && run_end > start) would have been true \
150 and already set values_range to Some.").0, i + 1));
151 } else if prev_end < end && run_end >= end {
152 current_run_end = current_run_end
153 .checked_add(&T::usize_as(end - prev_end))
154 .ok_or_else(|| {
155 ArrowError::InvalidArgumentError(
156 "run end overflow when extending RunEndEncoded array: \
157 use a larger run-end type (e.g. Int64 instead of Int32)"
158 .to_string(),
159 )
160 })?;
161 run_ends_bytes.extend_from_slice(current_run_end.to_byte_slice());
162
163 values_range = Some((values_range.expect("Unreachable: values_range cannot be None when prev_end < end && run_end >= end. \
165 Due to sequential processing and monotonic prev_end advancement, if we reach a run \
166 that spans beyond the slice end (run_end >= end), at least one previous condition \
167 must have matched first to set values_range. Either the first condition matched when \
168 the slice started (prev_end <= start && run_end > start), or the second condition \
169 matched for runs within the slice (prev_end >= start && run_end <= end).").0, i + 1));
170 break;
171 }
172
173 prev_end = run_end;
174 if prev_end >= end {
175 break;
176 }
177 } else {
178 break;
179 }
180 }
181 Ok((run_ends_bytes, values_range))
182}
183
184fn process_extends_batch<T: ArrowNativeType>(
186 mutable: &mut _MutableArrayData,
187 source_array_idx: usize,
188 run_ends_bytes: Vec<u8>,
189 values_range: Option<(usize, usize)>,
190) -> Result<(), ArrowError> {
191 if run_ends_bytes.is_empty() {
192 return Ok(());
193 }
194
195 mutable.child_data[0]
197 .data
198 .buffer1
199 .extend_from_slice(&run_ends_bytes);
200 mutable.child_data[0].data.len += run_ends_bytes.len() / std::mem::size_of::<T>();
201
202 let (start_idx, end_idx) =
204 values_range.expect("values_range should be Some if run_ends_bytes is not empty");
205 mutable.child_data[1].try_extend(source_array_idx, start_idx, end_idx)
206}
207
208pub fn build_extend(array: &ArrayData) -> Extend<'_> {
212 Box::new(
213 move |mutable: &mut _MutableArrayData, array_idx: usize, start: usize, len: usize| {
214 if len == 0 {
215 return Ok(());
216 }
217
218 let source_run_ends = &array.child_data()[0];
220 let source_buffer = &source_run_ends.buffers()[0];
221
222 let dest_run_end_type =
224 if let DataType::RunEndEncoded(run_ends_field, _) = &mutable.data_type {
225 run_ends_field.data_type()
226 } else {
227 panic!("extend called on non-RunEndEncoded mutable array");
228 };
229
230 macro_rules! build_and_process_impl {
232 ($run_end_type:ty) => {{
233 let dest_last_run_end =
234 get_last_run_end::<$run_end_type>(&mutable.child_data[0]);
235 let (run_ends_bytes, values_range) = build_extend_arrays::<$run_end_type>(
236 source_buffer,
237 source_run_ends.len(),
238 start + array.offset(),
239 len,
240 dest_last_run_end,
241 )?;
242 process_extends_batch::<$run_end_type>(
243 mutable,
244 array_idx,
245 run_ends_bytes,
246 values_range,
247 )?;
248 }};
249 }
250
251 match dest_run_end_type {
252 DataType::Int16 => build_and_process_impl!(i16),
253 DataType::Int32 => build_and_process_impl!(i32),
254 DataType::Int64 => build_and_process_impl!(i64),
255 _ => panic!("Invalid run end type for RunEndEncoded array: {dest_run_end_type}",),
256 }
257 Ok(())
258 },
259 )
260}
261
262#[cfg(test)]
263mod tests {
264 use super::*;
265 use crate::transform::MutableArrayData;
266 use crate::{ArrayData, ArrayDataBuilder};
267 use arrow_buffer::Buffer;
268 use arrow_schema::{DataType, Field};
269 use std::sync::Arc;
270
271 fn create_run_array_data(run_ends: Vec<i32>, values: ArrayData) -> ArrayData {
272 let run_ends_field = Arc::new(Field::new("run_ends", DataType::Int32, false));
273 let values_field = Arc::new(Field::new("values", values.data_type().clone(), true));
274 let data_type = DataType::RunEndEncoded(run_ends_field, values_field);
275
276 let last_run_end = if run_ends.is_empty() {
277 0
278 } else {
279 run_ends[run_ends.len() - 1] as usize
280 };
281
282 let run_ends_buffer = Buffer::from_vec(run_ends);
283 let run_ends_data = ArrayDataBuilder::new(DataType::Int32)
284 .len(run_ends_buffer.len() / std::mem::size_of::<i32>())
285 .add_buffer(run_ends_buffer)
286 .build()
287 .unwrap();
288
289 ArrayDataBuilder::new(data_type)
290 .len(last_run_end)
291 .add_child_data(run_ends_data)
292 .add_child_data(values)
293 .build()
294 .unwrap()
295 }
296
297 fn create_run_array_data_int16(run_ends: Vec<i16>, values: ArrayData) -> ArrayData {
298 let run_ends_field = Arc::new(Field::new("run_ends", DataType::Int16, false));
299 let values_field = Arc::new(Field::new("values", values.data_type().clone(), true));
300 let data_type = DataType::RunEndEncoded(run_ends_field, values_field);
301
302 let last_run_end = if run_ends.is_empty() {
303 0
304 } else {
305 run_ends[run_ends.len() - 1] as usize
306 };
307
308 let run_ends_buffer = Buffer::from_vec(run_ends);
309 let run_ends_data = ArrayDataBuilder::new(DataType::Int16)
310 .len(run_ends_buffer.len() / std::mem::size_of::<i16>())
311 .add_buffer(run_ends_buffer)
312 .build()
313 .unwrap();
314
315 ArrayDataBuilder::new(data_type)
316 .len(last_run_end)
317 .add_child_data(run_ends_data)
318 .add_child_data(values)
319 .build()
320 .unwrap()
321 }
322
323 fn create_run_array_data_int64(run_ends: Vec<i64>, values: ArrayData) -> ArrayData {
324 let run_ends_field = Arc::new(Field::new("run_ends", DataType::Int64, false));
325 let values_field = Arc::new(Field::new("values", values.data_type().clone(), true));
326 let data_type = DataType::RunEndEncoded(run_ends_field, values_field);
327
328 let last_run_end = if run_ends.is_empty() {
329 0
330 } else {
331 run_ends[run_ends.len() - 1] as usize
332 };
333
334 let run_ends_buffer = Buffer::from_vec(run_ends);
335 let run_ends_data = ArrayDataBuilder::new(DataType::Int64)
336 .len(run_ends_buffer.len() / std::mem::size_of::<i64>())
337 .add_buffer(run_ends_buffer)
338 .build()
339 .unwrap();
340
341 ArrayDataBuilder::new(data_type)
342 .len(last_run_end)
343 .add_child_data(run_ends_data)
344 .add_child_data(values)
345 .build()
346 .unwrap()
347 }
348
349 fn create_int32_array_data(values: Vec<i32>) -> ArrayData {
350 let buffer = Buffer::from_vec(values);
351 ArrayDataBuilder::new(DataType::Int32)
352 .len(buffer.len() / std::mem::size_of::<i32>())
353 .add_buffer(buffer)
354 .build()
355 .unwrap()
356 }
357
358 fn create_string_dict_array_data(values: Vec<&str>, dict_values: Vec<&str>) -> ArrayData {
359 let dict_offsets: Vec<i32> = dict_values
361 .iter()
362 .scan(0i32, |acc, s| {
363 let offset = *acc;
364 *acc += s.len() as i32;
365 Some(offset)
366 })
367 .chain(std::iter::once(
368 dict_values.iter().map(|s| s.len()).sum::<usize>() as i32,
369 ))
370 .collect();
371
372 let dict_data: Vec<u8> = dict_values.iter().flat_map(|s| s.bytes()).collect();
373
374 let dict_array = ArrayDataBuilder::new(DataType::Utf8)
375 .len(dict_values.len())
376 .add_buffer(Buffer::from_vec(dict_offsets))
377 .add_buffer(Buffer::from_vec(dict_data))
378 .build()
379 .unwrap();
380
381 let keys: Vec<i32> = values
383 .iter()
384 .map(|v| dict_values.iter().position(|d| d == v).unwrap() as i32)
385 .collect();
386
387 let dict_type = DataType::Dictionary(Box::new(DataType::Int32), Box::new(DataType::Utf8));
389
390 ArrayDataBuilder::new(dict_type)
391 .len(values.len())
392 .add_buffer(Buffer::from_vec(keys))
393 .add_child_data(dict_array)
394 .build()
395 .unwrap()
396 }
397
398 #[test]
399 fn test_extend_nulls_int32() {
400 let values = create_int32_array_data(vec![42]);
402
403 let ree_array = create_run_array_data(vec![5], values);
405
406 let mut mutable = MutableArrayData::new(vec![&ree_array], true, 10);
407
408 mutable.try_extend_nulls(3).unwrap();
409 mutable.try_extend(0, 0, 5).unwrap();
410 mutable.try_extend_nulls(3).unwrap();
411
412 let result = mutable.freeze();
414 let run_ends_buffer = &result.child_data()[0].buffers()[0];
415 let run_ends_slice = run_ends_buffer.as_slice();
416
417 assert_eq!(result.child_data()[0].len(), 3);
419 let first_run_end = i32::from_ne_bytes(run_ends_slice[0..4].try_into().unwrap());
420 let second_run_end = i32::from_ne_bytes(run_ends_slice[4..8].try_into().unwrap());
421 let third_run_end = i32::from_ne_bytes(run_ends_slice[8..12].try_into().unwrap());
422 assert_eq!(first_run_end, 3);
423 assert_eq!(second_run_end, 8);
424 assert_eq!(third_run_end, 11);
425
426 assert_eq!(result.child_data()[1].len(), 3); let values_buffer = &result.child_data()[1].buffers()[0];
429 let values_slice = values_buffer.as_slice();
430
431 let second_value = i32::from_ne_bytes(values_slice[4..8].try_into().unwrap());
433
434 assert_eq!(second_value, 42);
436
437 let values_array = &result.child_data()[1];
439 assert!(values_array.is_null(0));
441 assert!(values_array.is_valid(1));
443 assert!(values_array.is_null(2));
445 }
446
447 #[test]
448 fn test_extend_nulls_int16() {
449 let values = create_int32_array_data(vec![42]);
451
452 let ree_array = create_run_array_data_int16(vec![5i16], values);
454
455 let mut mutable = MutableArrayData::new(vec![&ree_array], true, 10);
456
457 mutable.try_extend(0, 0, 5).unwrap();
459
460 mutable.try_extend_nulls(3).unwrap();
462
463 let result = mutable.freeze();
465 let run_ends_buffer = &result.child_data()[0].buffers()[0];
466 let run_ends_slice = run_ends_buffer.as_slice();
467
468 assert_eq!(result.child_data()[0].len(), 2);
470 let first_run_end = i16::from_ne_bytes(run_ends_slice[0..2].try_into().unwrap());
471 let second_run_end = i16::from_ne_bytes(run_ends_slice[2..4].try_into().unwrap());
472 assert_eq!(first_run_end, 5);
473 assert_eq!(second_run_end, 8);
474 }
475
476 #[test]
477 fn test_extend_nulls_int64() {
478 let values = create_int32_array_data(vec![42]);
480
481 let ree_array = create_run_array_data_int64(vec![5i64], values);
483
484 let mut mutable = MutableArrayData::new(vec![&ree_array], true, 10);
485
486 mutable.try_extend(0, 0, 5).unwrap();
488
489 mutable.try_extend_nulls(3).unwrap();
491
492 let result = mutable.freeze();
494 let run_ends_buffer = &result.child_data()[0].buffers()[0];
495 let run_ends_slice = run_ends_buffer.as_slice();
496
497 assert_eq!(result.child_data()[0].len(), 2);
499 let first_run_end = i64::from_ne_bytes(run_ends_slice[0..8].try_into().unwrap());
500 let second_run_end = i64::from_ne_bytes(run_ends_slice[8..16].try_into().unwrap());
501 assert_eq!(first_run_end, 5);
502 assert_eq!(second_run_end, 8);
503 }
504
505 #[test]
506 fn test_extend_int32() {
507 let values = create_int32_array_data(vec![10, 20]);
509
510 let ree_array = create_run_array_data(vec![2, 5], values);
512
513 let mut mutable = MutableArrayData::new(vec![&ree_array], false, 10);
514
515 mutable.try_extend(0, 0, 5).unwrap();
517
518 let result = mutable.freeze();
519
520 assert_eq!(result.len(), 5); assert!(!result.child_data()[0].is_empty()); assert_eq!(result.child_data()[0].len(), result.child_data()[1].len()); }
527
528 #[test]
529 fn test_extend_empty() {
530 let values = create_int32_array_data(vec![]);
531 let ree_array = create_run_array_data(vec![], values);
532
533 let mut mutable = MutableArrayData::new(vec![&ree_array], false, 10);
534 mutable.try_extend(0, 0, 0).unwrap();
535
536 let result = mutable.freeze();
537 assert_eq!(result.len(), 0);
538 assert_eq!(result.child_data()[0].len(), 0);
539 }
540
541 #[test]
542 fn test_build_extend_arrays_int16() {
543 let buffer = Buffer::from_vec(vec![3i16, 5i16, 8i16]);
544 let (run_ends_bytes, values_range) =
545 build_extend_arrays::<i16>(&buffer, 3, 2, 4, 0i16).unwrap();
546
547 assert_eq!(run_ends_bytes.len(), 3 * std::mem::size_of::<i16>());
554 assert_eq!(values_range, Some((0, 3)));
555
556 let expected_bytes = [1i16, 3i16, 4i16]
558 .iter()
559 .flat_map(|&val| val.to_ne_bytes())
560 .collect::<Vec<u8>>();
561 assert_eq!(run_ends_bytes, expected_bytes);
562 }
563
564 #[test]
565 fn test_build_extend_arrays_int64() {
566 let buffer = Buffer::from_vec(vec![3i64, 5i64, 8i64]);
567 let (run_ends_bytes, values_range) =
568 build_extend_arrays::<i64>(&buffer, 3, 2, 4, 0i64).unwrap();
569
570 assert_eq!(run_ends_bytes.len(), 3 * std::mem::size_of::<i64>());
572 assert_eq!(values_range, Some((0, 3)));
573
574 let expected_bytes = [1i64, 3i64, 4i64]
576 .iter()
577 .flat_map(|&val| val.to_ne_bytes())
578 .collect::<Vec<u8>>();
579 assert_eq!(run_ends_bytes, expected_bytes);
580 }
581
582 #[test]
583 fn test_extend_string_dict() {
584 let dict_values = vec!["hello", "world"];
586 let values = create_string_dict_array_data(vec!["hello", "world"], dict_values);
587
588 let ree_array = create_run_array_data(vec![2, 5], values);
590
591 let mut mutable = MutableArrayData::new(vec![&ree_array], false, 10);
592
593 mutable.try_extend(0, 0, 5).unwrap();
595
596 let result = mutable.freeze();
597
598 assert_eq!(result.len(), 5); assert!(!result.child_data()[0].is_empty()); assert_eq!(result.child_data()[0].len(), result.child_data()[1].len()); assert_eq!(result.child_data()[0].len(), 2);
607 assert_eq!(result.child_data()[1].len(), 2);
608 }
609
610 #[test]
611 fn test_extend_nulls_overflow_i16() {
612 let values = create_int32_array_data(vec![42]);
613 let ree_array = create_run_array_data_int16(vec![5], values);
615 let mut mutable = MutableArrayData::new(vec![&ree_array], true, 10);
616
617 mutable.try_extend(0, 0, 5_usize).unwrap();
619
620 let err = mutable.try_extend_nulls(i16::MAX as usize).unwrap_err();
622 assert!(
623 err.to_string().contains("run end overflow"),
624 "unexpected error: {err}"
625 );
626 }
627
628 #[test]
629 fn test_extend_nulls_overflow_i32() {
630 let values = create_int32_array_data(vec![42]);
631 let ree_array = create_run_array_data(vec![10], values);
633 let mut mutable = MutableArrayData::new(vec![&ree_array], true, 10);
634
635 mutable.try_extend(0, 0, 10_usize).unwrap();
637
638 let err = mutable.try_extend_nulls(i32::MAX as usize).unwrap_err();
640 assert!(
641 err.to_string().contains("run end overflow"),
642 "unexpected error: {err}"
643 );
644 }
645
646 #[test]
647 fn test_build_extend_overflow_i16() {
648 let values = create_int32_array_data(vec![10]);
650 let source_array = create_run_array_data_int16(vec![20], values);
651
652 let dest_values = create_int32_array_data(vec![42]);
654 let dest_array = create_run_array_data_int16(vec![i16::MAX - 5], dest_values);
655
656 let mut mutable = MutableArrayData::new(vec![&source_array, &dest_array], false, 10);
657
658 mutable.try_extend(1, 0, (i16::MAX - 5) as usize).unwrap();
660
661 let err = mutable.try_extend(0, 0, 20).unwrap_err();
663 assert!(
664 err.to_string().contains("run end overflow"),
665 "unexpected error: {err}"
666 );
667 }
668
669 #[test]
670 fn test_build_extend_overflow_i32() {
671 let values = create_int32_array_data(vec![10]);
673 let source_array = create_run_array_data(vec![100], values);
674
675 let dest_values = create_int32_array_data(vec![42]);
677 let dest_array = create_run_array_data(vec![i32::MAX - 50], dest_values);
678
679 let mut mutable = MutableArrayData::new(vec![&source_array, &dest_array], false, 10);
680
681 mutable.try_extend(1, 0, (i32::MAX - 50) as usize).unwrap();
683
684 let err = mutable.try_extend(0, 0, 100).unwrap_err();
686 assert!(
687 err.to_string().contains("run end overflow"),
688 "unexpected error: {err}"
689 );
690 }
691}