1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
//! The typed heart of every Series column.
use crate::prelude::*;
use arrow::{array::*, bitmap::Bitmap};
use itertools::Itertools;
use polars_arrow::prelude::ValueSize;
use std::marker::PhantomData;
use std::sync::Arc;

pub mod ops;
#[macro_use]
pub mod arithmetic;
pub mod boolean;
pub mod builder;
pub mod cast;
pub mod comparison;
pub mod float;
pub mod iterator;
pub mod kernels;
#[cfg(feature = "ndarray")]
mod ndarray;

#[cfg(feature = "object")]
#[cfg_attr(docsrs, doc(cfg(feature = "object")))]
pub mod object;
#[cfg(feature = "random")]
#[cfg_attr(docsrs, doc(cfg(feature = "random")))]
mod random;
#[cfg(feature = "strings")]
#[cfg_attr(docsrs, doc(cfg(feature = "strings")))]
pub mod strings;
#[cfg(any(
    feature = "temporal",
    feature = "dtype-datetime",
    feature = "dtype-date"
))]
#[cfg_attr(docsrs, doc(cfg(feature = "temporal")))]
pub mod temporal;
mod trusted_len;
pub mod upstream_traits;
use arrow::array::Array;
mod bitwise;
#[cfg(feature = "dtype-categorical")]
pub(crate) mod categorical;
pub(crate) mod list;
pub(crate) mod logical;

use polars_arrow::prelude::*;

#[cfg(feature = "dtype-categorical")]
use crate::chunked_array::categorical::RevMapping;
use crate::utils::{slice_offsets, CustomIterTools};
use std::mem;

#[cfg(not(feature = "dtype-categorical"))]
pub struct RevMapping {}

pub type ChunkIdIter<'a> = std::iter::Map<std::slice::Iter<'a, ArrayRef>, fn(&ArrayRef) -> usize>;

/// # ChunkedArray
///
/// Every Series contains a `ChunkedArray<T>`. Unlike Series, ChunkedArray's are typed. This allows
/// us to apply closures to the data and collect the results to a `ChunkedArray` of the same type `T`.
/// Below we use an apply to use the cosine function to the values of a `ChunkedArray`.
///
/// ```rust
/// # use polars_core::prelude::*;
/// fn apply_cosine(ca: &Float32Chunked) -> Float32Chunked {
///     ca.apply(|v| v.cos())
/// }
/// ```
///
/// If we would like to cast the result we could use a Rust Iterator instead of an `apply` method.
/// Note that Iterators are slightly slower as the null values aren't ignored implicitly.
///
/// ```rust
/// # use polars_core::prelude::*;
/// fn apply_cosine_and_cast(ca: &Float32Chunked) -> Float64Chunked {
///     ca.into_iter()
///         .map(|opt_v| {
///         opt_v.map(|v| v.cos() as f64)
///     }).collect()
/// }
/// ```
///
/// Another option is to first cast and then use an apply.
///
/// ```rust
/// # use polars_core::prelude::*;
/// fn apply_cosine_and_cast(ca: &Float32Chunked) -> Float64Chunked {
///     ca.apply_cast_numeric(|v| v.cos() as f64)
/// }
/// ```
///
/// ## Conversion between Series and ChunkedArray's
/// Conversion from a `Series` to a `ChunkedArray` is effortless.
///
/// ```rust
/// # use polars_core::prelude::*;
/// fn to_chunked_array(series: &Series) -> Result<&Int32Chunked>{
///     series.i32()
/// }
///
/// fn to_series(ca: Int32Chunked) -> Series {
///     ca.into_series()
/// }
/// ```
///
/// # Iterators
///
/// `ChunkedArrays` fully support Rust native [Iterator](https://doc.rust-lang.org/std/iter/trait.Iterator.html)
/// and [DoubleEndedIterator](https://doc.rust-lang.org/std/iter/trait.DoubleEndedIterator.html) traits, thereby
/// giving access to all the excellent methods available for [Iterators](https://doc.rust-lang.org/std/iter/trait.Iterator.html).
///
/// ```rust
/// # use polars_core::prelude::*;
///
/// fn iter_forward(ca: &Float32Chunked) {
///     ca.into_iter()
///         .for_each(|opt_v| println!("{:?}", opt_v))
/// }
///
/// fn iter_backward(ca: &Float32Chunked) {
///     ca.into_iter()
///         .rev()
///         .for_each(|opt_v| println!("{:?}", opt_v))
/// }
/// ```
///
/// # Memory layout
///
/// `ChunkedArray`'s use [Apache Arrow](https://github.com/apache/arrow) as backend for the memory layout.
/// Arrows memory is immutable which makes it possible to make multiple zero copy (sub)-views from a single array.
///
/// To be able to append data, Polars uses chunks to append new memory locations, hence the `ChunkedArray<T>` data structure.
/// Appends are cheap, because it will not lead to a full reallocation of the whole array (as could be the case with a Rust Vec).
///
/// However, multiple chunks in a `ChunkArray` will slow down the Iterators, arithmetic and other operations.
/// When multiplying two `ChunkArray'`s with different chunk sizes they cannot utilize [SIMD](https://en.wikipedia.org/wiki/SIMD) for instance.
/// However, when chunk size don't match, Iterators will be used to do the operation (instead of arrows upstream implementation, which may utilize SIMD) and
/// the result will be a single chunked array.
///
/// **The key takeaway is that by applying operations on a `ChunkArray` of multiple chunks, the results will converge to
/// a `ChunkArray` of a single chunk!** It is recommended to leave them as is. If you want to have predictable performance
/// (no unexpected re-allocation of memory), it is advised to call the [rechunk](chunked_array/chunkops/trait.ChunkOps.html) after
/// multiple append operations.
pub struct ChunkedArray<T> {
    pub(crate) field: Arc<Field>,
    pub(crate) chunks: Vec<ArrayRef>,
    phantom: PhantomData<T>,
    /// maps categorical u32 indexes to String values
    pub(crate) categorical_map: Option<Arc<RevMapping>>,
    /// first bit: sorted
    /// second_bit: sorted reverse
    /// third bit dtype list: fast_explode
    ///     - unset: unknown or not all arrays have at least one value
    ///     - set: all list arrays are filled (this allows for cheap explode)
    pub(crate) bit_settings: u8,
}

impl<T> ChunkedArray<T> {
    #[cfg(feature = "asof_join")]
    pub(crate) fn is_sorted(&self) -> bool {
        self.bit_settings & 1 != 0
    }

    #[cfg(feature = "asof_join")]
    pub(crate) fn is_sorted_reverse(&self) -> bool {
        self.bit_settings & 1 << 1 != 0
    }

    /// Set the 'sorted' bit meta info.
    pub(crate) fn set_sorted(&mut self, reverse: bool) {
        if reverse {
            self.bit_settings |= 1 << 1
        } else {
            self.bit_settings |= 1
        }
    }

    /// Get the index of the first non null value in this ChunkedArray.
    pub fn first_non_null(&self) -> Option<usize> {
        let mut offset = 0;
        for (_, null_bitmap) in self.null_bits() {
            if let Some(null_bitmap) = null_bitmap {
                for (idx, is_valid) in null_bitmap.iter().enumerate() {
                    if is_valid {
                        return Some(offset + idx);
                    }
                }
                offset += null_bitmap.len()
            } else {
                return Some(offset);
            }
        }
        None
    }

    /// Get the buffer of bits representing null values
    pub fn null_bits(&self) -> impl Iterator<Item = (usize, Option<&Bitmap>)> + '_ {
        self.chunks
            .iter()
            .map(|arr| (arr.null_count(), arr.validity()))
    }

    /// Shrink the capacity of this array to fit it's length.
    pub fn shrink_to_fit(&mut self) {
        self.chunks = vec![arrow::compute::concat::concatenate(
            self.chunks.iter().map(|a| &**a).collect_vec().as_slice(),
        )
        .unwrap()
        .into()];
    }

    /// Unpack a Series to the same physical type.
    ///
    /// # Safety
    ///
    /// This is unsafe as the dtype may be incorrect and
    /// is assumed to be correct in other safe code.
    pub(crate) unsafe fn unpack_series_matching_physical_type(
        &self,
        series: &Series,
    ) -> &ChunkedArray<T> {
        let series_trait = &**series;
        if self.dtype() == series.dtype() {
            &*(series_trait as *const dyn SeriesTrait as *const ChunkedArray<T>)
        } else {
            use DataType::*;
            match (self.dtype(), series.dtype()) {
                (Int64, Datetime) | (Int32, Date) => {
                    &*(series_trait as *const dyn SeriesTrait as *const ChunkedArray<T>)
                }
                _ => panic!(
                    "cannot unpack series {:?} into matching type {:?}",
                    series,
                    self.dtype()
                ),
            }
        }
    }

    /// Series to ChunkedArray<T>
    pub fn unpack_series_matching_type(&self, series: &Series) -> Result<&ChunkedArray<T>> {
        if self.dtype() == series.dtype() {
            // Safety
            // dtype will be correct.
            Ok(unsafe { self.unpack_series_matching_physical_type(series) })
        } else {
            Err(PolarsError::DataTypeMisMatch(
                format!(
                    "cannot unpack series {:?} into matching type {:?}",
                    series,
                    self.dtype()
                )
                .into(),
            ))
        }
    }

    /// Combined length of all the chunks.
    pub fn len(&self) -> usize {
        self.chunks.iter().fold(0, |acc, arr| acc + arr.len())
    }

    /// Check if ChunkedArray is empty.
    pub fn is_empty(&self) -> bool {
        self.len() == 0
    }

    /// Unique id representing the number of chunks
    pub fn chunk_id(&self) -> ChunkIdIter {
        self.chunks.iter().map(|chunk| chunk.len())
    }

    /// A reference to the chunks
    pub fn chunks(&self) -> &Vec<ArrayRef> {
        &self.chunks
    }

    /// Returns true if contains a single chunk and has no null values
    pub fn is_optimal_aligned(&self) -> bool {
        self.chunks.len() == 1 && self.null_count() == 0
    }

    /// Count the null values.
    #[inline]
    pub fn null_count(&self) -> usize {
        self.chunks.iter().map(|arr| arr.null_count()).sum()
    }

    /// Take a view of top n elements
    pub fn limit(&self, num_elements: usize) -> Self {
        self.slice(0, num_elements)
    }

    /// Append arrow array in place.
    ///
    /// ```rust
    /// # use polars_core::prelude::*;
    /// let mut array = Int32Chunked::new_from_slice("array", &[1, 2]);
    /// let array_2 = Int32Chunked::new_from_slice("2nd", &[3]);
    ///
    /// array.append(&array_2);
    /// assert_eq!(Vec::from(&array), [Some(1), Some(2), Some(3)])
    /// ```
    pub fn append_array(&mut self, other: ArrayRef) -> Result<()> {
        if matches!(self.dtype(), DataType::Categorical) {
            return Err(PolarsError::InvalidOperation(
                "append_array not supported for categorical type".into(),
            ));
        }
        if self.field.data_type() == other.data_type() {
            self.chunks.push(other);
            Ok(())
        } else {
            Err(PolarsError::DataTypeMisMatch(
                format!(
                    "cannot append array of type {:?} in array of type {:?}",
                    other.data_type(),
                    self.dtype()
                )
                .into(),
            ))
        }
    }

    /// Create a new ChunkedArray from self, where the chunks are replaced.
    fn copy_with_chunks(&self, chunks: Vec<ArrayRef>) -> Self {
        ChunkedArray {
            field: self.field.clone(),
            chunks,
            phantom: PhantomData,
            categorical_map: self.categorical_map.clone(),
            bit_settings: self.bit_settings,
        }
    }

    /// Slice the array. The chunks are reallocated the underlying data slices are zero copy.
    ///
    /// When offset is negative it will be counted from the end of the array.
    /// This method will never error,
    /// and will slice the best match when offset, or length is out of bounds
    pub fn slice(&self, offset: i64, length: usize) -> Self {
        let (raw_offset, slice_len) = slice_offsets(offset, length, self.len());

        let mut remaining_length = slice_len;
        let mut remaining_offset = raw_offset;
        let mut new_chunks = vec![];

        for chunk in &self.chunks {
            let chunk_len = chunk.len();
            if remaining_offset > 0 && remaining_offset >= chunk_len {
                remaining_offset -= chunk_len;
                continue;
            }
            let take_len;
            if remaining_length + remaining_offset > chunk_len {
                take_len = chunk_len - remaining_offset;
            } else {
                take_len = remaining_length;
            }

            new_chunks.push(chunk.slice(remaining_offset, take_len).into());
            remaining_length -= take_len;
            remaining_offset = 0;
            if remaining_length == 0 {
                break;
            }
        }
        self.copy_with_chunks(new_chunks)
    }

    /// Get a mask of the null values.
    pub fn is_null(&self) -> BooleanChunked {
        if self.null_count() == 0 {
            return BooleanChunked::full("is_null", false, self.len());
        }
        let chunks = self
            .chunks
            .iter()
            .map(|arr| {
                let bitmap = arr
                    .validity()
                    .map(|bitmap| !bitmap)
                    .unwrap_or_else(|| Bitmap::new_zeroed(arr.len()));
                Arc::new(BooleanArray::from_data_default(bitmap, None)) as ArrayRef
            })
            .collect_vec();
        BooleanChunked::new_from_chunks("is_null", chunks)
    }

    /// Get a mask of the valid values.
    pub fn is_not_null(&self) -> BooleanChunked {
        if self.null_count() == 0 {
            return BooleanChunked::full("is_not_null", true, self.len());
        }
        let chunks = self
            .chunks
            .iter()
            .map(|arr| {
                let bitmap = arr
                    .validity()
                    .cloned()
                    .unwrap_or_else(|| !(&Bitmap::new_zeroed(arr.len())));
                Arc::new(BooleanArray::from_data_default(bitmap, None)) as ArrayRef
            })
            .collect_vec();
        BooleanChunked::new_from_chunks("is_not_null", chunks)
    }

    /// Get data type of ChunkedArray.
    pub fn dtype(&self) -> &DataType {
        self.field.data_type()
    }

    /// Get the head of the ChunkedArray
    pub fn head(&self, length: Option<usize>) -> Self {
        match length {
            Some(len) => self.slice(0, std::cmp::min(len, self.len())),
            None => self.slice(0, std::cmp::min(10, self.len())),
        }
    }

    /// Get the tail of the ChunkedArray
    pub fn tail(&self, length: Option<usize>) -> Self {
        let len = match length {
            Some(len) => std::cmp::min(len, self.len()),
            None => std::cmp::min(10, self.len()),
        };
        self.slice(-(len as i64), len)
    }

    /// Name of the ChunkedArray.
    pub fn name(&self) -> &str {
        self.field.name()
    }

    /// Get a reference to the field.
    pub fn ref_field(&self) -> &Field {
        &self.field
    }

    /// Rename this ChunkedArray.
    pub fn rename(&mut self, name: &str) {
        self.field = Arc::new(Field::new(name, self.field.data_type().clone()))
    }
}

impl<T> ChunkedArray<T>
where
    T: PolarsDataType,
    ChunkedArray<T>: ChunkOps,
{
    /// Should be used to match the chunk_id of another ChunkedArray.
    /// # Panics
    /// It is the callers responsibility to ensure that this ChunkedArray has a single chunk.
    pub(crate) fn match_chunks<I>(&self, chunk_id: I) -> Self
    where
        I: Iterator<Item = usize>,
    {
        debug_assert!(self.chunks.len() == 1);
        // Takes a ChunkedArray containing a single chunk
        let slice = |ca: &Self| {
            let array = &ca.chunks[0];

            let mut offset = 0;
            let chunks = chunk_id
                .map(|len| {
                    let out = array.slice(offset, len).into();
                    offset += len;
                    out
                })
                .collect();

            Self::new_from_chunks(self.name(), chunks)
        };

        if self.chunks.len() != 1 {
            let out = self.rechunk();
            slice(&out)
        } else {
            slice(self)
        }
    }
}

impl<T> ChunkedArray<T>
where
    T: PolarsDataType,
{
    /// Create a new ChunkedArray from existing chunks.
    pub fn new_from_chunks(name: &str, chunks: Vec<ArrayRef>) -> Self {
        // prevent List<Null> if the inner list type is known.
        let datatype = if matches!(T::get_dtype(), DataType::List(_)) {
            if let Some(arr) = chunks.get(0) {
                arr.data_type().into()
            } else {
                T::get_dtype()
            }
        } else {
            T::get_dtype()
        };
        let field = Arc::new(Field::new(name, datatype));
        ChunkedArray {
            field,
            chunks,
            phantom: PhantomData,
            categorical_map: None,
            bit_settings: 0,
        }
    }
}

impl<T> ChunkedArray<T>
where
    T: PolarsNumericType,
{
    /// Create a new ChunkedArray by taking ownership of the AlignedVec. This operation is zero copy.
    pub fn new_from_aligned_vec(name: &str, v: AlignedVec<T::Native>) -> Self {
        let arr = to_array::<T>(v, None);
        Self::new_from_chunks(name, vec![arr])
    }

    /// Nullify values in slice with an existing null bitmap
    pub fn new_from_owned_with_null_bitmap(
        name: &str,
        values: AlignedVec<T::Native>,
        buffer: Option<Bitmap>,
    ) -> Self {
        let arr = to_array::<T>(values, buffer);
        ChunkedArray {
            field: Arc::new(Field::new(name, T::get_dtype())),
            chunks: vec![arr],
            phantom: PhantomData,
            categorical_map: None,
            ..Default::default()
        }
    }
}

pub(crate) trait AsSinglePtr {
    /// Rechunk and return a ptr to the start of the array
    fn as_single_ptr(&mut self) -> Result<usize> {
        Err(PolarsError::InvalidOperation(
            "operation as_single_ptr not supported for this dtype".into(),
        ))
    }
}

impl<T> AsSinglePtr for ChunkedArray<T>
where
    T: PolarsNumericType,
{
    fn as_single_ptr(&mut self) -> Result<usize> {
        let mut ca = self.rechunk();
        mem::swap(&mut ca, self);
        let a = self.data_views().next().unwrap();
        let ptr = a.as_ptr();
        Ok(ptr as usize)
    }
}

impl AsSinglePtr for BooleanChunked {}
impl AsSinglePtr for ListChunked {}
impl AsSinglePtr for Utf8Chunked {}
impl AsSinglePtr for CategoricalChunked {}
#[cfg(feature = "object")]
impl<T> AsSinglePtr for ObjectChunked<T> {}

impl<T> ChunkedArray<T>
where
    T: PolarsNumericType,
{
    /// Contiguous slice
    pub fn cont_slice(&self) -> Result<&[T::Native]> {
        if self.chunks.len() == 1 && self.chunks[0].null_count() == 0 {
            Ok(self.downcast_iter().next().map(|arr| arr.values()).unwrap())
        } else {
            Err(PolarsError::NoSlice)
        }
    }

    /// Get slices of the underlying arrow data.
    /// NOTE: null values should be taken into account by the user of these slices as they are handled
    /// separately
    pub fn data_views(&self) -> impl Iterator<Item = &[T::Native]> + DoubleEndedIterator {
        self.downcast_iter().map(|arr| arr.values().as_slice())
    }

    #[allow(clippy::wrong_self_convention)]
    pub fn into_no_null_iter(
        &self,
    ) -> impl Iterator<Item = T::Native>
           + '_
           + Send
           + Sync
           + ExactSizeIterator
           + DoubleEndedIterator
           + TrustedLen {
        // .copied was significantly slower in benchmark, next call did not inline?
        #[allow(clippy::map_clone)]
        self.data_views()
            .flatten()
            .map(|v| *v)
            .trust_my_length(self.len())
    }
}

impl<T> Clone for ChunkedArray<T> {
    fn clone(&self) -> Self {
        ChunkedArray {
            field: self.field.clone(),
            chunks: self.chunks.clone(),
            phantom: PhantomData,
            categorical_map: self.categorical_map.clone(),
            bit_settings: self.bit_settings,
        }
    }
}

impl<T> AsRef<ChunkedArray<T>> for ChunkedArray<T> {
    fn as_ref(&self) -> &ChunkedArray<T> {
        self
    }
}

impl ValueSize for ListChunked {
    fn get_values_size(&self) -> usize {
        self.chunks
            .iter()
            .fold(0usize, |acc, arr| acc + arr.get_values_size())
    }
}

impl ValueSize for Utf8Chunked {
    fn get_values_size(&self) -> usize {
        self.chunks
            .iter()
            .fold(0usize, |acc, arr| acc + arr.get_values_size())
    }
}

impl ListChunked {
    /// Get the inner data type of the list.
    pub fn inner_dtype(&self) -> DataType {
        match self.dtype() {
            DataType::List(dt) => *dt.clone(),
            _ => unreachable!(),
        }
    }
}

pub(crate) fn to_primitive<T: PolarsNumericType>(
    values: AlignedVec<T::Native>,
    validity: Option<Bitmap>,
) -> PrimitiveArray<T::Native> {
    PrimitiveArray::from_data(T::get_dtype().to_arrow(), values.into(), validity)
}

pub(crate) fn to_array<T: PolarsNumericType>(
    values: AlignedVec<T::Native>,
    validity: Option<Bitmap>,
) -> ArrayRef {
    Arc::new(to_primitive::<T>(values, validity))
}

impl<T: PolarsNumericType> From<PrimitiveArray<T::Native>> for ChunkedArray<T> {
    fn from(a: PrimitiveArray<T::Native>) -> Self {
        ChunkedArray::new_from_chunks("", vec![Arc::new(a)])
    }
}

#[cfg(test)]
pub(crate) mod test {
    use crate::prelude::*;

    pub(crate) fn get_chunked_array() -> Int32Chunked {
        ChunkedArray::new_from_slice("a", &[1, 2, 3])
    }

    #[test]
    fn test_sort() {
        let a = Int32Chunked::new_from_slice("a", &[1, 9, 3, 2]);
        let b = a
            .sort(false)
            .into_iter()
            .map(|opt| opt.unwrap())
            .collect::<Vec<_>>();
        assert_eq!(b, [1, 2, 3, 9]);
        let a = Utf8Chunked::new_from_slice("a", &["b", "a", "c"]);
        let a = a.sort(false);
        let b = a.into_iter().collect::<Vec<_>>();
        assert_eq!(b, [Some("a"), Some("b"), Some("c")]);
    }

    #[test]
    fn arithmetic() {
        let s1 = get_chunked_array();
        println!("{:?}", s1.chunks);
        let s2 = &s1;
        let s1 = &s1;
        println!("{:?}", s1 + s2);
        println!("{:?}", s1 - s2);
        println!("{:?}", s1 * s2);
    }

    #[test]
    fn iter() {
        let s1 = get_chunked_array();
        // sum
        assert_eq!(s1.into_iter().fold(0, |acc, val| { acc + val.unwrap() }), 6)
    }

    #[test]
    fn limit() {
        let a = get_chunked_array();
        let b = a.limit(2);
        println!("{:?}", b);
        assert_eq!(b.len(), 2)
    }

    #[test]
    fn filter() {
        let a = get_chunked_array();
        let b = a
            .filter(&BooleanChunked::new_from_slice(
                "filter",
                &[true, false, false],
            ))
            .unwrap();
        assert_eq!(b.len(), 1);
        assert_eq!(b.into_iter().next(), Some(Some(1)));
    }

    #[test]
    fn aggregates_numeric() {
        let a = get_chunked_array();
        assert_eq!(a.max(), Some(3));
        assert_eq!(a.min(), Some(1));
        assert_eq!(a.sum(), Some(6))
    }

    #[test]
    fn take() {
        let a = get_chunked_array();
        let new = a.take([0usize, 1].iter().copied().into()).unwrap();
        assert_eq!(new.len(), 2)
    }

    #[test]
    fn get() {
        let mut a = get_chunked_array();
        assert_eq!(AnyValue::Int32(2), a.get_any_value(1));
        // check if chunks indexes are properly determined
        a.append_array(a.chunks[0].clone()).unwrap();
        assert_eq!(AnyValue::Int32(1), a.get_any_value(3));
    }

    #[test]
    fn cast() {
        let a = get_chunked_array();
        let b = a.cast(&DataType::Int64).unwrap();
        assert_eq!(b.dtype(), &ArrowDataType::Int64)
    }

    fn assert_slice_equal<T>(ca: &ChunkedArray<T>, eq: &[T::Native])
    where
        ChunkedArray<T>: ChunkOps,
        T: PolarsNumericType,
    {
        assert_eq!(
            ca.into_iter().map(|opt| opt.unwrap()).collect::<Vec<_>>(),
            eq
        )
    }

    #[test]
    fn slice() {
        let mut first = UInt32Chunked::new_from_slice("first", &[0, 1, 2]);
        let second = UInt32Chunked::new_from_slice("second", &[3, 4, 5]);
        first.append(&second);
        assert_slice_equal(&first.slice(0, 3), &[0, 1, 2]);
        assert_slice_equal(&first.slice(0, 4), &[0, 1, 2, 3]);
        assert_slice_equal(&first.slice(1, 4), &[1, 2, 3, 4]);
        assert_slice_equal(&first.slice(3, 2), &[3, 4]);
        assert_slice_equal(&first.slice(3, 3), &[3, 4, 5]);
        assert_slice_equal(&first.slice(-3, 3), &[3, 4, 5]);
        assert_slice_equal(&first.slice(-6, 6), &[0, 1, 2, 3, 4, 5]);

        assert_eq!(first.slice(-7, 2).len(), 2);
        assert_eq!(first.slice(-3, 4).len(), 3);
        assert_eq!(first.slice(3, 4).len(), 3);
        assert_eq!(first.slice(10, 4).len(), 0);
    }

    #[test]
    fn sorting() {
        let s = UInt32Chunked::new_from_slice("", &[9, 2, 4]);
        let sorted = s.sort(false);
        assert_slice_equal(&sorted, &[2, 4, 9]);
        let sorted = s.sort(true);
        assert_slice_equal(&sorted, &[9, 4, 2]);

        let s: Utf8Chunked = ["b", "a", "z"].iter().collect();
        let sorted = s.sort(false);
        assert_eq!(
            sorted.into_iter().collect::<Vec<_>>(),
            &[Some("a"), Some("b"), Some("z")]
        );
        let sorted = s.sort(true);
        assert_eq!(
            sorted.into_iter().collect::<Vec<_>>(),
            &[Some("z"), Some("b"), Some("a")]
        );
        let s: Utf8Chunked = [Some("b"), None, Some("z")].iter().copied().collect();
        let sorted = s.sort(false);
        assert_eq!(
            sorted.into_iter().collect::<Vec<_>>(),
            &[None, Some("b"), Some("z")]
        );
    }

    #[test]
    fn reverse() {
        let s = UInt32Chunked::new_from_slice("", &[1, 2, 3]);
        // path with continuous slice
        assert_slice_equal(&s.reverse(), &[3, 2, 1]);
        // path with options
        let s = UInt32Chunked::new_from_opt_slice("", &[Some(1), None, Some(3)]);
        assert_eq!(Vec::from(&s.reverse()), &[Some(3), None, Some(1)]);
        let s = BooleanChunked::new_from_slice("", &[true, false]);
        assert_eq!(Vec::from(&s.reverse()), &[Some(false), Some(true)]);

        let s = Utf8Chunked::new_from_slice("", &["a", "b", "c"]);
        assert_eq!(Vec::from(&s.reverse()), &[Some("c"), Some("b"), Some("a")]);

        let s = Utf8Chunked::new_from_opt_slice("", &[Some("a"), None, Some("c")]);
        assert_eq!(Vec::from(&s.reverse()), &[Some("c"), None, Some("a")]);
    }

    #[test]
    fn test_null_sized_chunks() {
        let mut s = Float64Chunked::new_from_slice("s", &Vec::<f64>::new());
        s.append(&Float64Chunked::new_from_slice("s2", &[1., 2., 3.]));
        dbg!(&s);

        let s = Float64Chunked::new_from_slice("s", &Vec::<f64>::new());
        dbg!(&s.into_iter().next());
    }

    #[test]
    #[cfg(feature = "dtype-categorical")]
    fn test_iter_categorical() {
        use crate::reset_string_cache;
        use crate::SINGLE_LOCK;
        let _lock = SINGLE_LOCK.lock();
        reset_string_cache();
        let ca =
            Utf8Chunked::new_from_opt_slice("", &[Some("foo"), None, Some("bar"), Some("ham")]);
        let ca = ca.cast(&DataType::Categorical).unwrap();
        let ca = ca.categorical().unwrap();
        let v: Vec<_> = ca.into_iter().collect();
        assert_eq!(v, &[Some(0), None, Some(1), Some(2)]);
    }

    #[test]
    #[ignore]
    fn test_shrink_to_fit() {
        let mut builder = Utf8ChunkedBuilder::new("foo", 2048, 100 * 2048);
        builder.append_value("foo");
        let mut arr = builder.finish();
        let before = arr
            .chunks()
            .iter()
            .map(|arr| arrow::compute::aggregate::estimated_bytes_size(arr.as_ref()))
            .sum::<usize>();
        arr.shrink_to_fit();
        let after = arr
            .chunks()
            .iter()
            .map(|arr| arrow::compute::aggregate::estimated_bytes_size(arr.as_ref()))
            .sum::<usize>();
        assert!(before > after);
    }
}