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

//! Contains the compression attribute definition
//! and methods to compress and decompress data.


// private modules make non-breaking changes easier
mod zip;
mod rle;
mod piz;
mod pxr24;
mod b44;


use std::convert::TryInto;
use std::mem::size_of;
use half::f16;
use crate::meta::attribute::{IntegerBounds, SampleType, ChannelList};
use crate::error::{Result, Error, usize_to_i32};
use crate::meta::header::Header;


/// A byte vector.
pub type ByteVec = Vec<u8>;

/// A byte slice.
pub type Bytes<'s> = &'s [u8];

/// Specifies which compression method to use.
/// Use uncompressed data for fastest loading and writing speeds.
/// Use RLE compression for fast loading and writing with slight memory savings.
/// Use ZIP compression for slow processing with large memory savings.
#[derive(Debug, Clone, Copy, PartialEq)]
pub enum Compression {

    /// Store uncompressed values.
    /// Produces large files that can be read and written very quickly.
    /// Consider using RLE instead, as it provides some compression with almost equivalent speed.
    Uncompressed,

    /// Produces slightly smaller files
    /// that can still be read and written rather quickly.
    /// The compressed file size is usually between 60 and 75 percent of the uncompressed size.
    /// Works best for images with large flat areas, such as masks and abstract graphics.
    /// This compression method is lossless.
    RLE,

    /// Uses ZIP compression to compress each line. Slowly produces small images
    /// which can be read with moderate speed. This compression method is lossless.
    /// Might be slightly faster but larger than `ZIP16´.
    ZIP1,  // TODO ZIP { individual_lines: bool, compression_level: Option<u8> }  // TODO specify zip compression level?

    /// Uses ZIP compression to compress blocks of 16 lines. Slowly produces small images
    /// which can be read with moderate speed. This compression method is lossless.
    /// Might be slightly slower but smaller than `ZIP1´.
    ZIP16, // TODO collapse with ZIP1

    /// PIZ compression works well for noisy and natural images. Works better with larger tiles.
    /// Only supported for flat images, but not for deep data.
    /// This compression method is lossless.
    // A wavelet transform is applied to the pixel data, and the result is Huffman-
    // encoded. This scheme tends to provide the best compression ratio for the types of
    // images that are typically processed at Industrial Light & Magic. Files are
    // compressed and decompressed at roughly the same speed. For photographic
    // images with film grain, the files are reduced to between 35 and 55 percent of their
    // uncompressed size.
    // PIZ compression works well for scan-line based files, and also for tiled files with
    // large tiles, but small tiles do not shrink much. (PIZ-compressed data start with a
    // relatively long header; if the input to the compressor is short, adding the header
    // tends to offset any size reduction of the input.)
    PIZ,

    /// Like `ZIP1`, but reduces precision of `f32` images to `f24`.
    /// Therefore, this is lossless compression for `f16` and `u32` data, lossy compression for `f32` data.
    /// This compression method works well for depth
    /// buffers and similar images, where the possible range of values is very large, but
    /// where full 32-bit floating-point accuracy is not necessary. Rounding improves
    /// compression significantly by eliminating the pixels' 8 least significant bits, which
    /// tend to be very noisy, and therefore difficult to compress.
    /// This produces really small image files. Only supported for flat images, not for deep data.
    // After reducing 32-bit floating-point data to 24 bits by rounding (while leaving 16-bit
    // floating-point data unchanged), differences between horizontally adjacent pixels
    // are compressed with zlib, similar to ZIP. PXR24 compression preserves image
    // channels of type HALF and UINT exactly, but the relative error of FLOAT data
    // increases to about ???.
    PXR24, // TODO specify zip compression level?

    /// This is a lossy compression method for f16 images.
    /// It's the predecessor of the `B44A` compression,
    /// which has improved compression rates for uniformly colored areas.
    /// You should probably use `B44A` instead of the plain `B44`.
    ///
    /// Only supported for flat images, not for deep data.
    // lossy 4-by-4 pixel block compression,
    // flat fields are compressed more
    // Channels of type HALF are split into blocks of four by four pixels or 32 bytes. Each
    // block is then packed into 14 bytes, reducing the data to 44 percent of their
    // uncompressed size. When B44 compression is applied to RGB images in
    // combination with luminance/chroma encoding (see below), the size of the
    // compressed pixels is about 22 percent of the size of the original RGB data.
    // Channels of type UINT or FLOAT are not compressed.
    // Decoding is fast enough to allow real-time playback of B44-compressed OpenEXR
    // image sequences on commodity hardware.
    // The size of a B44-compressed file depends on the number of pixels in the image,
    // but not on the data in the pixels. All images with the same resolution and the same
    // set of channels have the same size. This can be advantageous for systems that
    // support real-time playback of image sequences; the predictable file size makes it
    // easier to allocate space on storage media efficiently.
    // B44 compression is only supported for flat images.
    B44, // TODO B44 { optimize_uniform_areas: bool }

    /// This is a lossy compression method for f16 images.
    /// All f32 and u32 channels will be stored without compression.
    /// All the f16 pixels are divided into 4x4 blocks.
    /// Each block is then compressed as a whole.
    ///
    /// The 32 bytes of a block will require only ~14 bytes after compression,
    /// independent of the actual pixel contents. With chroma subsampling,
    /// a block will be compressed to ~7 bytes.
    /// Uniformly colored blocks will be compressed to ~3 bytes.
    ///
    /// The 512 bytes of an f32 block will not be compressed at all.
    ///
    /// Should be fast enough for realtime playback.
    /// Only supported for flat images, not for deep data.
    B44A, // TODO collapse with B44

    /// __This lossy compression is not yet supported by this implementation.__
    // lossy DCT based compression, in blocks
    // of 32 scanlines. More efficient for partial buffer access.
    DWAA(Option<f32>), // TODO does this have a default value? make this non optional? default Compression Level setting is 45.0

    /// __This lossy compression is not yet supported by this implementation.__
    // lossy DCT based compression, in blocks
    // of 256 scanlines. More efficient space
    // wise and faster to decode full frames
    // than DWAA_COMPRESSION.
    DWAB(Option<f32>), // TODO collapse with B44. default Compression Level setting is 45.0
}

impl std::fmt::Display for Compression {
    fn fmt(&self, formatter: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        write!(formatter, "{} compression", match self {
            Compression::Uncompressed => "no",
            Compression::RLE => "rle",
            Compression::ZIP1 => "zip line",
            Compression::ZIP16 => "zip block",
            Compression::B44 => "b44",
            Compression::B44A => "b44a",
            Compression::DWAA(_) => "dwaa",
            Compression::DWAB(_) => "dwab",
            Compression::PIZ => "piz",
            Compression::PXR24 => "pxr24",
        })
    }
}



impl Compression {

    /// Compress the image section of bytes.
    pub fn compress_image_section(self, header: &Header, uncompressed_native_endian: ByteVec, pixel_section: IntegerBounds) -> Result<ByteVec> {
        let max_tile_size = header.max_block_pixel_size();

        assert!(pixel_section.validate(Some(max_tile_size)).is_ok(), "decompress tile coordinate bug");
        if header.deep { assert!(self.supports_deep_data()) }

        use self::Compression::*;
        let compressed_little_endian = match self {
            Uncompressed => {
                return Ok(convert_current_to_little_endian(
                    uncompressed_native_endian, &header.channels, pixel_section
                ))
            },

            // we need to clone here, because we might have to fallback to the uncompressed data later (when compressed data is larger than raw data)
            ZIP16 => zip::compress_bytes(&header.channels, uncompressed_native_endian.clone(), pixel_section),
            ZIP1 => zip::compress_bytes(&header.channels, uncompressed_native_endian.clone(), pixel_section),
            RLE => rle::compress_bytes(&header.channels, uncompressed_native_endian.clone(), pixel_section),
            PIZ => piz::compress(&header.channels, uncompressed_native_endian.clone(), pixel_section),
            PXR24 => pxr24::compress(&header.channels, uncompressed_native_endian.clone(), pixel_section),
            B44 => b44::compress(&header.channels, uncompressed_native_endian.clone(), pixel_section, false),
            B44A => b44::compress(&header.channels, uncompressed_native_endian.clone(), pixel_section, true),
            _ => return Err(Error::unsupported(format!("yet unimplemented compression method: {}", self)))
        };

        let compressed_little_endian = compressed_little_endian.map_err(|_|
            Error::invalid(format!("pixels cannot be compressed ({})", self))
        )?;

        if self == Uncompressed || compressed_little_endian.len() < uncompressed_native_endian.len() {
            // only write compressed if it actually is smaller than raw
            Ok(compressed_little_endian)
        }
        else {
            // if we do not use compression, manually convert uncompressed data
            Ok(convert_current_to_little_endian(uncompressed_native_endian, &header.channels, pixel_section))
        }
    }

    /// Decompress the image section of bytes.
    pub fn decompress_image_section(self, header: &Header, compressed: ByteVec, pixel_section: IntegerBounds, pedantic: bool) -> Result<ByteVec> {
        let max_tile_size = header.max_block_pixel_size();

        assert!(pixel_section.validate(Some(max_tile_size)).is_ok(), "decompress tile coordinate bug");
        if header.deep { assert!(self.supports_deep_data()) }

        let expected_byte_size = pixel_section.size.area() * header.channels.bytes_per_pixel; // FIXME this needs to account for subsampling anywhere

        // note: always true where self == Uncompressed
        if compressed.len() == expected_byte_size {
            // the compressed data was larger than the raw data, so the small raw data has been written
            Ok(convert_little_endian_to_current(compressed, &header.channels, pixel_section))
        }
        else {
            use self::Compression::*;
            let bytes = match self {
                Uncompressed => Ok(convert_little_endian_to_current(compressed, &header.channels, pixel_section)),
                ZIP16 => zip::decompress_bytes(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                ZIP1 => zip::decompress_bytes(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                RLE => rle::decompress_bytes(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                PIZ => piz::decompress(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                PXR24 => pxr24::decompress(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                B44 | B44A => b44::decompress(&header.channels, compressed, pixel_section, expected_byte_size, pedantic),
                _ => return Err(Error::unsupported(format!("yet unimplemented compression method: {}", self)))
            };

            // map all errors to compression errors
            let bytes = bytes
                .map_err(|decompression_error| match decompression_error {
                    Error::NotSupported(message) =>
                        Error::unsupported(format!("yet unimplemented compression special case ({})", message)),

                    error => Error::invalid(format!(
                        "compressed {:?} data ({})",
                        self, error.to_string()
                    )),
                })?;

            if bytes.len() != expected_byte_size {
                Err(Error::invalid("decompressed data"))
            }

            else { Ok(bytes) }
        }
    }

    /// For scan line images and deep scan line images, one or more scan lines may be
    /// stored together as a scan line block. The number of scan lines per block
    /// depends on how the pixel data are compressed.
    pub fn scan_lines_per_block(self) -> usize {
        use self::Compression::*;
        match self {
            Uncompressed | RLE   | ZIP1    => 1,
            ZIP16 | PXR24                  => 16,
            PIZ   | B44   | B44A | DWAA(_) => 32,
            DWAB(_)                        => 256,
        }
    }

    /// Deep data can only be compressed using RLE or ZIP compression.
    pub fn supports_deep_data(self) -> bool {
        use self::Compression::*;
        match self {
            Uncompressed | RLE | ZIP1 => true,
            _ => false,
        }
    }

    /// Most compression methods will reconstruct the exact pixel bytes,
    /// but some might throw away unimportant data for specific types of samples.
    pub fn is_lossless_for(self, sample_type: SampleType) -> bool {
        use self::Compression::*;
        match self {
            PXR24 => sample_type != SampleType::F32, // pxr reduces f32 to f24
            B44 | B44A => sample_type != SampleType::F16, // b44 only compresses f16 values, others are left uncompressed
            Uncompressed | RLE | ZIP1 | ZIP16 | PIZ => true,
            DWAB(_) | DWAA(_) => false,
        }
    }

    /// Most compression methods will reconstruct the exact pixel bytes,
    /// but some might throw away unimportant data in some cases.
    pub fn may_loose_data(self) -> bool {
        use self::Compression::*;
        match self {
            Uncompressed | RLE | ZIP1 | ZIP16 | PIZ => false,
            PXR24 | B44 | B44A | DWAB(_) | DWAA(_)  => true,
        }
    }

    /// Most compression methods will reconstruct the exact pixel bytes,
    /// but some might replace NaN with zeroes.
    pub fn supports_nan(self) -> bool {
        use self::Compression::*;
        match self {
            B44 | B44A | DWAB(_) | DWAA(_) => false, // TODO dwa might support it?
            _ => true
        }
    }

}

// see https://github.com/AcademySoftwareFoundation/openexr/blob/6a9f8af6e89547bcd370ae3cec2b12849eee0b54/OpenEXR/IlmImf/ImfMisc.cpp#L1456-L1541

#[allow(unused)] // allows the extra parameters to be unused
fn convert_current_to_little_endian(mut bytes: ByteVec, channels: &ChannelList, rectangle: IntegerBounds) -> ByteVec {
    #[cfg(target = "big_endian")]
    reverse_block_endianness(&mut byte_vec, channels, rectangle);

    bytes
}

#[allow(unused)] // allows the extra parameters to be unused
fn convert_little_endian_to_current(mut bytes: ByteVec, channels: &ChannelList, rectangle: IntegerBounds) -> ByteVec {
    #[cfg(target = "big_endian")]
    reverse_block_endianness(&mut bytes, channels, rectangle);

    bytes
}

#[allow(unused)] // unused when on little endian system
fn reverse_block_endianness(bytes: &mut [u8], channels: &ChannelList, rectangle: IntegerBounds){
    let mut remaining_bytes: &mut [u8] = bytes;

    for y in rectangle.position.y() .. rectangle.end().y() {
        for channel in &channels.list {
            let line_is_subsampled = mod_p(y, usize_to_i32(channel.sampling.y())) != 0;
            if line_is_subsampled { continue; }

            let sample_count = rectangle.size.width() / channel.sampling.x();

            match channel.sample_type {
                SampleType::F16 => remaining_bytes = chomp_convert_n::<f16>(reverse_2_bytes, remaining_bytes, sample_count),
                SampleType::F32 => remaining_bytes = chomp_convert_n::<f32>(reverse_4_bytes, remaining_bytes, sample_count),
                SampleType::U32 => remaining_bytes = chomp_convert_n::<u32>(reverse_4_bytes, remaining_bytes, sample_count),
            }
        }
    }

    #[inline]
    fn chomp_convert_n<T>(convert_single_value: fn(&mut[u8]), mut bytes: &mut [u8], count: usize) -> &mut [u8] {
        let type_size = size_of::<T>();
        let (line_bytes, rest) = bytes.split_at_mut(count * type_size);
        let value_byte_chunks = line_bytes.chunks_exact_mut(type_size);

        for value_bytes in value_byte_chunks {
            convert_single_value(value_bytes);
        }

        rest
    }

    debug_assert!(remaining_bytes.is_empty(), "not all bytes were converted to little endian");
}

#[inline]
fn reverse_2_bytes(bytes: &mut [u8]){
    // this code seems like it could be optimized easily by the compiler
    let two_bytes: [u8; 2] = bytes.try_into().expect("invalid byte count");
    bytes.copy_from_slice(&[two_bytes[1], two_bytes[0]]);
}

#[inline]
fn reverse_4_bytes(bytes: &mut [u8]){
    let four_bytes: [u8; 4] = bytes.try_into().expect("invalid byte count");
    bytes.copy_from_slice(&[four_bytes[3], four_bytes[2], four_bytes[1], four_bytes[0]]);
}

#[inline]
fn div_p (x: i32, y: i32) -> i32 {
    if x >= 0 {
        if y >= 0 { x  / y }
        else { -(x  / -y) }
    }
    else {
        if y >= 0 { -((y-1-x) / y) }
        else { (-y-1-x) / -y }
    }
}

#[inline]
fn mod_p(x: i32, y: i32) -> i32 {
    x - y * div_p(x, y)
}

/// A collection of functions used to prepare data for compression.
mod optimize_bytes {

    /// Integrate over all differences to the previous value in order to reconstruct sample values.
    pub fn differences_to_samples(buffer: &mut [u8]) {
        // The naive implementation is very simple:
        //
        // for index in 1..buffer.len() {
        //    buffer[index] = (buffer[index - 1] as i32 + buffer[index] as i32 - 128) as u8;
        // }
        //
        // But we process elements in pairs to take advantage of instruction-level parallelism.
        // When computations within a pair do not depend on each other, they can be processed in parallel.
        // Since this function is responsible for a very large chunk of execution time,
        // this tweak alone improves decoding performance of RLE images by 20%.
        if let Some(first) = buffer.get(0) {
            let mut previous = *first as i16;
            for chunk in &mut buffer[1..].chunks_exact_mut(2) {
                // no bounds checks here due to indices and chunk size being constant
                let diff0 = chunk[0] as i16;
                let diff1 = chunk[1] as i16;
                // these two computations do not depend on each other, unlike in the naive version,
                // so they can be executed by the CPU in parallel via instruction-level parallelism
                let sample0 = (previous + diff0 - 128) as u8;
                let sample1 = (previous + diff0 + diff1 - 128 * 2) as u8;
                chunk[0] = sample0;
                chunk[1] = sample1;
                previous = sample1 as i16;
            }
            // handle the remaining element at the end not processed by the loop over pairs, if present
            for elem in &mut buffer[1..].chunks_exact_mut(2).into_remainder().iter_mut() {
                let sample = (previous + *elem as i16 - 128) as u8;
                *elem = sample;
                previous = sample as i16;
            }
        }
    }

    /// Derive over all values in order to produce differences to the previous value.
    pub fn samples_to_differences(buffer: &mut [u8]){
        // naive version:
        // for index in (1..buffer.len()).rev() {
        //     buffer[index] = (buffer[index] as i32 - buffer[index - 1] as i32 + 128) as u8;
        // }
        //
        // But we process elements in batches to take advantage of autovectorization.
        // If the target platform has no vector instructions (e.g. 32-bit ARM without `-C target-cpu=native`)
        // this will instead take advantage of instruction-level parallelism.
        if let Some(first) = buffer.get(0) {
            let mut previous = *first as i16;
            // Chunk size is 16 because we process bytes (8 bits),
            // and 8*16 = 128 bits is the size of a typical SIMD register.
            // Even WASM has 128-bit SIMD registers.
            for chunk in &mut buffer[1..].chunks_exact_mut(16) {
                // no bounds checks here due to indices and chunk size being constant
                let sample0 = chunk[0] as i16;
                let sample1 = chunk[1] as i16;
                let sample2 = chunk[2] as i16;
                let sample3 = chunk[3] as i16;
                let sample4 = chunk[4] as i16;
                let sample5 = chunk[5] as i16;
                let sample6 = chunk[6] as i16;
                let sample7 = chunk[7] as i16;
                let sample8 = chunk[8] as i16;
                let sample9 = chunk[9] as i16;
                let sample10 = chunk[10] as i16;
                let sample11 = chunk[11] as i16;
                let sample12 = chunk[12] as i16;
                let sample13 = chunk[13] as i16;
                let sample14 = chunk[14] as i16;
                let sample15 = chunk[15] as i16;
                // Unlike in decoding, computations in here are truly independent from each other,
                // which enables the compiler to vectorize this loop.
                // Even if the target platform has no vector instructions,
                // so using more parallelism doesn't imply doing more work,
                // and we're not really limited in how wide we can go.
                chunk[0] = (sample0 - previous + 128) as u8;
                chunk[1] = (sample1 - sample0 + 128) as u8;
                chunk[2] = (sample2 - sample1 + 128) as u8;
                chunk[3] = (sample3 - sample2 + 128) as u8;
                chunk[4] = (sample4 - sample3 + 128) as u8;
                chunk[5] = (sample5 - sample4 + 128) as u8;
                chunk[6] = (sample6 - sample5 + 128) as u8;
                chunk[7] = (sample7 - sample6 + 128) as u8;
                chunk[8] = (sample8 - sample7 + 128) as u8;
                chunk[9] = (sample9 - sample8 + 128) as u8;
                chunk[10] = (sample10 - sample9 + 128) as u8;
                chunk[11] = (sample11 - sample10 + 128) as u8;
                chunk[12] = (sample12 - sample11 + 128) as u8;
                chunk[13] = (sample13 - sample12 + 128) as u8;
                chunk[14] = (sample14 - sample13 + 128) as u8;
                chunk[15] = (sample15 - sample14 + 128) as u8;
                previous = sample15;
            }
            // Handle the remaining element at the end not processed by the loop over batches, if present
            // This is what the iterator-based version of this function would look like without vectorization
            for elem in &mut buffer[1..].chunks_exact_mut(16).into_remainder().iter_mut() {
                let diff = (*elem as i16 - previous + 128) as u8;
                previous = *elem as i16;
                *elem = diff;
            }
        }
    }

    use std::cell::Cell;
    thread_local! {
        // A buffer for reusing between invocations of interleaving and deinterleaving.
        // Allocating memory is cheap, but zeroing or otherwise initializing it is not.
        // Doing it hundreds of times (once per block) would be expensive.
        // This optimization brings down the time spent in interleaving from 15% to 5%.
        static SCRATCH_SPACE: Cell<Vec<u8>> = Cell::new(Vec::new());
    }

    fn with_reused_buffer<F>(length: usize, mut func: F) where F: FnMut(&mut [u8]) {
        SCRATCH_SPACE.with(|scratch_space| {
            // reuse a buffer if we've already initialized one
            let mut buffer = scratch_space.take();
            if buffer.len() < length {
                // Efficiently create a zeroed Vec by requesting zeroed memory from the OS.
                // This is slightly faster than a `memcpy()` plus `memset()` that would happen otherwise,
                // but is not a big deal either way since it's not a hot codepath.
                buffer = vec![0u8; length];
            }

            // call the function
            func(&mut buffer[..length]);

            // save the internal buffer for reuse
            scratch_space.set(buffer);
        });
    }

    /// Interleave the bytes such that the second half of the array is every other byte.
    pub fn interleave_byte_blocks(separated: &mut [u8]) {
        with_reused_buffer(separated.len(), |interleaved| {

            // Split the two halves that we are going to interleave.
            let (first_half, second_half) = separated.split_at((separated.len() + 1) / 2);
            // The first half can be 1 byte longer than the second if the length of the input is odd,
            // but the loop below only processes numbers in pairs.
            // To handle it, preserve the last element of the first slice, to be handled after the loop.
            let first_half_last = first_half.last();
            // Truncate the first half to match the lenght of the second one; more optimizer-friendly
            let first_half_iter = &first_half[..second_half.len()];

            // Main loop that performs the interleaving
            for ((first, second), interleaved) in first_half_iter.iter().zip(second_half.iter())
                .zip(interleaved.chunks_exact_mut(2)) {
                    // The length of each chunk is known to be 2 at compile time,
                    // and each index is also a constant.
                    // This allows the compiler to remove the bounds checks.
                    interleaved[0] = *first;
                    interleaved[1] = *second;
            }

            // If the length of the slice was odd, restore the last element of the first half that we saved
            if interleaved.len() % 2 == 1 {
                if let Some(value) = first_half_last {
                    // we can unwrap() here because we just checked that the lenght is non-zero:
                    // `% 2 == 1` will fail for zero
                    *interleaved.last_mut().unwrap() = *value;
                }
            }

            // write out the results
            separated.copy_from_slice(&interleaved);
        });
    }

/// Separate the bytes such that the second half contains every other byte.
/// This performs deinterleaving - the inverse of interleaving.
pub fn separate_bytes_fragments(source: &mut [u8]) {
    with_reused_buffer(source.len(), |separated| {

        // Split the two halves that we are going to interleave.
        let (first_half, second_half) = separated.split_at_mut((source.len() + 1) / 2);
        // The first half can be 1 byte longer than the second if the length of the input is odd,
        // but the loop below only processes numbers in pairs.
        // To handle it, preserve the last element of the input, to be handled after the loop.
        let last = source.last();
        let first_half_iter = &mut first_half[..second_half.len()];

        // Main loop that performs the deinterleaving
        for ((first, second), interleaved) in first_half_iter.iter_mut().zip(second_half.iter_mut())
            .zip(source.chunks_exact(2)) {
                // The length of each chunk is known to be 2 at compile time,
                // and each index is also a constant.
                // This allows the compiler to remove the bounds checks.
                *first = interleaved[0];
                *second = interleaved[1];
        }

        // If the length of the slice was odd, restore the last element of the input that we saved
        if source.len() % 2 == 1 {
            if let Some(value) = last {
                // we can unwrap() here because we just checked that the lenght is non-zero:
                // `% 2 == 1` will fail for zero
                *first_half.last_mut().unwrap() = *value;
            }
        }

        // write out the results
        source.copy_from_slice(&separated);
    });
}


    #[cfg(test)]
    pub mod test {

        #[test]
        fn roundtrip_interleave(){
            let source = vec![ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ];
            let mut modified = source.clone();

            super::separate_bytes_fragments(&mut modified);
            super::interleave_byte_blocks(&mut modified);

            assert_eq!(source, modified);
        }

        #[test]
        fn roundtrip_derive(){
            let source = vec![ 0, 1, 2, 7, 4, 5, 6, 7, 13, 9, 10 ];
            let mut modified = source.clone();

            super::samples_to_differences(&mut modified);
            super::differences_to_samples(&mut modified);

            assert_eq!(source, modified);
        }

    }
}


#[cfg(test)]
pub mod test {
    use super::*;
    use crate::meta::attribute::ChannelDescription;
    use crate::block::samples::IntoNativeSample;

    #[test]
    fn roundtrip_endianness_mixed_channels(){
        let a32 = ChannelDescription::new("A", SampleType::F32, true);
        let y16 = ChannelDescription::new("Y", SampleType::F16, true);
        let channels = ChannelList::new(smallvec![ a32, y16 ]);

        let data = vec![
            23582740683_f32.to_ne_bytes().as_slice(),
            35827420683_f32.to_ne_bytes().as_slice(),
            27406832358_f32.to_f16().to_ne_bytes().as_slice(),
            74062358283_f32.to_f16().to_ne_bytes().as_slice(),

            52582740683_f32.to_ne_bytes().as_slice(),
            45827420683_f32.to_ne_bytes().as_slice(),
            15406832358_f32.to_f16().to_ne_bytes().as_slice(),
            65062358283_f32.to_f16().to_ne_bytes().as_slice(),
        ].into_iter().flatten().map(|x| *x).collect();

        roundtrip_convert_endianness(
            data, &channels,
            IntegerBounds::from_dimensions((2, 2))
        );
    }

    fn roundtrip_convert_endianness(
        current_endian: ByteVec, channels: &ChannelList, rectangle: IntegerBounds
    ){
        let little_endian = convert_current_to_little_endian(
            current_endian.clone(), channels, rectangle
        );

        let current_endian_decoded = convert_little_endian_to_current(
            little_endian.clone(), channels, rectangle
        );

        assert_eq!(current_endian, current_endian_decoded, "endianness conversion failed");
    }
}