part of dart._engine;
// Copyright 2013 The Flutter Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.



typedef SegmentationResult = ({Uint32List words, Uint32List graphemes, Uint32List breaks});

// The cache numbers below were picked based on the following logic.
//
// Most paragraphs in an app are small (e.g. icons, button labels, etc). These
// paragraphs are also cheap to cache. So we cache a lot of them. 100,000 of
// them amounts to a worst case of 5MB (10-character long text + words uint list
// + graphemes uint list + breaks uint list).
//
// Large paragraphs are less common (a handful per page), but are expensive to
// cache. So we cache fewer of them. 20 of them at a length of 50,000 characters
// amount to a memory usage of 5MB (50,000-character long text + words uint list
// + graphemes uint list + breaks uint list).
//
// Medium paragraphs are somewhere in between. 10,000 of them amount to a worst
// case of 5MB (100-character long text + words uint list + graphemes uint list
// + breaks uint list).

typedef SegmentationCacheSpec = ({int cacheSize, int maxTextLength});

const SegmentationCacheSpec kSmallParagraphCacheSpec = (cacheSize: 100000, maxTextLength: 10);
const SegmentationCacheSpec kMediumParagraphCacheSpec = (cacheSize: 10000, maxTextLength: 100);
const SegmentationCacheSpec kLargeParagraphCacheSpec = (cacheSize: 20, maxTextLength: 50000);

typedef SegmentationCache =
    ({
      LruCache<String, SegmentationResult> small,
      LruCache<String, SegmentationResult> medium,
      LruCache<String, SegmentationResult> large,
    });

/// Caches segmentation results for small, medium and large paragraphts.
///
/// Paragraphs are frequently re-created because of style or font changes, while
/// their text contents remain the same. This cache is effective at
/// short-circuiting the segmentation of such paragraphs.
final SegmentationCache segmentationCache = (
  small: LruCache<String, SegmentationResult>(kSmallParagraphCacheSpec.cacheSize),
  medium: LruCache<String, SegmentationResult>(kMediumParagraphCacheSpec.cacheSize),
  large: LruCache<String, SegmentationResult>(kLargeParagraphCacheSpec.cacheSize),
);

extension SegmentationCacheExtensions on SegmentationCache {
  /// Gets the appropriate cache for the given [text].
  LruCache<String, SegmentationResult>? getCacheForText(String text) {
    if (text.length <= kSmallParagraphCacheSpec.maxTextLength) {
      return small;
    }
    if (text.length <= kMediumParagraphCacheSpec.maxTextLength) {
      return medium;
    }
    if (text.length <= kLargeParagraphCacheSpec.maxTextLength) {
      return large;
    }
    return null;
  }

  /// Clears all the caches.
  void clear() {
    small.clear();
    medium.clear();
    large.clear();
  }
}

/// Injects required ICU data into the [builder].
///
/// This should only be used with the CanvasKit Chromium variant that's compiled
/// without ICU data.
void injectClientICU(SkParagraphBuilder builder) {
  assert(
    canvasKit.ParagraphBuilder.RequiresClientICU(),
    'This method should only be used with the CanvasKit Chromium variant.',
  );

  final SegmentationResult result = segmentText(builder.getText());
  builder.setWordsUtf16(result.words);
  builder.setGraphemeBreaksUtf16(result.graphemes);
  builder.setLineBreaksUtf16(result.breaks);
}

/// Segments the [text] into words, graphemes and line breaks.
///
/// Caches results in [segmentationCache].
SegmentationResult segmentText(String text) {
  final LruCache<String, SegmentationResult>? cache = segmentationCache.getCacheForText(text);
  final SegmentationResult? cachedResult = cache?[text];

  final SegmentationResult result;
  if (cachedResult != null) {
    result = cachedResult;
  } else {
    result = (
      words: fragmentUsingIntlSegmenter(text, IntlSegmenterGranularity.word),
      graphemes: fragmentUsingIntlSegmenter(text, IntlSegmenterGranularity.grapheme),
      breaks: fragmentUsingV8LineBreaker(text),
    );
  }

  // Save or promote to most recently used.
  cache?.cache(text, result);
  return result;
}

/// The granularity at which to segment text.
///
/// To find all supported granularities, see:
/// - https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Intl/Segmenter/Segmenter
enum IntlSegmenterGranularity { grapheme, word }

final Map<IntlSegmenterGranularity, DomSegmenter> _intlSegmenters =
    <IntlSegmenterGranularity, DomSegmenter>{
      IntlSegmenterGranularity.grapheme: createIntlSegmenter(granularity: 'grapheme'),
      IntlSegmenterGranularity.word: createIntlSegmenter(granularity: 'word'),
    };

Uint32List fragmentUsingIntlSegmenter(String text, IntlSegmenterGranularity granularity) {
  final DomSegmenter segmenter = _intlSegmenters[granularity]!;
  final DomIteratorWrapper<DomSegment> iterator = segmenter.segment(text).iterator();

  final List<int> breaks = <int>[];
  while (iterator.moveNext()) {
    breaks.add(iterator.current.index);
  }
  breaks.add(text.length);
  return Uint32List.fromList(breaks);
}

// These are the soft/hard line break values expected by Skia's SkParagraph.
const int _kSoftLineBreak = 0;
const int _kHardLineBreak = 1;

final DomV8BreakIterator _v8LineBreaker = createV8BreakIterator();

Uint32List fragmentUsingV8LineBreaker(String text) {
  final List<LineBreakFragment> fragments = breakLinesUsingV8BreakIterator(
    text,
    text.toJS,
    _v8LineBreaker,
  );

  final int size = (fragments.length + 1) * 2;
  final Uint32List typedArray = Uint32List(size);

  typedArray[0] = 0; // start index
  typedArray[1] = _kSoftLineBreak; // break type

  for (int i = 0; i < fragments.length; i++) {
    final LineBreakFragment fragment = fragments[i];
    final int uint32Index = 2 + i * 2;
    typedArray[uint32Index] = fragment.end;
    typedArray[uint32Index + 1] =
        fragment.type == LineBreakType.mandatory ? _kHardLineBreak : _kSoftLineBreak;
  }

  return typedArray;
}