SkPlainTextEditor: from experimental to modules

Change-Id: I8896283ee3a57af926a43f6647e27059d52dd7a8
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/237146
Reviewed-by: Hal Canary <halcanary@google.com>
Commit-Queue: Hal Canary <halcanary@google.com>
diff --git a/modules/skplaintexteditor/src/editor.cpp b/modules/skplaintexteditor/src/editor.cpp
new file mode 100644
index 0000000..4b66d80
--- /dev/null
+++ b/modules/skplaintexteditor/src/editor.cpp
@@ -0,0 +1,512 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+
+#include "modules/skplaintexteditor/include/editor.h"
+
+#include "include/core/SkCanvas.h"
+#include "include/core/SkExecutor.h"
+#include "include/core/SkPath.h"
+#include "src/utils/SkUTF.h"
+
+#include "modules/skplaintexteditor/src/shape.h"
+
+#include <algorithm>
+
+using namespace SkPlainTextEditor;
+
+static inline SkRect offset(SkRect r, SkIPoint p) {
+    return r.makeOffset((float)p.x(), (float)p.y());
+}
+
+static constexpr SkRect kUnsetRect{-FLT_MAX, -FLT_MAX, -FLT_MAX, -FLT_MAX};
+
+static bool valid_utf8(const char* ptr, size_t size) { return SkUTF::CountUTF8(ptr, size) >= 0; }
+
+// Kind of like Python's readlines(), but without any allocation.
+// Calls f() on each line.
+// F is [](const char*, size_t) -> void
+template <typename F>
+static void readlines(const void* data, size_t size, F f) {
+    const char* start = (const char*)data;
+    const char* end = start + size;
+    const char* ptr = start;
+    while (ptr < end) {
+        while (*ptr++ != '\n' && ptr < end) {}
+        size_t len = ptr - start;
+        SkASSERT(len > 0);
+        f(start, len);
+        start = ptr;
+    }
+}
+
+static const StringSlice remove_newline(const char* str, size_t len) {
+    return SkASSERT((str != nullptr) || (len == 0)),
+           StringSlice(str, (len > 0 && str[len - 1] == '\n') ? len - 1 : len);
+}
+
+void Editor::markDirty(TextLine* line) {
+    line->fBlob = nullptr;
+    line->fShaped = false;
+    line->fWordBoundaries = std::vector<bool>();
+}
+
+void Editor::setFont(SkFont font) {
+    if (font != fFont) {
+        fFont = std::move(font);
+        fNeedsReshape = true;
+        for (auto& l : fLines) { this->markDirty(&l); }
+    }
+}
+
+void Editor::setWidth(int w) {
+    if (fWidth != w) {
+        fWidth = w;
+        fNeedsReshape = true;
+        for (auto& l : fLines) { this->markDirty(&l); }
+    }
+}
+static SkPoint to_point(SkIPoint p) { return {(float)p.x(), (float)p.y()}; }
+
+Editor::TextPosition Editor::getPosition(SkIPoint xy) {
+    Editor::TextPosition approximatePosition;
+    this->reshapeAll();
+    for (size_t j = 0; j < fLines.size(); ++j) {
+        const TextLine& line = fLines[j];
+        SkIRect lineRect = {0,
+                            line.fOrigin.y(),
+                            fWidth,
+                            j + 1 < fLines.size() ? fLines[j + 1].fOrigin.y() : INT_MAX};
+        if (const SkTextBlob* b = line.fBlob.get()) {
+            SkIRect r = b->bounds().roundOut();
+            r.offset(line.fOrigin);
+            lineRect.join(r);
+        }
+        if (!lineRect.contains(xy.x(), xy.y())) {
+            continue;
+        }
+        SkPoint pt = to_point(xy - line.fOrigin);
+        const std::vector<SkRect>& pos = line.fCursorPos;
+        for (size_t i = 0; i < pos.size(); ++i) {
+            if (pos[i] != kUnsetRect && pos[i].contains(pt.x(), pt.y())) {
+                return Editor::TextPosition{i, j};
+            }
+        }
+        approximatePosition = {xy.x() <= line.fOrigin.x() ? 0 : line.fText.size(), j};
+    }
+    return approximatePosition;
+}
+
+static inline bool is_utf8_continuation(char v) {
+    return ((unsigned char)v & 0b11000000) ==
+                               0b10000000;
+}
+
+static const char* next_utf8(const char* p, const char* end) {
+    if (p < end) {
+        do {
+            ++p;
+        } while (p < end && is_utf8_continuation(*p));
+    }
+    return p;
+}
+
+static const char* align_utf8(const char* p, const char* begin) {
+    while (p > begin && is_utf8_continuation(*p)) {
+        --p;
+    }
+    return p;
+}
+
+static const char* prev_utf8(const char* p, const char* begin) {
+    return p > begin ? align_utf8(p - 1, begin) : begin;
+}
+
+SkRect Editor::getLocation(Editor::TextPosition cursor) {
+    this->reshapeAll();
+    cursor = this->move(Editor::Movement::kNowhere, cursor);
+    if (fLines.size() > 0) {
+        const TextLine& cLine = fLines[cursor.fParagraphIndex];
+        SkRect pos = {0, 0, 0, 0};
+        if (cursor.fTextByteIndex < cLine.fCursorPos.size()) {
+            pos = cLine.fCursorPos[cursor.fTextByteIndex];
+        }
+        pos.fRight = pos.fLeft + 1;
+        pos.fLeft -= 1;
+        return offset(pos, cLine.fOrigin);
+    }
+    return SkRect{0, 0, 0, 0};
+}
+
+static size_t count_char(const StringSlice& string, char value) {
+    size_t count = 0;
+    for (char c : string) { if (c == value) { ++count; } }
+    return count;
+}
+
+Editor::TextPosition Editor::insert(TextPosition pos, const char* utf8Text, size_t byteLen) {
+    if (!valid_utf8(utf8Text, byteLen) || 0 == byteLen) {
+        return pos;
+    }
+    pos = this->move(Editor::Movement::kNowhere, pos);
+    fNeedsReshape = true;
+    if (pos.fParagraphIndex < fLines.size()) {
+        fLines[pos.fParagraphIndex].fText.insert(pos.fTextByteIndex, utf8Text, byteLen);
+        this->markDirty(&fLines[pos.fParagraphIndex]);
+    } else {
+        SkASSERT(pos.fParagraphIndex == fLines.size());
+        SkASSERT(pos.fTextByteIndex == 0);
+        fLines.push_back(Editor::TextLine(StringSlice(utf8Text, byteLen)));
+    }
+    pos = Editor::TextPosition{pos.fTextByteIndex + byteLen, pos.fParagraphIndex};
+    size_t newlinecount = count_char(fLines[pos.fParagraphIndex].fText, '\n');
+    if (newlinecount > 0) {
+        StringSlice src = std::move(fLines[pos.fParagraphIndex].fText);
+        std::vector<TextLine>::const_iterator next = fLines.begin() + pos.fParagraphIndex + 1;
+        fLines.insert(next, newlinecount, TextLine());
+        TextLine* line = &fLines[pos.fParagraphIndex];
+        readlines(src.begin(), src.size(), [&line](const char* str, size_t l) {
+            (line++)->fText = remove_newline(str, l);
+        });
+    }
+    return pos;
+}
+
+Editor::TextPosition Editor::remove(TextPosition pos1, TextPosition pos2) {
+    pos1 = this->move(Editor::Movement::kNowhere, pos1);
+    pos2 = this->move(Editor::Movement::kNowhere, pos2);
+    auto cmp = [](const Editor::TextPosition& u, const Editor::TextPosition& v) { return u < v; };
+    Editor::TextPosition start = std::min(pos1, pos2, cmp);
+    Editor::TextPosition end = std::max(pos1, pos2, cmp);
+    if (start == end || start.fParagraphIndex == fLines.size()) {
+        return start;
+    }
+    fNeedsReshape = true;
+    if (start.fParagraphIndex == end.fParagraphIndex) {
+        SkASSERT(end.fTextByteIndex > start.fTextByteIndex);
+        fLines[start.fParagraphIndex].fText.remove(
+                start.fTextByteIndex, end.fTextByteIndex - start.fTextByteIndex);
+        this->markDirty(&fLines[start.fParagraphIndex]);
+    } else {
+        SkASSERT(end.fParagraphIndex < fLines.size());
+        auto& line = fLines[start.fParagraphIndex];
+        line.fText.remove(start.fTextByteIndex,
+                          line.fText.size() - start.fTextByteIndex);
+        line.fText.insert(start.fTextByteIndex,
+                          fLines[end.fParagraphIndex].fText.begin() + end.fTextByteIndex,
+                          fLines[end.fParagraphIndex].fText.size() - end.fTextByteIndex);
+        this->markDirty(&line);
+        fLines.erase(fLines.begin() + start.fParagraphIndex + 1,
+                     fLines.begin() + end.fParagraphIndex + 1);
+    }
+    return start;
+}
+
+static void append(char** dst, size_t* count, const char* src, size_t n) {
+    if (*dst) {
+        ::memcpy(*dst, src, n);
+        *dst += n;
+    }
+    *count += n;
+}
+
+size_t Editor::copy(TextPosition pos1, TextPosition pos2, char* dst) const {
+    size_t size = 0;
+    pos1 = this->move(Editor::Movement::kNowhere, pos1);
+    pos2 = this->move(Editor::Movement::kNowhere, pos2);
+    auto cmp = [](const Editor::TextPosition& u, const Editor::TextPosition& v) { return u < v; };
+    Editor::TextPosition start = std::min(pos1, pos2, cmp);
+    Editor::TextPosition end = std::max(pos1, pos2, cmp);
+    if (start == end || start.fParagraphIndex == fLines.size()) {
+        return size;
+    }
+    if (start.fParagraphIndex == end.fParagraphIndex) {
+        SkASSERT(end.fTextByteIndex > start.fTextByteIndex);
+        auto& str = fLines[start.fParagraphIndex].fText;
+        append(&dst, &size, str.begin() + start.fTextByteIndex,
+               end.fTextByteIndex - start.fTextByteIndex);
+        return size;
+    }
+    SkASSERT(end.fParagraphIndex < fLines.size());
+    const std::vector<TextLine>::const_iterator firstP = fLines.begin() + start.fParagraphIndex;
+    const std::vector<TextLine>::const_iterator lastP  = fLines.begin() + end.fParagraphIndex;
+    const auto& first = firstP->fText;
+    const auto& last  = lastP->fText;
+
+    append(&dst, &size, first.begin() + start.fTextByteIndex, first.size() - start.fTextByteIndex);
+    for (auto line = firstP + 1; line < lastP; ++line) {
+        append(&dst, &size, "\n", 1);
+        append(&dst, &size, line->fText.begin(), line->fText.size());
+    }
+    append(&dst, &size, "\n", 1);
+    append(&dst, &size, last.begin(), end.fTextByteIndex);
+    return size;
+}
+
+static inline const char* begin(const StringSlice& s) { return s.begin(); }
+
+static inline const char* end(const StringSlice& s) { return s.end(); }
+
+static size_t align_column(const StringSlice& str, size_t p) {
+    if (p >= str.size()) {
+        return str.size();
+    }
+    return align_utf8(begin(str) + p, begin(str)) - begin(str);
+}
+
+// returns smallest i such that list[i] > value.  value > list[i-1]
+// Use a binary search since list is monotonic
+template <typename T>
+static size_t find_first_larger(const std::vector<T>& list, T value) {
+    return (size_t)(std::upper_bound(list.begin(), list.end(), value) - list.begin());
+}
+
+static size_t find_closest_x(const std::vector<SkRect>& bounds, float x, size_t b, size_t e) {
+    if (b >= e) {
+        return b;
+    }
+    SkASSERT(e <= bounds.size());
+    size_t best_index = b;
+    float best_diff = ::fabsf(bounds[best_index].x() - x);
+    for (size_t i = b + 1; i < e; ++i) {
+        float d = ::fabsf(bounds[i].x() - x);
+        if (d < best_diff) {
+            best_diff = d;
+            best_index = i;
+        }
+    }
+    return best_index;
+}
+
+Editor::TextPosition Editor::move(Editor::Movement move, Editor::TextPosition pos) const {
+    if (fLines.empty()) {
+        return {0, 0};
+    }
+    // First thing: fix possible bad input values.
+    if (pos.fParagraphIndex >= fLines.size()) {
+        pos.fParagraphIndex = fLines.size() - 1;
+        pos.fTextByteIndex = fLines[pos.fParagraphIndex].fText.size();
+    } else {
+        pos.fTextByteIndex = align_column(fLines[pos.fParagraphIndex].fText, pos.fTextByteIndex);
+    }
+
+    SkASSERT(pos.fParagraphIndex < fLines.size());
+    SkASSERT(pos.fTextByteIndex <= fLines[pos.fParagraphIndex].fText.size());
+
+    SkASSERT(pos.fTextByteIndex == fLines[pos.fParagraphIndex].fText.size() ||
+             !is_utf8_continuation(fLines[pos.fParagraphIndex].fText.begin()[pos.fTextByteIndex]));
+
+    switch (move) {
+        case Editor::Movement::kNowhere:
+            break;
+        case Editor::Movement::kLeft:
+            if (0 == pos.fTextByteIndex) {
+                if (pos.fParagraphIndex > 0) {
+                    --pos.fParagraphIndex;
+                    pos.fTextByteIndex = fLines[pos.fParagraphIndex].fText.size();
+                }
+            } else {
+                const auto& str = fLines[pos.fParagraphIndex].fText;
+                pos.fTextByteIndex =
+                    prev_utf8(begin(str) + pos.fTextByteIndex, begin(str)) - begin(str);
+            }
+            break;
+        case Editor::Movement::kRight:
+            if (fLines[pos.fParagraphIndex].fText.size() == pos.fTextByteIndex) {
+                if (pos.fParagraphIndex + 1 < fLines.size()) {
+                    ++pos.fParagraphIndex;
+                    pos.fTextByteIndex = 0;
+                }
+            } else {
+                const auto& str = fLines[pos.fParagraphIndex].fText;
+                pos.fTextByteIndex =
+                    next_utf8(begin(str) + pos.fTextByteIndex, end(str)) - begin(str);
+            }
+            break;
+        case Editor::Movement::kHome:
+            {
+                const std::vector<size_t>& list = fLines[pos.fParagraphIndex].fLineEndOffsets;
+                size_t f = find_first_larger(list, pos.fTextByteIndex);
+                pos.fTextByteIndex = f > 0 ? list[f - 1] : 0;
+            }
+            break;
+        case Editor::Movement::kEnd:
+            {
+                const std::vector<size_t>& list = fLines[pos.fParagraphIndex].fLineEndOffsets;
+                size_t f = find_first_larger(list, pos.fTextByteIndex);
+                if (f < list.size()) {
+                    pos.fTextByteIndex = list[f] > 0 ? list[f] - 1 : 0;
+                } else {
+                    pos.fTextByteIndex = fLines[pos.fParagraphIndex].fText.size();
+                }
+            }
+            break;
+        case Editor::Movement::kUp:
+            {
+                SkASSERT(pos.fTextByteIndex < fLines[pos.fParagraphIndex].fCursorPos.size());
+                float x = fLines[pos.fParagraphIndex].fCursorPos[pos.fTextByteIndex].left();
+                const std::vector<size_t>& list = fLines[pos.fParagraphIndex].fLineEndOffsets;
+                size_t f = find_first_larger(list, pos.fTextByteIndex);
+                // list[f] > value.  value > list[f-1]
+                if (f > 0) {
+                    // not the first line in paragraph.
+                    pos.fTextByteIndex = find_closest_x(fLines[pos.fParagraphIndex].fCursorPos, x,
+                                                        (f == 1) ? 0 : list[f - 2],
+                                                        list[f - 1]);
+                } else if (pos.fParagraphIndex > 0) {
+                    --pos.fParagraphIndex;
+                    const auto& newLine = fLines[pos.fParagraphIndex];
+                    size_t r = newLine.fLineEndOffsets.size();
+                    if (r > 0) {
+                        pos.fTextByteIndex = find_closest_x(newLine.fCursorPos, x,
+                                                            newLine.fLineEndOffsets[r - 1],
+                                                            newLine.fCursorPos.size());
+                    } else {
+                        pos.fTextByteIndex = find_closest_x(newLine.fCursorPos, x, 0,
+                                                            newLine.fCursorPos.size());
+                    }
+                }
+                pos.fTextByteIndex =
+                    align_column(fLines[pos.fParagraphIndex].fText, pos.fTextByteIndex);
+            }
+            break;
+        case Editor::Movement::kDown:
+            {
+                const std::vector<size_t>& list = fLines[pos.fParagraphIndex].fLineEndOffsets;
+                float x = fLines[pos.fParagraphIndex].fCursorPos[pos.fTextByteIndex].left();
+
+                size_t f = find_first_larger(list, pos.fTextByteIndex);
+                if (f < list.size()) {
+                    const auto& bounds = fLines[pos.fParagraphIndex].fCursorPos;
+                    pos.fTextByteIndex = find_closest_x(bounds, x, list[f],
+                                                        f + 1 < list.size() ? list[f + 1]
+                                                                            : bounds.size());
+                } else if (pos.fParagraphIndex + 1 < fLines.size()) {
+                    ++pos.fParagraphIndex;
+                    const auto& bounds = fLines[pos.fParagraphIndex].fCursorPos;
+                    const std::vector<size_t>& l2 = fLines[pos.fParagraphIndex].fLineEndOffsets;
+                    pos.fTextByteIndex = find_closest_x(bounds, x, 0,
+                                                        l2.size() > 0 ? l2[0] : bounds.size());
+                } else {
+                    pos.fTextByteIndex = fLines[pos.fParagraphIndex].fText.size();
+                }
+                pos.fTextByteIndex =
+                    align_column(fLines[pos.fParagraphIndex].fText, pos.fTextByteIndex);
+            }
+            break;
+        case Editor::Movement::kWordLeft:
+            {
+                if (pos.fTextByteIndex == 0) {
+                    pos = this->move(Editor::Movement::kLeft, pos);
+                    break;
+                }
+                const std::vector<bool>& words = fLines[pos.fParagraphIndex].fWordBoundaries;
+                SkASSERT(words.size() == fLines[pos.fParagraphIndex].fText.size());
+                do {
+                    --pos.fTextByteIndex;
+                } while (pos.fTextByteIndex > 0 && !words[pos.fTextByteIndex]);
+            }
+            break;
+        case Editor::Movement::kWordRight:
+            {
+                const StringSlice& text = fLines[pos.fParagraphIndex].fText;
+                if (pos.fTextByteIndex == text.size()) {
+                    pos = this->move(Editor::Movement::kRight, pos);
+                    break;
+                }
+                const std::vector<bool>& words = fLines[pos.fParagraphIndex].fWordBoundaries;
+                SkASSERT(words.size() == text.size());
+                do {
+                    ++pos.fTextByteIndex;
+                } while (pos.fTextByteIndex < text.size() && !words[pos.fTextByteIndex]);
+            }
+            break;
+
+    }
+    return pos;
+}
+
+void Editor::paint(SkCanvas* c, PaintOpts options) {
+    this->reshapeAll();
+    if (!c) {
+        return;
+    }
+
+    c->drawPaint(SkPaint(options.fBackgroundColor));
+
+    SkPaint selection = SkPaint(options.fSelectionColor);
+    auto cmp = [](const Editor::TextPosition& u, const Editor::TextPosition& v) { return u < v; };
+    for (TextPosition pos = std::min(options.fSelectionBegin, options.fSelectionEnd, cmp),
+                      end = std::max(options.fSelectionBegin, options.fSelectionEnd, cmp);
+         pos < end;
+         pos = this->move(Editor::Movement::kRight, pos))
+    {
+        SkASSERT(pos.fParagraphIndex < fLines.size());
+        const TextLine& l = fLines[pos.fParagraphIndex];
+        c->drawRect(offset(l.fCursorPos[pos.fTextByteIndex], l.fOrigin), selection);
+    }
+
+    if (fLines.size() > 0) {
+        c->drawRect(Editor::getLocation(options.fCursor), SkPaint(options.fCursorColor));
+    }
+
+    SkPaint foreground = SkPaint(options.fForegroundColor);
+    for (const TextLine& line : fLines) {
+        if (line.fBlob) {
+            c->drawTextBlob(line.fBlob.get(), line.fOrigin.x(), line.fOrigin.y(), foreground);
+        }
+    }
+}
+
+void Editor::reshapeAll() {
+    if (fNeedsReshape) {
+        if (fLines.empty()) {
+            fLines.push_back(TextLine());
+        }
+        float shape_width = (float)(fWidth);
+        #ifdef SK_EDITOR_GO_FAST
+        SkSemaphore semaphore;
+        std::unique_ptr<SkExecutor> executor = SkExecutor::MakeFIFOThreadPool(100);
+        int jobCount = 0;
+        for (TextLine& line : fLines) {
+            if (!line.fShaped) {
+                executor->add([&]() {
+                    ShapeResult result = Shape(line.fText.begin(), line.fText.size(),
+                                               fFont, fLocale, shape_width);
+                    line.fBlob           = std::move(result.blob);
+                    line.fLineEndOffsets = std::move(result.lineBreakOffsets);
+                    line.fCursorPos      = std::move(result.glyphBounds);
+                    line.fWordBoundaries = std::move(result.wordBreaks);
+                    line.fHeight         = result.verticalAdvance;
+                    line.fShaped = true;
+                    semaphore.signal();
+                }
+                ++jobCount;
+            });
+        }
+        while (jobCount-- > 0) { semaphore.wait(); }
+        #else
+        int i = 0;
+        for (TextLine& line : fLines) {
+            if (!line.fShaped) {
+                ShapeResult result = Shape(line.fText.begin(), line.fText.size(),
+                                           fFont, fLocale, shape_width);
+                line.fBlob           = std::move(result.blob);
+                line.fLineEndOffsets = std::move(result.lineBreakOffsets);
+                line.fCursorPos      = std::move(result.glyphBounds);
+                line.fWordBoundaries = std::move(result.wordBreaks);
+                line.fHeight         = result.verticalAdvance;
+                line.fShaped = true;
+            }
+            ++i;
+        }
+        #endif
+        int y = 0;
+        for (TextLine& line : fLines) {
+            line.fOrigin = {0, y};
+            y += line.fHeight;
+        }
+        fHeight = y;
+        fNeedsReshape = false;
+    }
+}
+
diff --git a/modules/skplaintexteditor/src/shape.cpp b/modules/skplaintexteditor/src/shape.cpp
new file mode 100644
index 0000000..9acba0d
--- /dev/null
+++ b/modules/skplaintexteditor/src/shape.cpp
@@ -0,0 +1,294 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+
+#include "modules/skplaintexteditor/src/shape.h"
+
+#include "include/core/SkFont.h"
+#include "include/core/SkFontMetrics.h"
+#include "include/core/SkPoint.h"
+#include "include/core/SkRefCnt.h"
+#include "include/core/SkScalar.h"
+#include "include/core/SkString.h"
+#include "include/core/SkTextBlob.h"
+#include "include/core/SkTypes.h"
+#include "include/private/SkTFitsIn.h"
+#include "modules/skplaintexteditor/src/word_boundaries.h"
+#include "modules/skshaper/include/SkShaper.h"
+#include "src/core/SkTextBlobPriv.h"
+#include "src/utils/SkUTF.h"
+
+#include <limits.h>
+#include <string.h>
+
+
+using namespace SkPlainTextEditor;
+
+namespace {
+class RunHandler final : public SkShaper::RunHandler {
+public:
+    RunHandler(const char* utf8Text, size_t) : fUtf8Text(utf8Text) {}
+    using RunCallback = void (*)(void* context,
+                                 const char* utf8Text,
+                                  size_t utf8TextBytes,
+                                  size_t glyphCount,
+                                  const SkGlyphID* glyphs,
+                                  const SkPoint* positions,
+                                  const uint32_t* clusters,
+                                  const SkFont& font);
+    void setRunCallback(RunCallback f, void* context) {
+        fCallbackContext = context;
+        fCallbackFunction = f;
+    }
+
+    sk_sp<SkTextBlob> makeBlob();
+    SkPoint endPoint() const { return fOffset; }
+    SkPoint finalPosition() const { return fCurrentPosition; }
+
+    void beginLine() override;
+    void runInfo(const RunInfo&) override;
+    void commitRunInfo() override;
+    SkShaper::RunHandler::Buffer runBuffer(const RunInfo&) override;
+    void commitRunBuffer(const RunInfo&) override;
+    void commitLine() override;
+
+    const std::vector<size_t>& lineEndOffsets() const { return fLineEndOffsets; }
+
+    SkRect finalRect(const SkFont& font) const {
+        if (0 == fMaxRunAscent || 0 == fMaxRunDescent) {
+            SkFontMetrics metrics;
+            font.getMetrics(&metrics);
+            return {fCurrentPosition.x(),
+                    fCurrentPosition.y(),
+                    fCurrentPosition.x() + font.getSize(),
+                    fCurrentPosition.y() + metrics.fDescent - metrics.fAscent};
+        } else {
+            return {fCurrentPosition.x(),
+                    fCurrentPosition.y() + fMaxRunAscent,
+                    fCurrentPosition.x() + font.getSize(),
+                    fCurrentPosition.y() + fMaxRunDescent};
+        }
+    }
+
+
+private:
+    SkTextBlobBuilder fBuilder;
+    std::vector<size_t> fLineEndOffsets;
+    const SkGlyphID* fCurrentGlyphs = nullptr;
+    const SkPoint* fCurrentPoints = nullptr;
+    void* fCallbackContext = nullptr;
+    RunCallback fCallbackFunction = nullptr;
+    char const * const fUtf8Text;
+    size_t fTextOffset = 0;
+    uint32_t* fClusters = nullptr;
+    int fClusterOffset = 0;
+    int fGlyphCount = 0;
+    SkScalar fMaxRunAscent = 0;
+    SkScalar fMaxRunDescent = 0;
+    SkScalar fMaxRunLeading = 0;
+    SkPoint fCurrentPosition = {0, 0};
+    SkPoint fOffset = {0, 0};
+};
+}  // namespace
+
+void RunHandler::beginLine() {
+    fCurrentPosition = fOffset;
+    fMaxRunAscent = 0;
+    fMaxRunDescent = 0;
+    fMaxRunLeading = 0;
+}
+
+void RunHandler::runInfo(const SkShaper::RunHandler::RunInfo& info) {
+    SkFontMetrics metrics;
+    info.fFont.getMetrics(&metrics);
+    fMaxRunAscent = SkTMin(fMaxRunAscent, metrics.fAscent);
+    fMaxRunDescent = SkTMax(fMaxRunDescent, metrics.fDescent);
+    fMaxRunLeading = SkTMax(fMaxRunLeading, metrics.fLeading);
+}
+
+void RunHandler::commitRunInfo() {
+    fCurrentPosition.fY -= fMaxRunAscent;
+}
+
+SkShaper::RunHandler::Buffer RunHandler::runBuffer(const RunInfo& info) {
+    int glyphCount = SkTFitsIn<int>(info.glyphCount) ? info.glyphCount : INT_MAX;
+    int utf8RangeSize = SkTFitsIn<int>(info.utf8Range.size()) ? info.utf8Range.size() : INT_MAX;
+
+    const auto& runBuffer = SkTextBlobBuilderPriv::AllocRunTextPos(&fBuilder, info.fFont, glyphCount,
+                                                                   utf8RangeSize, SkString());
+    fCurrentGlyphs = runBuffer.glyphs;
+    fCurrentPoints = runBuffer.points();
+
+    if (runBuffer.utf8text && fUtf8Text) {
+        memcpy(runBuffer.utf8text, fUtf8Text + info.utf8Range.begin(), utf8RangeSize);
+    }
+    fClusters = runBuffer.clusters;
+    fGlyphCount = glyphCount;
+    fClusterOffset = info.utf8Range.begin();
+
+    return {runBuffer.glyphs,
+            runBuffer.points(),
+            nullptr,
+            runBuffer.clusters,
+            fCurrentPosition};
+}
+
+void RunHandler::commitRunBuffer(const RunInfo& info) {
+    // for (size_t i = 0; i < info.glyphCount; ++i) {
+    //     SkASSERT(fClusters[i] >= info.utf8Range.begin());
+    //     // this fails for khmer example.
+    //     SkASSERT(fClusters[i] <  info.utf8Range.end());
+    // }
+    if (fCallbackFunction) {
+        fCallbackFunction(fCallbackContext,
+                          fUtf8Text,
+                          info.utf8Range.end(),
+                          info.glyphCount,
+                          fCurrentGlyphs,
+                          fCurrentPoints,
+                          fClusters,
+                          info.fFont);
+    }
+    SkASSERT(0 <= fClusterOffset);
+    for (int i = 0; i < fGlyphCount; ++i) {
+        SkASSERT(fClusters[i] >= (unsigned)fClusterOffset);
+        fClusters[i] -= fClusterOffset;
+    }
+    fCurrentPosition += info.fAdvance;
+    fTextOffset = SkTMax(fTextOffset, info.utf8Range.end());
+}
+
+void RunHandler::commitLine() {
+    if (fLineEndOffsets.empty() || fTextOffset > fLineEndOffsets.back()) {
+        // Ensure that fLineEndOffsets is monotonic.
+        fLineEndOffsets.push_back(fTextOffset);
+    }
+    fOffset += { 0, fMaxRunDescent + fMaxRunLeading - fMaxRunAscent };
+}
+
+sk_sp<SkTextBlob> RunHandler::makeBlob() {
+    return fBuilder.make();
+}
+
+static SkRect selection_box(const SkFontMetrics& metrics,
+                            float advance,
+                            SkPoint pos) {
+    if (fabsf(advance) < 1.0f) {
+        advance = copysignf(1.0f, advance);
+    }
+    return SkRect{pos.x(),
+                  pos.y() + metrics.fAscent,
+                  pos.x() + advance,
+                  pos.y() + metrics.fDescent}.makeSorted();
+}
+
+static void set_character_bounds(void* context,
+                                 const char* utf8Text,
+                                 size_t utf8TextBytes,
+                                 size_t glyphCount,
+                                 const SkGlyphID* glyphs,
+                                 const SkPoint* positions,
+                                 const uint32_t* clusters,
+                                 const SkFont& font)
+{
+    SkASSERT(context);
+    SkASSERT(glyphCount > 0);
+    SkRect* cursors = (SkRect*)context;
+
+    SkFontMetrics metrics;
+    font.getMetrics(&metrics);
+    std::unique_ptr<float[]> advances(new float[glyphCount]);
+    font.getWidths(glyphs, glyphCount, advances.get());
+
+    // Loop over each cluster in this run.
+    size_t clusterStart = 0;
+    for (size_t glyphIndex = 0; glyphIndex < glyphCount; ++glyphIndex) {
+        if (glyphIndex + 1 < glyphCount  // more glyphs
+            && clusters[glyphIndex] == clusters[glyphIndex + 1]) {
+            continue; // multi-glyph cluster
+        }
+        unsigned textBegin = clusters[glyphIndex];
+        unsigned textEnd = utf8TextBytes;
+        for (size_t i = 0; i < glyphCount; ++i) {
+            if (clusters[i] >= textEnd) {
+                textEnd = clusters[i] + 1;
+            }
+        }
+        for (size_t i = 0; i < glyphCount; ++i) {
+            if (clusters[i] > textBegin && clusters[i] < textEnd) {
+                textEnd = clusters[i];
+                if (textEnd == textBegin + 1) { break; }
+            }
+        }
+        SkASSERT(glyphIndex + 1 > clusterStart);
+        unsigned clusterGlyphCount = glyphIndex + 1 - clusterStart;
+        const SkPoint* clusterGlyphPositions = &positions[clusterStart];
+        const float* clusterAdvances = &advances[clusterStart];
+        clusterStart = glyphIndex + 1;  // for next loop
+
+        SkRect clusterBox = selection_box(metrics, clusterAdvances[0], clusterGlyphPositions[0]);
+        for (unsigned i = 1; i < clusterGlyphCount; ++i) { // multiple glyphs
+            clusterBox.join(selection_box(metrics, clusterAdvances[i], clusterGlyphPositions[i]));
+        }
+        if (textBegin + 1 == textEnd) {  // single byte, fast path.
+            cursors[textBegin] = clusterBox;
+            continue;
+        }
+        int textCount = textEnd - textBegin;
+        int codePointCount = SkUTF::CountUTF8(utf8Text + textBegin, textCount);
+        if (codePointCount == 1) {  // single codepoint, fast path.
+            cursors[textBegin] = clusterBox;
+            continue;
+        }
+
+        float width = clusterBox.width() / codePointCount;
+        SkASSERT(width > 0);
+        const char* ptr = utf8Text + textBegin;
+        const char* end = utf8Text + textEnd;
+        float x = clusterBox.left();
+        while (ptr < end) {  // for each codepoint in cluster
+            const char* nextPtr = ptr;
+            SkUTF::NextUTF8(&nextPtr, end);
+            int firstIndex = ptr - utf8Text;
+            float nextX = x + width;
+            cursors[firstIndex] = SkRect{x, clusterBox.top(), nextX, clusterBox.bottom()};
+            x = nextX;
+            ptr = nextPtr;
+        }
+    }
+}
+
+ShapeResult SkPlainTextEditor::Shape(const char* utf8Text,
+                          size_t textByteLen,
+                          const SkFont& font,
+                          const char* locale,
+                          float width)
+{
+    ShapeResult result;
+    if (SkUTF::CountUTF8(utf8Text, textByteLen) < 0) {
+        utf8Text = nullptr;
+        textByteLen = 0;
+    }
+    std::unique_ptr<SkShaper> shaper = SkShaper::Make();
+    float height = font.getSpacing();
+    RunHandler runHandler(utf8Text, textByteLen);
+    if (textByteLen) {
+        result.glyphBounds.resize(textByteLen);
+        for (SkRect& c : result.glyphBounds) {
+            c = SkRect{-FLT_MAX, -FLT_MAX, -FLT_MAX, -FLT_MAX};
+        }
+        runHandler.setRunCallback(set_character_bounds, result.glyphBounds.data());
+        // TODO: make use of locale in shaping.
+        shaper->shape(utf8Text, textByteLen, font, true, width, &runHandler);
+        if (runHandler.lineEndOffsets().size() > 1) {
+            result.lineBreakOffsets = runHandler.lineEndOffsets();
+            SkASSERT(result.lineBreakOffsets.size() > 0);
+            result.lineBreakOffsets.pop_back();
+        }
+        height = std::max(height, runHandler.endPoint().y());
+        result.blob = runHandler.makeBlob();
+    }
+    result.glyphBounds.push_back(runHandler.finalRect(font));
+    result.verticalAdvance = (int)ceilf(height);
+    result.wordBreaks = GetUtf8WordBoundaries(utf8Text, textByteLen, locale);
+    return result;
+}
diff --git a/modules/skplaintexteditor/src/shape.h b/modules/skplaintexteditor/src/shape.h
new file mode 100644
index 0000000..c5585eb
--- /dev/null
+++ b/modules/skplaintexteditor/src/shape.h
@@ -0,0 +1,26 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+
+#include "include/core/SkFont.h"
+#include "include/core/SkTextBlob.h"
+
+#include <cstddef>
+#include <vector>
+
+namespace SkPlainTextEditor {
+
+struct ShapeResult {
+    sk_sp<SkTextBlob> blob;
+    std::vector<std::size_t> lineBreakOffsets;
+    std::vector<SkRect> glyphBounds;
+    std::vector<bool> wordBreaks;
+    int verticalAdvance;
+};
+
+ShapeResult Shape(const char* ut8text,
+                  size_t textByteLen,
+                  const SkFont& font,
+                  const char* locale,
+                  float width);
+
+}
diff --git a/modules/skplaintexteditor/src/stringslice.cpp b/modules/skplaintexteditor/src/stringslice.cpp
new file mode 100644
index 0000000..3f7da00
--- /dev/null
+++ b/modules/skplaintexteditor/src/stringslice.cpp
@@ -0,0 +1,82 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+
+#include "modules/skplaintexteditor/include/stringslice.h"
+
+#include <algorithm>
+#include <cassert>
+#include <cstdlib>
+#include <cstring>
+
+using namespace SkPlainTextEditor;
+
+void StringSlice::FreeWrapper::operator()(void* t) { std::free(t); }
+
+StringSlice::StringSlice(StringSlice&& that)
+    : fPtr(std::move(that.fPtr))
+    , fLength(that.fLength)
+    , fCapacity(that.fCapacity)
+{
+    that.fLength = 0;
+    that.fCapacity = 0;
+}
+
+StringSlice& StringSlice::operator=(StringSlice&& that) {
+    if (this != &that) {
+        this->~StringSlice();
+        new (this)StringSlice(std::move(that));
+    }
+    return *this;
+}
+
+StringSlice& StringSlice::operator=(const StringSlice& that) {
+    if (this != &that) {
+        fLength = 0;
+        if (that.size() > 0) {
+            this->insert(0, that.begin(), that.size());
+        }
+    }
+    return *this;
+}
+
+void StringSlice::insert(std::size_t offset, const char* text, std::size_t length) {
+    if (length) {
+        offset = std::min(fLength, offset);
+        this->reserve(fLength + length);
+        char* s = fPtr.get();
+        assert(s);
+        if (offset != fLength) {
+            std::memmove(s + offset + length, s + offset, fLength - offset);
+        }
+        if (text) {
+            std::memcpy(s + offset, text, length);
+        } else {
+            std::memset(s + offset, 0, length);
+        }
+        fLength += length;
+    }
+}
+
+void StringSlice::remove(std::size_t offset, std::size_t length) {
+    if (length && offset < fLength) {
+        length = std::min(length, fLength - offset);
+        assert(length > 0);
+        assert(length + offset <= fLength);
+        if (length + offset < fLength) {
+            char* s = fPtr.get();
+            assert(s);
+            std::memmove(s + offset, s + offset + length, fLength - (length + offset));
+        }
+        fLength -= length;
+    }
+}
+
+void StringSlice::realloc(std::size_t size) {
+    // round up to multiple of (1 << kBits) bytes
+    static constexpr unsigned kBits = 4;
+    fCapacity = size ? (((size - 1) >> kBits) + 1) << kBits : 0;
+    assert(fCapacity % (1u << kBits) == 0);
+    assert(fCapacity >= size);
+    fPtr.reset((char*)std::realloc(fPtr.release(), fCapacity));
+    assert(fCapacity >= fLength);
+}
diff --git a/modules/skplaintexteditor/src/word_boundaries.cpp b/modules/skplaintexteditor/src/word_boundaries.cpp
new file mode 100644
index 0000000..17c8cd3
--- /dev/null
+++ b/modules/skplaintexteditor/src/word_boundaries.cpp
@@ -0,0 +1,50 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+
+#include "modules/skplaintexteditor/src/word_boundaries.h"
+
+#include <unicode/brkiter.h>
+#include <unicode/unistr.h>
+
+#include <memory>
+
+std::vector<bool> GetUtf8WordBoundaries(const char* begin, size_t byteCount, const char* locale) {
+    static constexpr UBreakIteratorType kIteratorType = UBRK_WORD;
+    struct UTextCloser {
+        void operator()(UText* p) { (void)utext_close(p); }
+    };
+    struct UBreakCloser {
+        void operator()(UBreakIterator* p) { (void)ubrk_close(p); }
+    };
+
+    std::vector<bool> result;
+    if (0 == byteCount) {
+        return result;
+    }
+    result.resize(byteCount);
+
+    UText utf8UText = UTEXT_INITIALIZER;
+    UErrorCode errorCode = U_ZERO_ERROR;
+    (void)utext_openUTF8(&utf8UText, begin, byteCount, &errorCode);
+    std::unique_ptr<UText, UTextCloser> autoclose1(&utf8UText);
+    if (U_FAILURE(errorCode)) {
+        return result;
+    }
+    UBreakIterator* iter = ubrk_open(kIteratorType, locale, nullptr, 0, &errorCode);
+    std::unique_ptr<UBreakIterator, UBreakCloser> autoclose2(iter);
+    if (U_FAILURE(errorCode)) {
+        return result;
+    }
+    ubrk_setUText(iter, &utf8UText, &errorCode);
+    if (U_FAILURE(errorCode)) {
+        return result;
+    }
+    int pos = ubrk_first(iter);
+    while (pos != icu::BreakIterator::DONE) {
+        if ((unsigned)pos < (unsigned)byteCount) {
+            result[pos] = true;
+        }
+        pos = ubrk_next(iter);
+    }
+    return result;
+}
diff --git a/modules/skplaintexteditor/src/word_boundaries.h b/modules/skplaintexteditor/src/word_boundaries.h
new file mode 100644
index 0000000..b6541df
--- /dev/null
+++ b/modules/skplaintexteditor/src/word_boundaries.h
@@ -0,0 +1,12 @@
+// Copyright 2019 Google LLC.
+// Use of this source code is governed by a BSD-style license that can be found in the LICENSE file.
+#ifndef word_boundaries_DEFINED
+#define word_boundaries_DEFINED
+
+#include <cstddef>
+#include <vector>
+
+// TODO: Decide if this functionality should be moved into SkShaper as an extra utility.
+std::vector<bool> GetUtf8WordBoundaries(const char* begin, std::size_t byteLen, const char* locale);
+
+#endif  // word_boundaries_DEFINED