feat(editor): put caret at pointer coords when clicking on selected text element (#10970)

packages/element/src/textElement.ts

@@ -441,7 +441,8 @@ const VALID_CONTAINER_TYPES = new Set([
 
 export const isValidTextContainer = (element: {
   type: ExcalidrawElementType;
-}) => VALID_CONTAINER_TYPES.has(element.type);
+}): element is ExcalidrawTextContainer =>
+  VALID_CONTAINER_TYPES.has(element.type);
 
 export const computeContainerDimensionForBoundText = (
   dimension: number,

packages/element/src/textWrapping.ts

@@ -4,6 +4,22 @@ import { charWidth, getLineWidth } from "./textMeasurements";
 
 import type { FontString } from "./types";
 
+/**
+ * This module approximates browser-like soft wrapping for Excalidraw text.
+ *
+ * The flow is:
+ * 1. `parseTokens()` splits a hard line into breakable tokens using a unicode-aware regex.
+ * 2. `getWrappedTextLines()` reflows each hard line into one or more visual lines and
+ *    records where each visual line came from in the source text.
+ * 3. `wrapLine()` assembles tokens into lines, and `wrapWord()` handles a single token
+ *    that is wider than the available width.
+ * 4. `trimLine()` / `trimLineEndAtSoftBreak()` mirror browser behavior around trailing
+ *    whitespace so the rendered text stays consistent with what users see on canvas.
+ *
+ * Mostly, you'll want to use wrapText(). getWrappedTextLines() is for callers
+ * that need metadata such as mapping visual lines back to `originalText`
+ * for caret placement or future editor features.
+ */
 let cachedCjkRegex: RegExp | undefined;
 let cachedLineBreakRegex: RegExp | undefined;
 let cachedEmojiRegex: RegExp | undefined;
@@ -358,6 +374,10 @@ const Break = {
 
 /**
  * Breaks the line into the tokens based on the found line break opporutnities.
+ *
+ * Note: tokenization normalizes to NFC first so decomposed graphemes are treated as
+ * their composed variants for wrapping. Any code that needs exact source offsets should
+ * keep in mind that this assumes the input text is already NFC-normalized.
  */
 export const parseTokens = (line: string) => {
   const breakLineRegex = getLineBreakRegex();
@@ -370,56 +390,120 @@ export const parseTokens = (line: string) => {
 
 /**
  * Wraps the original text into the lines based on the given width.
+ *
+ * This is a convenience adapter over `getWrappedTextLines()` for call sites
+ * that only need the rendered wrapped string and not the source offsets.
  */
 export const wrapText = (
   text: string,
   font: FontString,
   maxWidth: number,
 ): string => {
+  return getWrappedTextLines(text, font, maxWidth)
+    .map((line) => line.text)
+    .join("\n");
+};
+
+/**
+ * A single rendered visual line produced from the original text.
+ *
+ * `start` and `end` are end-exclusive code-unit offsets into the original text, and do
+ * not include synthetic soft line breaks inserted by this module. If trailing whitespace
+ * was trimmed away at a wrap boundary, `end` points to the last rendered character.
+ */
+export type WrappedTextLine = {
+  text: string;
+  start: number;
+  end: number;
+};
+
+/**
+ * Splits only on existing hard line breaks and preserves original offsets.
+ */
+const getHardLineBreaks = (text: string): WrappedTextLine[] => {
+  let offset = 0;
+
+  return text.split("\n").map((line) => {
+    const start = offset;
+    const end = start + line.length;
+
+    offset = end + 1;
+
+    return {
+      text: line,
+      start,
+      end,
+    };
+  });
+};
+
+/**
+ * Returns the rendered visual lines together with their source offsets.
+ *
+ * This is the source-of-truth wrapping pipeline for callers that need more than the
+ * final wrapped string, for example caret placement or future editor/rich-text mapping.
+ */
+export const getWrappedTextLines = (
+  text: string,
+  font: FontString,
+  maxWidth: number,
+): WrappedTextLine[] => {
   // if maxWidth is not finite or NaN which can happen in case of bugs in
   // computation, we need to make sure we don't continue as we'll end up
   // in an infinite loop
   if (!Number.isFinite(maxWidth) || maxWidth < 0) {
-    return text;
+    return getHardLineBreaks(text);
   }
 
-  const lines: Array<string> = [];
-  const originalLines = text.split("\n");
+  const lines: WrappedTextLine[] = [];
+  let offset = 0;
 
-  for (const originalLine of originalLines) {
-    const currentLineWidth = getLineWidth(originalLine, font);
+  for (const originalLine of text.split("\n")) {
+    const originalLineWidth = getLineWidth(originalLine, font);
 
-    if (currentLineWidth <= maxWidth) {
-      lines.push(originalLine);
-      continue;
+    if (originalLineWidth <= maxWidth) {
+      lines.push({
+        text: originalLine,
+        start: offset,
+        end: offset + originalLine.length,
+      });
+    } else {
+      lines.push(...wrapLine(originalLine, font, maxWidth, offset));
     }
 
-    const wrappedLine = wrapLine(originalLine, font, maxWidth);
-    lines.push(...wrappedLine);
+    offset += originalLine.length + 1;
   }
 
-  return lines.join("\n");
+  return lines;
 };
 
 /**
- * Wraps the original line into the lines based on the given width.
+ * Wraps a single hard line into one or more visual lines.
+ *
+ * The line-local offsets are tracked in original-text code units so
+ * we can map the visual line back to the source.
  */
 const wrapLine = (
   line: string,
   font: FontString,
   maxWidth: number,
-): string[] => {
-  const lines: Array<string> = [];
+  lineStart: number,
+): WrappedTextLine[] => {
+  const lines: WrappedTextLine[] = [];
   const tokens = parseTokens(line);
-  const tokenIterator = tokens[Symbol.iterator]();
 
   let currentLine = "";
+  let currentLineStart = lineStart;
+  let currentLineEnd = lineStart;
   let currentLineWidth = 0;
+  // Tracks the next token's code-unit position in the original source string.
+  let tokenOffset = lineStart;
+  let tokenIndex = 0;
 
-  let iterator = tokenIterator.next();
-
-  while (!iterator.done) {
-    const token = iterator.value;
+  while (tokenIndex < tokens.length) {
+    const token = tokens[tokenIndex];
+    const tokenStart = tokenOffset;
+    const tokenEnd = tokenStart + token.length;
     const testLine = currentLine + token;
 
     // cache single codepoint whitespace, CJK or emoji width calc. as kerning should not apply here
@@ -429,37 +513,59 @@ const wrapLine = (
 
     // build up the current line, skipping length check for possibly trailing whitespaces
     if (/\s/.test(token) || testLineWidth <= maxWidth) {
+      if (!currentLine) {
+        currentLineStart = tokenStart;
+      }
       currentLine = testLine;
+      currentLineEnd = tokenEnd;
       currentLineWidth = testLineWidth;
-      iterator = tokenIterator.next();
+      tokenOffset = tokenEnd;
+      tokenIndex++;
       continue;
     }
 
     // current line is empty => just the token (word) is longer than `maxWidth` and needs to be wrapped
     if (!currentLine) {
-      const wrappedWord = wrapWord(token, font, maxWidth);
-      const trailingLine = wrappedWord[wrappedWord.length - 1] ?? "";
+      const wrappedWord = wrapWord(token, font, maxWidth, tokenStart);
+      const trailingLine = wrappedWord[wrappedWord.length - 1] ?? {
+        text: "",
+        start: tokenStart,
+        end: tokenStart,
+      };
       const precedingLines = wrappedWord.slice(0, -1);
 
       lines.push(...precedingLines);
 
       // trailing line of the wrapped word might still be joined with next token/s
-      currentLine = trailingLine;
-      currentLineWidth = getLineWidth(trailingLine, font);
-      iterator = tokenIterator.next();
+      currentLine = trailingLine.text;
+      currentLineStart = trailingLine.start;
+      currentLineEnd = trailingLine.end;
+      currentLineWidth = getLineWidth(trailingLine.text, font);
+      tokenOffset = tokenEnd;
+      tokenIndex++;
     } else {
       // push & reset, but don't iterate on the next token, as we didn't use it yet!
-      lines.push(currentLine.trimEnd());
+      lines.push(
+        trimLineEndAtSoftBreak(currentLine, currentLineStart, currentLineEnd),
+      );
 
       // purposefully not iterating and not setting `currentLine` to `token`, so that we could use a simple !currentLine check above
       currentLine = "";
+      currentLineStart = tokenStart;
+      currentLineEnd = tokenStart;
       currentLineWidth = 0;
     }
   }
 
   // iterator done, push the trailing line if exists
   if (currentLine) {
-    const trailingLine = trimLine(currentLine, font, maxWidth);
+    const trailingLine = trimLine(
+      currentLine,
+      currentLineStart,
+      currentLineEnd,
+      font,
+      maxWidth,
+    );
     lines.push(trailingLine);
   }
 
@@ -467,59 +573,100 @@ const wrapLine = (
 };
 
 /**
- * Wraps the word into the lines based on the given width.
+ * Wraps a single word that could not be placed on an empty line as-is.
  */
 const wrapWord = (
   word: string,
   font: FontString,
   maxWidth: number,
-): Array<string> => {
+  wordStart: number,
+): WrappedTextLine[] => {
   // multi-codepoint emojis are already broken apart and shouldn't be broken further
   if (getEmojiRegex().test(word)) {
-    return [word];
+    return [
+      {
+        text: word,
+        start: wordStart,
+        end: wordStart + word.length,
+      },
+    ];
   }
 
   satisfiesWordInvariant(word);
 
-  const lines: Array<string> = [];
+  const lines: WrappedTextLine[] = [];
   const chars = Array.from(word);
 
   let currentLine = "";
+  let currentLineStart = wordStart;
+  let currentLineEnd = wordStart;
   let currentLineWidth = 0;
+  let offset = wordStart;
 
   for (const char of chars) {
+    const charStart = offset;
+    const charEnd = charStart + char.length;
     const _charWidth = charWidth.calculate(char, font);
     const testLineWidth = currentLineWidth + _charWidth;
 
     if (testLineWidth <= maxWidth) {
+      if (!currentLine) {
+        currentLineStart = charStart;
+      }
       currentLine = currentLine + char;
+      currentLineEnd = charEnd;
       currentLineWidth = testLineWidth;
+      offset = charEnd;
       continue;
     }
 
     if (currentLine) {
-      lines.push(currentLine);
+      lines.push({
+        text: currentLine,
+        start: currentLineStart,
+        end: currentLineEnd,
+      });
     }
 
     currentLine = char;
+    currentLineStart = charStart;
+    currentLineEnd = charEnd;
     currentLineWidth = _charWidth;
+    offset = charEnd;
   }
 
   if (currentLine) {
-    lines.push(currentLine);
+    lines.push({
+      text: currentLine,
+      start: currentLineStart,
+      end: currentLineEnd,
+    });
   }
 
   return lines;
 };
 
 /**
- * Similarly to browsers, does not trim all trailing whitespaces, but only those exceeding the `maxWidth`.
+ * Trims trailing whitespace that is exceeding the `maxWidth`.
+ *
+ * Used for the trailing visual line of a hard line, where some trailing
+ * whitespace may still be visible if it fits into the available width.
  */
-const trimLine = (line: string, font: FontString, maxWidth: number) => {
+const trimLine = (
+  line: string,
+  start: number,
+  end: number,
+  font: FontString,
+  maxWidth: number,
+): WrappedTextLine => {
   const shouldTrimWhitespaces = getLineWidth(line, font) > maxWidth;
 
   if (!shouldTrimWhitespaces) {
-    return line;
+    return {
+      text: line,
+      start,
+      end,
+    };
   }
 
   // defensively default to `trimeEnd` in case the regex does not match
@@ -543,7 +690,30 @@ const trimLine = (line: string, font: FontString, maxWidth: number) => {
     trimmedLineWidth = testLineWidth;
   }
 
-  return trimmedLine;
+  return {
+    text: trimmedLine,
+    start,
+    end: end - (line.length - trimmedLine.length),
+  };
+};
+
+/**
+ * Used for internal soft-wrap boundaries, where trailing whitespace should not
+ * survive into the rendered line even though it still exists in the original
+ * text.
+ */
+const trimLineEndAtSoftBreak = (
+  line: string,
+  start: number,
+  end: number,
+): WrappedTextLine => {
+  const trimmedLine = line.trimEnd();
+
+  return {
+    text: trimmedLine,
+    start,
+    end: end - (line.length - trimmedLine.length),
+  };
 };
 
 /**

packages/element/tests/textWrapping.test.ts

@@ -1,4 +1,8 @@
-import { wrapText, parseTokens } from "../src/textWrapping";
+import {
+  getWrappedTextLines,
+  parseTokens,
+  wrapText,
+} from "../src/textWrapping";
 
 import type { FontString } from "../src/types";
 
@@ -102,6 +106,71 @@ describe("Test wrapText", () => {
     expect(res2).toBe(`\tA)\none\ntab\n- two\ntabs\n- 8\nspace\ns`);
   });
 
+  it("should retain original text offsets for wrapped lines", () => {
+    expect(getWrappedTextLines("Hello World!", font, 60)).toEqual([
+      {
+        text: "Hello",
+        start: 0,
+        end: 5,
+      },
+      {
+        text: "World!",
+        start: 6,
+        end: 12,
+      },
+    ]);
+  });
+
+  it("should exclude whitespace trimmed away at soft-wrap boundaries from line offsets", () => {
+    expect(getWrappedTextLines("  Hello  World", font, 90)).toEqual([
+      {
+        text: "  Hello",
+        start: 0,
+        end: 7,
+      },
+      {
+        text: "World",
+        start: 9,
+        end: 14,
+      },
+    ]);
+  });
+
+  it("should retain offsets when wrapping a single long token", () => {
+    expect(getWrappedTextLines("Excalidraw", font, 50)).toEqual([
+      {
+        text: "Excal",
+        start: 0,
+        end: 5,
+      },
+      {
+        text: "idraw",
+        start: 5,
+        end: 10,
+      },
+    ]);
+  });
+
+  it("should preserve empty hard lines in metadata", () => {
+    expect(getWrappedTextLines("A\n\nB", font, 100)).toEqual([
+      {
+        text: "A",
+        start: 0,
+        end: 1,
+      },
+      {
+        text: "",
+        start: 2,
+        end: 2,
+      },
+      {
+        text: "B",
+        start: 3,
+        end: 4,
+      },
+    ]);
+  });
+
   describe("When text is CJK", () => {
     it("should break each CJK character when width is very small", () => {
       // "안녕하세요" (Hangul) + "こんにちは世界" (Hiragana, Kanji) + "ｺﾝﾆﾁハ" (Katakana) + "你好" (Han) = "Hello Hello World Hello Hi"

packages/excalidraw/components/App.tsx

@@ -684,6 +684,11 @@ class App extends React.Component<AppProps, AppState> {
   lastPointerDownEvent: React.PointerEvent<HTMLElement> | null = null;
   lastPointerUpEvent: React.PointerEvent<HTMLElement> | PointerEvent | null =
     null;
+  // TODO this is a hack and we should ideally unify touch and pointer events
+  // and implement our own double click handling end-to-end (currently we're
+  // using a mix of native browser for click events and manual for touch -
+  // and browser doubleClick sucks to begin with)
+  lastPointerUpIsDoubleClick: boolean = false;
   lastPointerMoveEvent: PointerEvent | null = null;
   /** current frame pointer cords */
   lastPointerMoveCoords: { x: number; y: number } | null = null;
@@ -1438,6 +1443,21 @@ class App extends React.Component<AppProps, AppState> {
     return true;
   }
 
+  private isDoubleClick = (
+    lastPointerEvent:
+      | PointerEvent
+      | React.PointerEvent<HTMLElement>
+      | undefined
+      | null,
+    currentPointerEvent: PointerEvent | React.PointerEvent<HTMLElement>,
+  ) => {
+    return (
+      lastPointerEvent != null &&
+      currentPointerEvent.timeStamp - lastPointerEvent.timeStamp <=
+        TAP_TWICE_TIMEOUT
+    );
+  };
+
   private isIframeLikeElementCenter(
     el: ExcalidrawIframeLikeElement | null,
     event: React.PointerEvent<HTMLElement> | PointerEvent,
@@ -5617,8 +5637,14 @@ class App extends React.Component<AppProps, AppState> {
     element: ExcalidrawTextElement,
     {
       isExistingElement = false,
+      initialCaretSceneCoords = null,
     }: {
       isExistingElement?: boolean;
+      /**
+       * supply null if no caret positioning is desired, and instead
+       * text should be auto-selected
+       */
+      initialCaretSceneCoords?: { x: number; y: number } | null;
     },
   ) {
     const elementsMap = this.scene.getElementsMapIncludingDeleted();
@@ -5721,6 +5747,7 @@ class App extends React.Component<AppProps, AppState> {
       element,
       excalidrawContainer: this.excalidrawContainerRef.current,
       app: this,
+      initialCaretSceneCoords,
       // when text is selected, it's hard (at least on iOS) to re-position the
       // caret (i.e. deselect). There's not much use for always selecting
       // the text on edit anyway (and users can select-all from contextmenu
@@ -5744,6 +5771,68 @@ class App extends React.Component<AppProps, AppState> {
     });
   }
 
+  private getSelectedTextElement(
+    container?: ExcalidrawTextContainer | null,
+  ): NonDeleted<ExcalidrawTextElement> | null {
+    const selectedElements = this.scene.getSelectedElements(this.state);
+
+    if (selectedElements.length !== 1) {
+      return null;
+    }
+
+    const selectedElement = selectedElements[0]!;
+
+    if (isTextElement(selectedElement)) {
+      return selectedElement;
+    }
+
+    if (!container) {
+      return null;
+    }
+
+    return getBoundTextElement(
+      selectedElement,
+      this.scene.getNonDeletedElementsMap(),
+    );
+  }
+
+  private getSelectedTextEditingContainerAtPosition(
+    hitElement: NonDeletedExcalidrawElement | null,
+    sceneCoords: { x: number; y: number },
+  ): ExcalidrawTextContainer | null | undefined {
+    const selectedElements = this.scene.getSelectedElements(this.state);
+
+    if (
+      selectedElements.length !== 1 ||
+      !hitElement ||
+      hitElement.id !== selectedElements[0]!.id
+    ) {
+      return null;
+    }
+
+    const selectedElement = selectedElements[0]!;
+
+    if (isTextElement(selectedElement)) {
+      return null;
+    }
+
+    if (!isValidTextContainer(selectedElement)) {
+      return undefined;
+    }
+
+    const textElement = this.getSelectedTextElement(selectedElement);
+    const hitTextElement = this.getTextElementAtPosition(
+      sceneCoords.x,
+      sceneCoords.y,
+    );
+
+    if (!textElement || hitTextElement?.id !== textElement.id) {
+      return undefined;
+    }
+
+    return selectedElement;
+  }
+
   private getTextElementAtPosition(
     x: number,
     y: number,
@@ -5969,6 +6058,7 @@ class App extends React.Component<AppProps, AppState> {
     insertAtParentCenter = true,
     container,
     autoEdit = true,
+    initialCaretSceneCoords,
   }: {
     /** X position to insert text at */
     sceneX: number;
@@ -5978,6 +6068,7 @@ class App extends React.Component<AppProps, AppState> {
     insertAtParentCenter?: boolean;
     container?: ExcalidrawTextContainer | null;
     autoEdit?: boolean;
+    initialCaretSceneCoords?: { x: number; y: number };
   }) => {
     let shouldBindToContainer = false;
 
@@ -5998,24 +6089,9 @@ class App extends React.Component<AppProps, AppState> {
         shouldBindToContainer = true;
       }
     }
-    let existingTextElement: NonDeleted<ExcalidrawTextElement> | null = null;
-
-    const selectedElements = this.scene.getSelectedElements(this.state);
-
-    if (selectedElements.length === 1) {
-      if (isTextElement(selectedElements[0])) {
-        existingTextElement = selectedElements[0];
-      } else if (container) {
-        existingTextElement = getBoundTextElement(
-          selectedElements[0],
-          this.scene.getNonDeletedElementsMap(),
-        );
-      } else {
-        existingTextElement = this.getTextElementAtPosition(sceneX, sceneY);
-      }
-    } else {
-      existingTextElement = this.getTextElementAtPosition(sceneX, sceneY);
-    }
+    const existingTextElement =
+      this.getSelectedTextElement(container) ||
+      this.getTextElementAtPosition(sceneX, sceneY);
 
     const fontFamily =
       existingTextElement?.fontFamily || this.state.currentItemFontFamily;
@@ -6116,6 +6192,9 @@ class App extends React.Component<AppProps, AppState> {
     if (autoEdit || existingTextElement || container) {
       this.handleTextWysiwyg(element, {
         isExistingElement: !!existingTextElement,
+        initialCaretSceneCoords: existingTextElement
+          ? initialCaretSceneCoords
+          : null,
       });
     } else {
       this.setState({
@@ -6144,6 +6223,9 @@ class App extends React.Component<AppProps, AppState> {
   private handleCanvasDoubleClick = (
     event: React.MouseEvent<HTMLCanvasElement>,
   ) => {
+    if (this.state.editingTextElement) {
+      return;
+    }
     // case: double-clicking with arrow/line tool selected would both create
     // text and enter multiElement mode
     if (this.state.multiElement) {
@@ -7680,6 +7762,10 @@ class App extends React.Component<AppProps, AppState> {
     }
 
     this.removePointer(event);
+    this.lastPointerUpIsDoubleClick = this.isDoubleClick(
+      this.lastPointerUpEvent,
+      event,
+    );
     this.lastPointerUpEvent = event;
 
     if (!event.ctrlKey) {
@@ -8516,6 +8602,7 @@ class App extends React.Component<AppProps, AppState> {
       insertAtParentCenter: !event.altKey,
       container,
       autoEdit: false,
+      initialCaretSceneCoords: { x: sceneX, y: sceneY },
     });
 
     resetCursor(this.interactiveCanvas);
@@ -11001,6 +11088,35 @@ class App extends React.Component<AppProps, AppState> {
         return;
       }
 
+      const selectedTextEditingContainer =
+        this.getSelectedTextEditingContainerAtPosition(hitElement, sceneCoords);
+
+      if (
+        activeTool.type === this.state.preferredSelectionTool.type &&
+        !this.state.editingTextElement &&
+        !pointerDownState.drag.hasOccurred &&
+        !pointerDownState.hit.wasAddedToSelection &&
+        !childEvent.shiftKey &&
+        !childEvent[KEYS.CTRL_OR_CMD] &&
+        !childEvent.altKey &&
+        childEvent.pointerType !== "touch" &&
+        hitElement &&
+        ((isTextElement(hitElement) &&
+          this.state.selectedElementIds[hitElement.id] &&
+          this.scene.getSelectedElements(this.state).length === 1) ||
+          selectedTextEditingContainer)
+      ) {
+        this.startTextEditing({
+          sceneX: sceneCoords.x,
+          sceneY: sceneCoords.y,
+          container: selectedTextEditingContainer,
+          initialCaretSceneCoords: this.lastPointerUpIsDoubleClick
+            ? undefined
+            : sceneCoords,
+        });
+        return;
+      }
+
       if (!activeTool.locked && activeTool.type !== "freedraw" && newElement) {
         this.setState((prevState) => ({
           selectedElementIds: makeNextSelectedElementIds(

packages/excalidraw/components/Stats/stats.test.tsx

@@ -361,12 +361,10 @@ describe("stats for a non-generic element", () => {
     mouse.clickAt(20, 30);
     const editor = await getTextEditor();
     updateTextEditor(editor, "Hello!");
-    act(() => {
-      editor.blur();
-    });
+    Keyboard.exitTextEditor(editor);
 
     const text = h.elements[0] as ExcalidrawTextElement;
-    mouse.clickOn(text);
+    API.setSelectedElements([text]);
 
     elementStats = stats?.querySelector("#elementStats");
 

packages/excalidraw/wysiwyg/textWysiwyg.test.tsx

@@ -232,6 +232,67 @@ describe("textWysiwyg", () => {
       expect(h.elements.length).toBe(1);
     });
 
+    it("should edit selected bound text on single click", async () => {
+      const container = API.createElement({
+        type: "rectangle",
+        width: 160,
+        height: 70,
+        boundElements: [],
+      });
+      const text = API.createElement({
+        type: "text",
+        text: "Hello World!",
+        x: container.x + 20,
+        y: container.y + 20,
+        width: 120,
+        height: 25,
+        containerId: container.id,
+      });
+
+      API.setElements([container, text]);
+      API.updateElement(container, {
+        boundElements: [{ type: "text", id: text.id }],
+      });
+      API.setSelectedElements([container]);
+      UI.clickTool("selection");
+
+      mouse.clickAt(text.x + 26, text.y + 10);
+
+      const editor = await getTextEditor();
+
+      expect(editor).not.toBe(null);
+    });
+
+    it("should not edit selected bound text container when only the container was single-clicked", async () => {
+      const container = API.createElement({
+        type: "rectangle",
+        width: 160,
+        height: 70,
+        boundElements: [],
+      });
+      const text = API.createElement({
+        type: "text",
+        text: "Hello World!",
+        x: container.x + 20,
+        y: container.y + 20,
+        width: 120,
+        height: 25,
+        containerId: container.id,
+      });
+
+      API.setElements([container, text]);
+      API.updateElement(container, {
+        boundElements: [{ type: "text", id: text.id }],
+      });
+      API.setSelectedElements([container]);
+      UI.clickTool("selection");
+
+      mouse.clickAt(container.x + 5, container.y + 10);
+
+      expect(h.state.editingTextElement).toBe(null);
+      expect(await getTextEditor({ waitForEditor: false })).toBe(null);
+    });
+
     // FIXME too flaky. No one knows why.
     it.skip("should bump the version of a labeled arrow when the label is updated", async () => {
       const arrow = UI.createElement("arrow", {

packages/excalidraw/wysiwyg/textWysiwyg.tsx

@@ -10,7 +10,9 @@ import {
   isTestEnv,
   MIME_TYPES,
   applyDarkModeFilter,
+  isRTL,
 } from "@excalidraw/common";
+import { pointFrom, pointRotateRads, type Radians } from "@excalidraw/math";
 
 import {
   getTextFromElements,
@@ -33,8 +35,11 @@ import {
   getBoundTextElement,
 } from "@excalidraw/element";
 import { getTextWidth } from "@excalidraw/element";
+import { getLineHeightInPx } from "@excalidraw/element";
+import { getLineWidth } from "@excalidraw/element";
 import { normalizeText } from "@excalidraw/element";
 import { wrapText } from "@excalidraw/element";
+import { getWrappedTextLines } from "@excalidraw/element";
 import {
   isArrowElement,
   isBoundToContainer,
@@ -91,6 +96,103 @@ const getTransform = (
   return `translate(${translateX}px, ${translateY}px) scale(${zoom.value}) rotate(${degree}deg)`;
 };
 
+const getLineDirection = (text: string, offset: number) => {
+  const hardLineStart = text.lastIndexOf("\n", Math.max(0, offset - 1)) + 1;
+  const hardLineEnd = text.indexOf("\n", offset);
+  const hardLineText = text.slice(
+    hardLineStart,
+    hardLineEnd === -1 ? text.length : hardLineEnd,
+  );
+
+  return isRTL(hardLineText) ? "rtl" : "ltr";
+};
+
+const getCaretBoundaryOffsets = (text: string) => {
+  const offsets = [0];
+  let offset = 0;
+
+  for (const char of Array.from(text)) {
+    offset += char.length;
+    offsets.push(offset);
+  }
+
+  return offsets;
+};
+
+const getLineCaretOffsetFromNativeLayout = ({
+  text,
+  font,
+  lineHeightPx,
+  direction,
+  targetX,
+}: {
+  text: string;
+  font: ReturnType<typeof getFontString>;
+  lineHeightPx: number;
+  direction: "ltr" | "rtl";
+  targetX: number;
+}) => {
+  if (!text || !document.body || typeof document.createRange !== "function") {
+    return null;
+  }
+
+  const offsets = getCaretBoundaryOffsets(text);
+  const mirror = document.createElement("div");
+  const textNode = document.createTextNode(text);
+  const range = document.createRange();
+  const positions: number[] = [];
+
+  mirror.dir = direction;
+  Object.assign(mirror.style, {
+    position: "fixed",
+    top: "0",
+    left: "0",
+    margin: 0,
+    padding: 0,
+    border: 0,
+    opacity: "0",
+    pointerEvents: "none",
+    whiteSpace: "pre",
+    font,
+    lineHeight: `${lineHeightPx}px`,
+  });
+  mirror.append(textNode);
+  document.body.append(mirror);
+
+  try {
+    for (const offset of offsets) {
+      range.setStart(textNode, offset);
+      range.setEnd(textNode, offset);
+      const caretRect = range.getBoundingClientRect();
+
+      if (!Number.isFinite(caretRect.left)) {
+        return null;
+      }
+
+      positions.push(caretRect.left);
+    }
+  } catch {
+    return null;
+  } finally {
+    mirror.remove();
+  }
+
+  const leftEdge = Math.min(...positions);
+  let closestOffset = offsets[0];
+  let closestDistance = Infinity;
+
+  for (let index = 0; index < offsets.length; index++) {
+    const distance = Math.abs(positions[index] - leftEdge - targetX);
+
+    if (distance < closestDistance) {
+      closestDistance = distance;
+      closestOffset = offsets[index];
+    }
+  }
+
+  return closestOffset;
+};
+
 type SubmitHandler = () => void;
 
 export const textWysiwyg = ({
@@ -103,6 +205,7 @@ export const textWysiwyg = ({
   excalidrawContainer,
   app,
   autoSelect = true,
+  initialCaretSceneCoords = null,
 }: {
   id: ExcalidrawElement["id"];
   /**
@@ -119,7 +222,19 @@ export const textWysiwyg = ({
   excalidrawContainer: HTMLDivElement | null;
   app: App;
   autoSelect?: boolean;
+  initialCaretSceneCoords?: { x: number; y: number } | null;
 }): SubmitHandler => {
+  let currentTextLayout: {
+    angle: Radians;
+    font: ReturnType<typeof getFontString>;
+    height: number;
+    lineHeightPx: number;
+    textAlign: ExcalidrawTextElement["textAlign"];
+    width: number;
+    x: number;
+    y: number;
+  } | null = null;
+
   const textPropertiesUpdated = (
     updatedTextElement: ExcalidrawTextElement,
     editable: HTMLTextAreaElement,
@@ -254,6 +369,7 @@ export const textWysiwyg = ({
       height *= 1.05;
 
       const font = getFontString(updatedTextElement);
+      const angle = getTextElementAngle(updatedTextElement, container);
 
       // Make sure text editor height doesn't go beyond viewport
       const editorMaxHeight =
@@ -269,7 +385,7 @@ export const textWysiwyg = ({
         transform: getTransform(
           width,
           height,
-          getTextElementAngle(updatedTextElement, container),
+          angle,
           appState,
           maxWidth,
           editorMaxHeight,
@@ -283,6 +399,19 @@ export const textWysiwyg = ({
         opacity: updatedTextElement.opacity / 100,
         maxHeight: `${editorMaxHeight}px`,
       });
+      currentTextLayout = {
+        angle: angle as Radians,
+        font,
+        height: updatedTextElement.height,
+        lineHeightPx: getLineHeightInPx(
+          updatedTextElement.fontSize,
+          updatedTextElement.lineHeight,
+        ),
+        textAlign,
+        width: updatedTextElement.width,
+        x: coordX,
+        y: coordY,
+      };
       editable.scrollTop = 0;
       // For some reason updating font attribute doesn't set font family
       // hence updating font family explicitly for test environment
@@ -333,6 +462,71 @@ export const textWysiwyg = ({
   editable.value = element.originalText;
   updateWysiwygStyle();
 
+  const getCaretIndexFromInitialSceneCoords = () => {
+    if (!initialCaretSceneCoords || !currentTextLayout) {
+      return null;
+    }
+
+    const layout = currentTextLayout;
+    const center = pointFrom(
+      layout.x + layout.width / 2,
+      layout.y + layout.height / 2,
+    );
+    const [unrotatedX, unrotatedY] = pointRotateRads(
+      pointFrom(initialCaretSceneCoords.x, initialCaretSceneCoords.y),
+      center,
+      -layout.angle as Radians,
+    );
+    const localX = unrotatedX - layout.x;
+    const localY = unrotatedY - layout.y;
+    const lines = getWrappedTextLines(
+      editable.value,
+      layout.font,
+      whiteSpace === "pre-wrap" ? layout.width : Infinity,
+    );
+    const lineIndex = Math.max(
+      0,
+      Math.min(lines.length - 1, Math.floor(localY / layout.lineHeightPx)),
+    );
+    const line = lines[lineIndex];
+    const direction = getLineDirection(editable.value, line.start);
+    const lineWidth = getLineWidth(line.text, layout.font);
+    const lineStartX =
+      layout.textAlign === "center"
+        ? (layout.width - lineWidth) / 2
+        : layout.textAlign === "right"
+        ? layout.width - lineWidth
+        : 0;
+    const relativeX = localX - lineStartX;
+
+    if (!line.text) {
+      return line.start;
+    }
+
+    const lineCaretOffset = getLineCaretOffsetFromNativeLayout({
+      text: line.text,
+      font: layout.font,
+      lineHeightPx: layout.lineHeightPx,
+      direction,
+      targetX: relativeX,
+    });
+
+    return line.start + (lineCaretOffset || 0);
+  };
+
+  let pendingInitialSelection = (() => {
+    const caretIndex = getCaretIndexFromInitialSceneCoords();
+
+    if (caretIndex === null) {
+      return null;
+    }
+
+    return {
+      start: caretIndex,
+      end: caretIndex,
+    };
+  })();
+
   if (onChange) {
     editable.onpaste = async (event) => {
       // we need to synchronously get the MIME types so we can preventDefault()
@@ -696,6 +890,13 @@ export const textWysiwyg = ({
       // Otherwise, re-enable submit on blur and refocus the editor.
       editable.onblur = handleSubmit;
       editable.focus();
+      if (pendingInitialSelection) {
+        editable.setSelectionRange(
+          pendingInitialSelection.start,
+          pendingInitialSelection.end,
+        );
+        pendingInitialSelection = null;
+      }
     });
   };
 
@@ -786,7 +987,7 @@ export const textWysiwyg = ({
 
   let isDestroyed = false;
 
-  if (autoSelect) {
+  if (autoSelect && !pendingInitialSelection) {
     // select on init (focusing is done separately inside the bindBlurEvent()
     // because we need it to happen *after* the blur event from `pointerdown`)
     editable.select();