diff --git a/src/main/browser/dictation/browser-guest-dictation-text.ts b/src/main/browser/dictation/browser-guest-dictation-text.ts index 16c382f973c..b25eb98a9e0 100644 --- a/src/main/browser/dictation/browser-guest-dictation-text.ts +++ b/src/main/browser/dictation/browser-guest-dictation-text.ts @@ -1,4 +1,7 @@ -import { dictationSeparators } from '../../../shared/dictation-transcript-spacing' +import { + DICTATION_TEXT_BLOCK_SELECTOR, + dictationSeparators +} from '../../../shared/dictation-transcript-spacing' // The focused text field of this frame's document, looking through open shadow roots, or null. const FOCUSED_FIELD_EXPRESSION = `(() => { @@ -27,19 +30,30 @@ const READ_FOCUSED_FIELD_SCRIPT = `(() => { const selection = root.getSelection ? root.getSelection() : document.getSelection() if (!selection || selection.rangeCount === 0) return { before: '', after: '' } const caret = selection.getRangeAt(0) + const blockOf = (node) => { + const block = (node instanceof Element ? node : node.parentElement)?.closest(${JSON.stringify(DICTATION_TEXT_BLOCK_SELECTOR)}) + return block && field.contains(block) ? block : field + } const before = document.createRange() - before.selectNodeContents(field) + before.selectNodeContents(blockOf(caret.startContainer)) before.setEnd(caret.startContainer, caret.startOffset) const after = document.createRange() - after.selectNodeContents(field) + after.selectNodeContents(blockOf(caret.endContainer)) after.setStart(caret.endContainer, caret.endOffset) return { before: before.toString().slice(-limit), after: after.toString().slice(0, limit) } })()` // execCommand types into the frame that owns the field; webContents.insertText only reaches the top frame. +// Rich-text editors that model input themselves take a cancelable beforeinput first, as in Orca's own editors. const typeIntoFocusedFieldScript = (text: string): string => `(() => { + const text = ${JSON.stringify(text)} const field = ${FOCUSED_FIELD_EXPRESSION} - return field ? document.execCommand('insertText', false, ${JSON.stringify(text)}) : false + if (!field) return false + if (field.isContentEditable) { + const beforeInput = new InputEvent('beforeinput', { bubbles: true, cancelable: true, inputType: 'insertText', data: text }) + if (!field.dispatchEvent(beforeInput)) return true + } + return document.execCommand('insertText', false, text) })()` type FocusedFieldText = { before: string; after: string } diff --git a/src/main/browser/guest-lifecycle/input/shortcuts/browser-guest-dictation-shortcut.test.ts b/src/main/browser/guest-lifecycle/input/shortcuts/browser-guest-dictation-shortcut.test.ts index a90515addec..7905acaa995 100644 --- a/src/main/browser/guest-lifecycle/input/shortcuts/browser-guest-dictation-shortcut.test.ts +++ b/src/main/browser/guest-lifecycle/input/shortcuts/browser-guest-dictation-shortcut.test.ts @@ -47,17 +47,13 @@ function pressDictationInPage(isComposing: boolean) { } describe('dictation shortcut from a focused browser page', () => { - it('hands the chord to Orca so dictation can target the page', () => { - const { prevented, renderer } = pressDictationInPage(false) - - expect(prevented).toBe(true) - expect(renderer.send).toHaveBeenCalledWith('ui:dictationKeyDown') - }) - it('leaves the chord to the page while an IME composition owns it', () => { - const { prevented, renderer } = pressDictationInPage(true) + const composing = pressDictationInPage(true) + expect(composing.prevented).toBe(false) + expect(composing.renderer.send).not.toHaveBeenCalledWith('ui:dictationKeyDown') - expect(prevented).toBe(false) - expect(renderer.send).not.toHaveBeenCalledWith('ui:dictationKeyDown') + const typed = pressDictationInPage(false) + expect(typed.prevented).toBe(true) + expect(typed.renderer.send).toHaveBeenCalledWith('ui:dictationKeyDown') }) }) diff --git a/src/renderer/src/components/dictation/dictation-insertion-target.ts b/src/renderer/src/components/dictation/dictation-insertion-target.ts index 5d561158378..cb174ca6693 100644 --- a/src/renderer/src/components/dictation/dictation-insertion-target.ts +++ b/src/renderer/src/components/dictation/dictation-insertion-target.ts @@ -8,7 +8,10 @@ import { measureTextControlPasteByteLengthWithYield, pasteTextIntoTextControl } from '@/lib/text-control-paste' -import { dictationSeparators } from '../../../../shared/dictation-transcript-spacing' +import { + DICTATION_TEXT_BLOCK_SELECTOR, + dictationSeparators +} from '../../../../shared/dictation-transcript-spacing' import { focusedBrowserPageId } from '@/hooks/resolve-zoom-target' import { DICTATION_FOCUS_INPUT_EVENT } from './dictation-control-events' @@ -114,9 +117,6 @@ function readSelectionRangeIn(element: HTMLElement): Range | null { return range && element.contains(range.commonAncestorContainer) ? range : null } -const CONTENT_EDITABLE_BLOCK_SELECTOR = - 'p, li, h1, h2, h3, h4, h5, h6, pre, blockquote, td, th, div' - /** The selection's neighbours within its own block, so a new paragraph's start has none. */ function readContentEditableNeighbours(element: HTMLElement): { before: string; after: string } { const range = readSelectionRangeIn(element) @@ -125,7 +125,7 @@ function readContentEditableNeighbours(element: HTMLElement): { before: string; } const blockOf = (node: Node): Node => { const block = (node instanceof Element ? node : node.parentElement)?.closest( - CONTENT_EDITABLE_BLOCK_SELECTOR + DICTATION_TEXT_BLOCK_SELECTOR ) return block && element.contains(block) ? block : element } diff --git a/src/shared/dictation-transcript-spacing.ts b/src/shared/dictation-transcript-spacing.ts index 620dac490dd..8af1a19cf3d 100644 --- a/src/shared/dictation-transcript-spacing.ts +++ b/src/shared/dictation-transcript-spacing.ts @@ -1,3 +1,7 @@ +/** Blocks that bound a rich-text caret's neighbours, so a new paragraph's start has none. */ +export const DICTATION_TEXT_BLOCK_SELECTOR = + 'p, li, h1, h2, h3, h4, h5, h6, pre, blockquote, td, th, div' + const CJK_BOUNDARY_CHAR_RE = /^[\p{Script=Han}\p{Script=Hiragana}\p{Script=Katakana}\p{Script=Hangul}]$/u const NO_SPACE_BEFORE_CHAR_RE = /^[,.;:!?%。,、!?;:))\]}”’»]$/u