Offline Dictionary & Spellchecker Overhaul

Offline Dictionary & Spellchecker Overhaul

Accurate, dependable spellchecking is vital for publication publishing. Previous proofing iterations relied on third-party remote API services that introduced network latency, rate limits, missing everyday words (falsely flagging common terms like "highlight" and "highlighted"), offset drift that caused squiggly underlines to start mid-word or span into adjacent blocks, and unrelated suggestions. Version 5.8.5 delivered a complete architectural overhaul, bundling a dedicated 274,000+ word curated offline English lexicon, direct DOM text-node token scanning, and sub-millisecond local phonetic suggestions.

📖 Instant Offline Spellchecking with Contextual SuggestionsMisspelled words receive instant local proofing underlines with zero network latency and accurate suggestions on right-click.

Comprehensive Curated Offline English Lexicon (274,000+ Words)

Version 5.8.5 eliminated remote network proofing dependencies entirely:

• High-Speed Curated Lexicon: Bundled tools/proofing/dictionary.js combining SCOWL (standard Chromium/Firefox dictionary) and ENABLE1, containing 274,757 curated words loaded into an instant O(1) Set.

• Zero False Positives for Everyday Words: Words like highlight, highlighted, highlights, highlighter, and highlighting are fully recognized.

• US, UK & Commonwealth Orthography: Built-in support for both American and British variations (such as colour, favourite, theatre, realise, centre, dialogue, and travelling).

• Specialized Prepress & Publishing Vocabulary: Comprehensive vocabulary including bleed, kerning, leading, tracking, cmyk, dpi, ppi, vector, raster, imposition, and dropcap.

Direct Text-Node Token Scanning (Zero Offset Drift)

Underline misalignment and mid-word squiggles were resolved at the DOM architecture level:

• The Root Cause of Drift: Remote APIs calculated character offsets against innerText (which inserts newline characters for block elements like <div>, <p>, and <br>). In-app traversals counted textContent without newlines, causing DOM ranges to drift progressively across lines.

• Direct TreeWalker Scanning: Upgraded paintHighlights() in tools/proofing/spellcheck-engine.js to scan DOM text nodes directly using TreeWalker(NodeFilter.SHOW_TEXT) and regex word boundary tokenization (/[A-Za-z]+(?:['’][A-Za-z]+)?/g).

• Precise Range Construction: Range objects are constructed strictly on (textNode, match.index, match.index + rawWord.length), guaranteeing squiggly underlines never start mid-word, underline partial words, or span across block boundaries.

Instant Phonetic & Damerau-Levenshtein Suggestions

Contextual replacement suggestions are generated locally in sub-millisecond time:

• Local Phonetic Matching: Right-clicking misspelled words calculates intelligent phonetic and edit-distance matches locally (e.g. highlite → highlight, recieve → receive, seperate → separate).

• Casing Preservation: Capitalized inputs produce capitalized suggestions (Hightlight → Highlight), and uppercase inputs generate uppercase replacements.

• Full Options Suite Preservation: Preserved Custom Dictionary management with word deletion, Ignored Words, Ignored Grammar Rules, and Autocorrect shortcuts.

Tip: You can add specialized company names and proprietary technical terms directly to your Custom Dictionary in the Options dialog.