[Psi-Devel] [Patches] [Patch] Spell checker doesn't handle abbreviated words

[Michail Pishchagin]

On 14.07.2008, at 13:34, Norman Rasmussen wrote:
>> Could you please give some examples?
>
> discussed at http://conference.psi-im.org/psi@conference.psi-im.org/2008/06/28.html
> flysprayed at http://flyspray.psi-im.org/task/859
>
> words that incorrectly highlight: doesn't shouldn't
> words that incorrectly don't highlight: can'do good'spelling
> note: can't and won't don't highlight because can and won are valid  
> words

Ahhar, we need to put a unit-test around our highlighter.

>> One other problem that you approach doesn't fix is that aspell  
>> erroneously highlights numbers. And there's one more problem as  
>> well: we shouldn't highlight words on top of which the cursor is  
>> located.
>
>> I have a private prototype patch that fixes two problems I've  
>> specified, and I wanted to clean it up for inclusion in 0.13.
>
> yegh, so you might need to use: "[A-Za-z][A-Za-z']*[A-Za-z]" as the  
> regex instead.  Ugly Note: that this probably excludes non-latin  
> charecters, which isn't desired.  QtRegex doesn't easily support  
> matching only letter by the looks of it.  I guess you might be able  
> to work some magic with zero length lookaheads, but then the regex  
> starts to get too magic.
>
> I guess you could just match on "\\w[\\w']*\\w", and then check that  
> "\\d" fails too, that may be the 'best' technique to follow.


This solution works for me:

	rehighlightTimer_->stop();
	needRehighlight_ = previousBlockState() != -1;
	int cursorPosition = cursorPositionInCurrentBlock();

	// Underline
	QTextCharFormat tcf;
	tcf.setUnderlineColor(QBrush(QColor(255,0,0)));
	tcf.setUnderlineStyle(QTextCharFormat::SpellCheckUnderline);

	// Match words (minimally)
	static QRegExp expression("\\b\\w+\\b");
	static QRegExp number("^\\d+$");

	// Iterate through all words
	int index = text.indexOf(expression);
	while (index >= 0) {
		int length = expression.matchedLength();
		if (expression.cap().indexOf(number) >= 0) {
			// skipping numbers
		}
		else if (cursorPosition >= index && (cursorPosition <= index +  
length)) {
			needRehighlight_ = true;
		}
		else {
			if (!SpellChecker::instance()->isCorrect(expression.cap()))
				setFormat(index, length, tcf);
		}
		index = text.indexOf(expression, index + length);
	}

	setCurrentBlockState(needRehighlight_ ? 0 : -1);

Not shown is the code that re-highlights the text when cursor moves  
around the text field ;-)

-Michail