Enhancement? Option to clamp to words, sentences or other tokens

[JorenSix]

Hi Simon,

First of all thanks for developing this in the open! I am evaluating the text-annotator-js component to see if it fits for a project I am working on. It looks very promising.

One of the requirements for my project is that annotations should clamp to word boundaries. I have included the code below in the trimRange() method also mentioned in #66. Conceptually perhaps not the worst place to place such code. In relation to this I have a few questions:

• Is there already a way to clamp to words/sentences I perhaps overlooked?
• Is clamping functionality common enough to include in the annotator?
• Is there a better place to modify a range based on application rules?
• Would it be a good idea to add an optional callback to the annotator component which would then be used to clamp a range according to application rules?

Thanks in advance for considering these questions.

const word_boundaries = ['.',',',';',':','?','!',' ','\n','-'];
 
var charBeforeStart = startContainer.nodeValue.charAt(range.startOffset-1);
var newStartOffset = range.startOffset;
while(!word_boundaries.includes(charBeforeStart) && newStartOffset >= 0){
  newStartOffset = newStartOffset-1;
  charBeforeStart = startContainer.nodeValue.charAt(newStartOffset);
}
range.setStart(startContainer, newStartOffset);

var charAfterEnd = endContainer.nodeValue.charAt(range.endOffset);
var newEndOffset = range.endOffset;
while(!word_boundaries.includes(charAfterEnd) && newEndOffset < endContainer.nodeValue.length){
  newEndOffset = newEndOffset+1;
  charAfterEnd = endContainer.nodeValue.charAt(newEndOffset);
}
range.setEnd(endContainer, newEndOffset);

[rsimon]

Hi,

no clamping to words is not - properly - implemented anywhere right now. (The .trimRange() method was supposed to do it, but was buggy, as @oleksandr-danylchenko writes above).

Nonetheless, I'd personally welcome such a feature! It should IMO be made a config option, and possibly off by default. But I think it would be useful.

But just to clarify: if I read your code correctly, your goal is to clamp outwards, right? I.e. I have the following:

Lorem ipsum dolor sit amet

Your use case is that I'd select orem ipsu, and the selection would clamp to lorem ipsum, right? On the other hand, if I select ipsum (with a leading white space), the selection would not clamp to ipsum? (Which is what the original trimeRange() method did.)

I think both would be useful. But, as I said, definitely as a config option.

[JorenSix]

Thanks for the quick response!

The implementation is indeed outward clamping and the does indeed behave buggy if the selection starts with a space. Thanks for the notification. Perhaps there are other edge cases not covered as well e.g. with the non annotatable elements.

I do think the clamping (inward, outward, word boundary definitions) may be culture/application dependent. E.g. to count a - as a word boundry or not. Or to perhaps clamp annotations on sentence level. So I would suggest to configure it as a callback where the trimRange either does nothing (default?), is configured to use a best guess word level clamping, or a implemented in a user defined function. I think that would be the most flexible.

Thanks again! J