TextTransformation creates 30+ instances of TextWatcher

[Reinmar]

TextTransformation creates a new TextWatcher instance for every configured transformation. This is completely non-optimal. We should have one TextWatcher and check all transformations in it. And then react to matchetd:data and check whcih transformation we matched.

---

If you'd like to see this fixed sooner, add a 👍 reaction to this post.

[Reinmar]

cc @jodator

[Reinmar]

I'm thinking about something like this:

		const watcher = new TextWatcher( editor.model, text => {
			for ( const transformation of configuredTransformations ) {
				const from = normalizeFrom( transformation.from );
				const to = normalizeTo( transformation.to );
			
				const match = from.test( text );

				if ( match ) {
					// Return match meta data so we can access that in the `matched:data` event.
					return { match, transformation };
				}
			}
		} );

		watcher.on( 'matched:data', ( evt, data ) => {
			const transformation = data.transformation;

			// Apply the transformation...
		} );

Which means that the return value of TextWatcher callback must allow returning some metadata (in this case – the transformation object that we need later on in matched:data listener.

[jodator]

Yeah, I was worried about that and probably left for next step which supposed to be opening the API.

The problem with that might be different and might be deeper. Nah we can fix that as you proposed. I would only advise to not take this example as-is. Ie, the from/to might be stored somewhere to not create them every time user type. The for loop should be AFAP® (as-fast-as-possible) IMO.

The other things is that we could rethink this watcher API and how the test callbacks are registered but passing stuff in the data is OK.