Don't preserve whitespace from HTML in annotation text

This is an imperfect approach because it might miss empty block elements that
nevertheless start new paragraphs, but it works well for most webpages and
typical EPUBs. We can fine-tune more later.

https://forums.zotero.org/discussion/107685/z7-beta-zotero-introduces-extra-line-breaks-in-snapshot-annotations

src/dom/epub/epub-view.ts

@@ -16,7 +16,8 @@ import Epub, {
 } from "epubjs";
 import {
 	moveRangeEndsIntoTextNodes,
-	PersistentRange
+	PersistentRange,
+	splitRangeToTextNodes
 } from "../common/lib/range";
 import {
 	FragmentSelector,
@@ -350,10 +351,32 @@ class EPUBView extends DOMView<EPUBViewState, EPUBViewData> {
 		if (range.collapsed) {
 			return null;
 		}
-		let text = type == 'highlight' || type == 'underline' ? range.toString().trim() : undefined;
-		// If this annotation type wants text, but we didn't get any, abort
-		if (text === '') {
-			return null;
+		let text;
+		if (type == 'highlight' || type == 'underline') {
+			text = '';
+			let lastSplitRange;
+			for (let splitRange of splitRangeToTextNodes(range)) {
+				if (lastSplitRange) {
+					let lastSplitRangeContainer = closestElement(lastSplitRange.commonAncestorContainer);
+					let lastSplitRangeBlock = lastSplitRangeContainer && getContainingBlock(lastSplitRangeContainer);
+					let splitRangeContainer = closestElement(splitRange.commonAncestorContainer);
+					let splitRangeBlock = splitRangeContainer && getContainingBlock(splitRangeContainer);
+					if (lastSplitRangeBlock !== splitRangeBlock) {
+						text += '\n\n';
+					}
+				}
+				text += splitRange.toString().replace(/\s+/g, ' ');
+				lastSplitRange = splitRange;
+			}
+			text = text.trim();
+
+			// If this annotation type wants text, but we didn't get any, abort
+			if (!text) {
+				return null;
+			}
+		}
+		else {
+			text = undefined;
 		}
 
 		let selector = this.toSelector(range);