Ignore diagnostic comparisons in legacy Unicode property dumps

Stop collecting property members at Discrepancy sections and resume at the next property heading. This removes diagnostic-only code points from Unicode 2.1.8 Math, Alphabetic, and Diacritic sets.

Assisted-By: GPT-6
Issue: #110

Author: JLHwung

scripts/parse-blocks-scripts-properties.mjs

@@ -348,9 +348,7 @@ const parseProperties = async (version) => {
 					// Unassigned code points are already available as General_Category=Cn.
 					currentProperty === 'Unassigned Code Value'
 				) {
-					// Skip contents.
-					i++;
-					while (++i < lines.length && lines[i].trim() !== '');
+					currentProperty = undefined;
 					continue;
 				}
 				if (currentProperty === 'Not a Character') {
@@ -359,7 +357,12 @@ const parseProperties = async (version) => {
 				currentProperty = currentProperty.replaceAll(' ', '_');
 				currentProperty = findCanonicalName(currentProperty) ?? currentProperty;
 				map[currentProperty] = regenerate();
+			} else if (line.startsWith('Discrepancy ')) {
+				// Unicode 2.1.8 includes comparisons with UnicodeData after each
+				// record list. Those ranges are diagnostics, not property members.
+				currentProperty = undefined;
 			} else if (
+				currentProperty &&
 				(maybeRange = /^([A-F\d]{4,6})(?:\.\.([A-F\d]{4,6}))?(?=\s|$)/.exec(
 					line,
 				)) != null

tests/parser-formats.mjs

@@ -11,10 +11,61 @@ import parseScriptExtensions from '../scripts/parse-script-extensions.mjs';
 import {
 	parseBlocks,
 	parseDerivedCoreProperties,
+	parseProperties,
 	parseScripts,
 } from '../scripts/parse-blocks-scripts-properties.mjs';
 import utils from '../scripts/utils.mjs';
 
+test('Legacy property diagnostics end a record section until the next heading', async (t) => {
+	t.mock.method(
+		utils,
+		'readDataFile',
+		() => `
+Property dump for: 0x20000001 (Alphabetic)
+
+  0041..005A  (26 chars)
+Discrepancy A: Unilib but not UnicodeData
+
+0041
+Discrepancy B: UnicodeData but not Unilib
+
+4E00..9FA5  (20902 chars)
+Property dump for: 0x20010000 (Diacritic)
+
+0300
+  Discrepancy B: UnicodeData but not Unilib
+0640
+Property dump for: 0x01000000 (Bidi: Left-to-Right)
+
+0041
+
+Discrepancy B: UnicodeData but not Unilib
+
+0042
+
+Property dump for: 0x00000000 (Unassigned Code Value)
+
+0378
+
+Property dump for: 0x10000004 (White space)
+
+0020
+`,
+	);
+	const properties = await parseProperties('2.1.8');
+	t.assert.deepEqual(Object.keys(properties), [
+		'Alphabetic',
+		'Diacritic',
+		'White_Space',
+	]);
+	t.assert.deepEqual(
+		properties.Alphabetic.toArray(),
+		Array.from({ length: 26 }, (_, i) => 0x41 + i),
+	);
+	t.assert.deepEqual(properties.Diacritic.toArray(), [0x300]);
+	t.assert.deepEqual(properties.White_Space.toArray(), [0x20]);
+});
+
 test('Simple case mappings ignore comments after empty and explicit titlecase fields', async (t) => {
 	t.mock.method(utils, 'readDataFile', () =>
 		[

tests/parsers.mjs

@@ -17,6 +17,22 @@ import {
 	parseScripts,
 } from '../scripts/parse-blocks-scripts-properties.mjs';
 
+test('Unicode 2.1.8 property values exclude discrepancy diagnostics', async (t) => {
+	const properties = await parseProperties('2.1.8');
+	t.assert.equal(properties.Math.toArray().length, 345);
+	t.assert.equal(properties.Alphabetic.toArray().length, 15181);
+	t.assert.equal(properties.Diacritic.toArray().length, 274);
+	t.assert.ok(properties.Math.contains(0xffe9));
+	t.assert.ok(!properties.Math.contains(0xffe8));
+	t.assert.ok(properties.Alphabetic.contains(0x41));
+	t.assert.ok(!properties.Alphabetic.contains(0x2d0));
+	t.assert.ok(!properties.Alphabetic.contains(0x4e00));
+	t.assert.ok(properties.Diacritic.contains(0x300));
+	t.assert.ok(!properties.Diacritic.contains(0x640));
+	// The next property's real records still include the ideograph.
+	t.assert.ok(properties.Ideographic.contains(0x4e00));
+});
+
 test('Unicode 5.0 through 6.0 provide their historical correction aliases', async (t) => {
 	for (const version of ['5.0.0', '5.1.0', '5.2.0', '6.0.0']) {
 		const aliases = await parseNameAliases(version);