URL: Fix URLSearchParams decoding to not throw URIError

polyfills/URL/polyfill.js

@@ -60,7 +60,29 @@
 			return output.replace(/%20/g, '+');
 		}
 
+		// NOTE: URL API accepts inputs like `?x=%`, `?x=%a`, and `?x=%2sf`
+		// as literals, whereas legacy decodeURIComponent would throw
+		// URIError (as specified by ECMAScript).
+		//
+		// https://url.spec.whatwg.org/#percent-decode
+		function percent_decode(bytes) {
+			// NOTE:
+			// * Only decode pairs of exactly two bytes.
+			// * Only decode bytes in range 0-9, A-F, a-f.
+			// * Decode as many pairs at the same time as possible.
+			//   This is because we're not actually operating on internal bytes,
+			//   but on a valid UTF string, and the string must remain valid at
+			//   all times, and decodeURIComponent will throw when attempting to
+			//   decode a byte that represents only part of a codepoint, for example
+			//   "%7F" separately from "%7F%C3%BF".
+			return bytes.replace(/((%[0-9A-Fa-f]{2})*)/g, function (_, m) {
+				return decodeURIComponent(m);
+			});
+		}
+
 		// NOTE: Doesn't do the encoding/decoding dance
+		//
+		// https://url.spec.whatwg.org/#concept-urlencoded-parser
 		function urlencoded_parse(input, isindex) {
 			var sequences = input.split('&');
 			if (isindex && sequences[0].indexOf('=') === -1)
@@ -83,8 +105,8 @@
 			var output = [];
 			pairs.forEach(function (pair) {
 				output.push({
-					name: decodeURIComponent(pair.name),
-					value: decodeURIComponent(pair.value)
+					name: percent_decode(pair.name),
+					value: percent_decode(pair.value)
 				});
 			});
 			return output;

polyfills/URL/tests.js

@@ -594,6 +594,7 @@ describe('WPT tests', function () {
 		proclaim["throws"](function() { new URLSearchParams([[1,2,3]]); });
 	});
 
+	// Examples from wpt/url/urlencoded-parser.any.js
 	[
 		{
 			input: "z=b&a=b&z=a&a=a",
@@ -603,6 +604,26 @@ describe('WPT tests', function () {
 			input: "\uFFFD=x&\uFFFC&\uFFFD=a",
 			output: [["\uFFFC", ""], ["\uFFFD", "x"], ["\uFFFD", "a"]]
 		},
+		{
+			input: '%a=a',
+			output: [['%a', 'a']]
+		},
+		{
+			input: "id=0&value=%",
+			output: [['id', '0'], ['value', '%']]
+		},
+		{
+			input: "b=%2sf%2a",
+			output: [['b', '%2sf*']]
+		},
+		{
+			input: "b=%2%2af%2a",
+			output: [['b', '%2*f*']]
+		},
+		{
+			input: "b=%%2a",
+			output: [['b', '%*']]
+		},
 		{
 			input: "ffi&🌈", // 🌈 > code point, but < code unit because two code units
 			output: [["🌈", ""], ["ffi", ""]]