perf(tokenizer): skip runs of text in one loop when decoding entities

src/Tokenizer.spec.ts

@@ -191,6 +191,27 @@ describe("Tokenizer", () => {
 
         it("for multi-byte entities", () =>
             expect(tokenize("≧̸")).toMatchSnapshot());
+
+        it("for text and entities split across chunks", () =>
+            expect(
+                tokenize((tokenizer) => {
+                    tokenizer.write("one & two & three");
+                    tokenizer.write(" four<b>five</b>six &lt");
+                    tokenizer.end();
+                }),
+            ).toEqual([
+                ["ontext", 0, 10],
+                ["ontextentity", 38, 15],
+                ["ontext", 15, 21],
+                ["ontext", 21, 26],
+                ["onopentagname", 27, 28],
+                ["onopentagend", 28],
+                ["ontext", 29, 33],
+                ["onclosetag", 35, 36],
+                ["ontext", 37, 41],
+                ["ontextentity", 60, 44],
+                ["onend"],
+            ]));
     });
 
     it("should close comments on --!>", () => {

src/Tokenizer.ts

@@ -247,8 +247,20 @@ export default class Tokenizer {
             }
             this.state = State.BeforeTagName;
             this.sectionStart = this.index;
-        } else if (this.decodeEntities && c === CharCodes.Amp) {
-            this.startEntity();
+        } else if (this.decodeEntities) {
+            if (c === CharCodes.Amp) {
+                this.startEntity();
+                return;
+            }
+            // Skip to the next `<` or `&`, or the end of the buffer.
+            const { buffer, offset } = this;
+            let index = this.index - offset + 1;
+            while (index < buffer.length) {
+                const next = buffer.charCodeAt(index);
+                if (next === CharCodes.Lt || next === CharCodes.Amp) break;
+                index++;
+            }
+            this.index = index + offset - 1;
         }
     }