refactor(util): reuse BOM-stripped text (#46078)

packages/util/src/bom.test.ts

@@ -0,0 +1,21 @@
+import { expect, test } from "bun:test"
+import { Bom } from "./bom.js"
+
+test.each([
+  { prefix: "", bom: false, expected: undefined },
+  { prefix: "", bom: true, expected: "\uFEFF" },
+  { prefix: "\uFEFF", bom: false, expected: "" },
+  { prefix: "\uFEFF", bom: true, expected: undefined },
+  { prefix: "\uFEFF\uFEFF", bom: false, expected: "" },
+  { prefix: "\uFEFF\uFEFF", bom: true, expected: "\uFEFF" },
+])("syncBytes(%j)", (row) => {
+  const encoder = new TextEncoder()
+  const text = "a\uFEFF\u00e9"
+  const input = encoder.encode(row.prefix + text)
+
+  expect(Bom.syncBytes(input, row.bom)).toEqual({
+    text,
+    bytes: row.expected === undefined ? undefined : encoder.encode(row.expected + text),
+  })
+  expect(input).toEqual(encoder.encode(row.prefix + text))
+})

packages/util/src/bom.ts

@@ -27,7 +27,7 @@ export function decodeBytes(content: Uint8Array) {
 export function syncBytes(content: Uint8Array, bom: boolean) {
   const decoded = decode(content)
   const current = split(decoded)
-  const canonical = join(current.text, bom)
+  const canonical = bom ? value + current.text : current.text
   return { text: current.text, bytes: decoded === canonical ? undefined : new TextEncoder().encode(canonical) }
 }