Strip the byte-order mark at the start of the JSON string

Author: isaacs

index.js

@@ -53,11 +53,12 @@ class JSONParseError extends SyntaxError {
 }
 
 const parseJson = (txt, reviver, context) => {
+  const parseText = stripBOM(txt)
   context = context || 20
   try {
-    return JSON.parse(txt, reviver)
+    return JSON.parse(parseText, reviver)
   } catch (e) {
-    if (typeof txt !== 'string') {
+    if (typeof txt !== 'string' && !Buffer.isBuffer(txt)) {
       const isEmptyArray = Array.isArray(txt) && txt.length === 0
       throw Object.assign(new TypeError(
         `Cannot parse ${isEmptyArray ? 'an empty array' : String(txt)}`
@@ -67,9 +68,14 @@ const parseJson = (txt, reviver, context) => {
       })
     }
 
-    throw new JSONParseError(e, txt, context, parseJson)
+    throw new JSONParseError(e, parseText, context, parseJson)
   }
 }
 
+// Remove byte order marker. This catches EF BB BF (the UTF-8 BOM)
+// because the buffer-to-string conversion in `fs.readFileSync()`
+// translates it to FEFF, the UTF-16 BOM.
+const stripBOM = txt => String(txt).replace(/^\uFEFF/, '')
+
 module.exports = parseJson
 parseJson.JSONParseError = JSONParseError

test/index.js

@@ -11,10 +11,24 @@ t.test('parses JSON', t => {
       baz: [1, 2, 3, 'four']
     }
   })
-  t.deepEqual(JSON.parse(data), parseJson(data), 'does the same thing')
+  t.deepEqual(parseJson(data), JSON.parse(data), 'does the same thing')
   t.done()
 })
 
+t.test('parses JSON if it is a Buffer, removing BOM bytes', t => {
+  const str = JSON.stringify({
+    foo: 1,
+    bar: {
+      baz: [1, 2, 3, 'four']
+    }
+  })
+  const data = Buffer.from(str)
+  const bom = Buffer.concat([Buffer.from([0xEF, 0xBB, 0xBF]), data])
+  t.deepEqual(parseJson(data), JSON.parse(str))
+  t.deepEqual(parseJson(bom), JSON.parse(str), 'strips the byte order marker')
+  t.end()
+})
+
 t.test('throws SyntaxError for unexpected token', t => {
   const data = 'foo'
   t.throws(