JSON Validator

JSON is defined by RFC 8259 (also published as ECMA-404). The grammar is small, and nearly every validation failure comes from one of these rules:

• Strings use double quotes only. 'name' is not a string, and object keys must be strings, so {name: 1} is invalid too.
• No trailing commas after the last array element or object member.
• No comments of any kind: no //, /* */ or #.
• Numbers are decimal. No leading zeros (007), no leading +, no hex, no bare .5 or 5., and no NaN or Infinity.
• Literals are lowercase: true, false, null. True, None and undefined are errors.
• Control characters must be escaped. A raw tab or newline inside a string is invalid. Write \t and \n instead.
• Any value can be the top level. "hello", 42 and null are complete JSON documents. The older RFC 4627 required an object or array at the top level, and a few old parsers still enforce that.

Whitespace between tokens is free: spaces, tabs, line feeds and carriage returns can appear anywhere outside a string, so indentation never affects validity.

These are the messages current Chrome, Edge and Node.js (V8) print. The position is a zero-based character offset into the string. Newer V8 versions also print a one-based line and column after it, which is why position 7 shows as column 8.

Older Chrome versions phrased the first error as Unexpected token } in JSON at position 42, which still appears in many Stack Overflow answers. Firefox reports the same problem as JSON.parse: expected double-quoted property name at line 3 column 5 of the JSON data. The wording differs between engines, but the fixes are the same. For a string that fails because of its escapes, the JSON escape tool reports the exact bad sequence.

A lot of what people call JSON is really a superset of it. VS Code's settings.json and tsconfig.json are JSONC, which allows comments and trailing commas. JSON5 also allows single quotes, unquoted keys and hex numbers. These files work fine in the tools built for them and fail in a strict parser:

{
  // build output
  "outDir": "dist",
  "strict": true,
}

To turn such a file into real JSON, delete the comments and the comma after true. The Try to fix button handles the common cases: single quotes, unquoted keys, trailing commas and comments. Check its output before you rely on it. Comment removal is text-based, so it also cuts off a URL like "https://..." at the //.

NaN usually comes from a producer and not from hand editing. Python's json.dumps writes NaN and Infinity by default, so a pandas column containing a missing float produces output that browsers reject. Pass allow_nan=False to fail early, or convert the values to None first. JavaScript does the opposite: JSON.stringify(NaN) silently writes null.

RFC 8259 says JSON exchanged between systems must be UTF-8, and that implementations must not add a byte order mark (BOM). Some tools still write one. Windows PowerShell 5.1 with Out-File -Encoding utf8 is a common source. JSON.parse then fails on the very first character with Unexpected token '', where the token looks empty because U+FEFF is invisible. Check for it with head -c 3 file.json | xxd. If the output starts with efbb bf, the file has a BOM. Strip it with text.replace(/^\uFEFF/, '') in JavaScript, or open the file with encoding='utf-8-sig' in Python. This page trims surrounding whitespace, including a leading BOM, before parsing. A file can therefore pass here and still fail in a strict parser, so run the xxd check when the error is at position 0.

Duplicate keys are a quieter problem. The RFC only says names should be unique, and it leaves the behavior up to each parser. JSON.parse('{"a":1,"a":2}') returns {a: 2} without any warning, and Python behaves the same way. Because this validator uses JSON.parse, it accepts duplicates as well. To catch them in Python:

def no_dupes(pairs):
    keys = [k for k, _ in pairs]
    dupes = {k for k in keys if keys.count(k) > 1}
    if dupes:
        raise ValueError(f"duplicate keys: {dupes}")
    return dict(pairs)

json.loads(text, object_pairs_hook=no_dupes)

Once the document is valid, use the JSON formatter to indent it consistently. To find what changed between two versions of a config, compare them in the diff checker.