How to Measure JSON Size: Characters, Bytes and Gzip

In JavaScript, string.length counts UTF-16 code units, not characters and not bytes. ASCII text has the same length and byte count, and anything else does not:

Text      .length   UTF-8 bytes
hello     5         5
é         1         2
€         1         3
👍        2         4
日本語     3         9

The emoji has a length of 2 because it needs two UTF-16 code units. Even é is not fixed: written as one composed character it has length 1, but decomposed into e plus a combining accent, as some keyboards and macOS file names produce, it has length 2.

Languages also disagree on length. For the JSON text {"name":"José","note":"👍"}, JavaScript and Java report a length of 27, because both count UTF-16 code units, while Python reports 26, because it counts code points. All three agree that it is 30 bytes in UTF-8.

Python's json.dumps escapes every non-ASCII character by default, so its output is larger than JavaScript's for the same data:

Value     json.dumps (default)        bytes   ensure_ascii=False   bytes
é         "\u00e9"                     8       "é"                   4
👍        "\ud83d\udc4d"               14      "👍"                  6
日本語     "\u65e5\u672c\u8a9e"         20      "日本語"               11

For text-heavy data in non-Latin scripts, the default output can be close to twice the size. Both forms are valid JSON and parse to the same strings, but a size limit sees the difference.

• Limits on request bodies, files and stored records almost always count UTF-8 bytes of the uncompressed text. Compressing a request does not always help either: Express, for example, applies its body limit after decompressing.
• Transfer time and bandwidth depend on the compressed bytes. HTTP Content-Length describes the body as sent: a 7,991-byte JSON response sent with gzip carried Content-Length: 730.
• Text field limits, such as a maximum length for a name, count characters, and as shown above, languages disagree on what a character is. Check which unit the validating side uses.

const json = JSON.stringify(payload);
Buffer.byteLength(json);                  // Node
new TextEncoder().encode(json).length;    // browsers and Node
new Blob([json]).size;                    // browsers and Node

json_text = json.dumps(payload, ensure_ascii=False, separators=(",", ":"))
len(json_text.encode("utf-8"))

json.getBytes(StandardCharsets.UTF_8).length

Measure exactly what you send. If the client minifies, measure the minified text. If Python sends with the default ensure_ascii, measure that output, not a version with ensure_ascii=False. For the compressed size, compress it: gzipSync(json).length in Node or len(gzip.compress(data)) in Python.