Schema String Formats

As described in the Binary Strings example, Lua strings are raw byte sequences. Without a schema, luadata applies a UTF-8/Latin-1 heuristic to decide how to encode them. This works well for most cases, but some strings are genuinely binary — packed coordinates, serialized databases, compressed blobs — and a consumer may want a specific encoding.

JSON Schema's format keyword lets you declare how a string field's bytes should be encoded in JSON output:

{
  "type": "object",
  "properties": {
    "name": { "type": "string" },
    "serialized": { "type": "string", "format": "base64" },
    "coords": { "type": "string", "format": "bytes" }
  }
}

Three formats are supported:

• base64 — encode the raw bytes as a base64 string, ideal for binary blobs that need safe JSON transport.
• bytes — emit a JSON array of integer byte values [72, 101, ...], useful when consumers need to process individual bytes.
• latin1 — force the Latin-1 mapping (each byte becomes its code point), bypassing the UTF-8 heuristic.

Without a format, the default heuristic applies (UTF-8 if valid, Latin-1 otherwise).

name = "Thrall"
serialized = "packed-quest-data"
coords = "AB"

{
  "name": "Thrall",
  "serialized": "cGFja2VkLXF1ZXN0LWRhdGE=",
  "coords": [65, 66]
}

The name field has no format, so it encodes as a normal UTF-8 string. The serialized field uses base64, producing a compact encoded string. The coords field uses bytes, producing an array of integer byte values.

Try editing the schema above — change serialized to "format": "bytes" or remove the format entirely to see how the output changes.