zgo.at/jfmt

jfmt is my JSON formatter. There's a few things this does different than most JSON formatters I could find, which in my opinion makes the output much nicer.

Install with go install zgo.at/jfmt/cmd/jfmt@latest.

Or import zgo.at/jfmt to use in Go code.

---

It tries to be a little smarter about what should be on a single line. Things like this:

{
    "nums": [
        12,
        4,
        901,
        12,
        742,
    ],
    "strings": [
        "k",
        "p",
        "bas",
        "pqz",
        "q21",
    ],
    "obj" {
        "k": "v"
    }
}

Are just silly, and splitting it out over many lines like that isn't helpful. So that becomes:

{
    "nums":    [12, 4, 901, 12, 742],
    "strings": ["k", "p", "bas", "pqz", "q21"],
    "obj"      {"k": "v"}
}

It will still write to multiple lines if it's too long.

It will also align the keys; which I rather like.

Like many JSON formatting tools keys are sorted alphabetically, but it puts values that fit on a single line first, to avoid avoid things like this:

{
  "additional": false,
  "definitions": {
      [.. very long object ..]
   }
  "title": "JSON schema for Python project metadata and configuration",
  "type": "object",
}

I don't like these kind of "dangling primitives", so that that becomes:

{
  "additional": false,
  "title":      "JSON schema for Python project metadata and configuration",
  "type":       "object",
  "definitions": {
      [.. very long object ..]
   }
}

And things like this:

{
    "arr5": [
        [
            [
                [
                    [
                        {"type": "string", "value": "#"}
                    ]
                ]
            ]
        ]
    ]
}

Can just be written on a single line, too:

{
    "arr5": [[[[[{"type": "string", "value": "#"}]]]]],
}

There's a bunch of cases like this; in general it tries to be as concise as possible as long as it doesn't sacrifice readability.

---

The downside of this is that it's a bit slower and uses more memory. It's still plenty fast enough for almost all use cases; the largest file in schemastore.org (kubernetes-definitions.json, 17,500 lines/950K) takes about 120ms on my laptop (vs. 30ms with Go's default MarshalIndent). I spent almost zero effort optimising anything, so there's probably a lot to win here.