Most JSON serializers give you only two choices:

• compact machine output:

{"a":{"b":{"c":"abc"}},"x":{"y":{"z":"xyz"}}}

• or fully expanded “pretty-print”:

{
  "a": {
    "b": {
      "c": "abc"
    }
  },
  "x": {
    "y": {
      "z": "xyz"
    }
  }
}

I wanted something in between: the first is hard for humans to scan, and the second becomes extremely verbose on real-world nested data.

The Idea

I wrote a small Python module called jsonfold. Instead of replacing Python’s JSON serializer, it works as a lightweight post-processing filter on top of json.dump() output.

The formatter selectively:

• folds small containers back onto one line,
• packs short scalar sequences,
• keeps large or complex structures expanded.

Example output:

{
  "a": { "b": { "c": "abc" } },
  "x": { "y": { "z": "xyz" } }
}

Why This Approach?

I did not want to rebuild a serializer - there are many good serializers (including the built-in json.dump()) that can efficiently process anything from simple data structures (list/dict) to custom classes and Python @dataclass objects, perform transformations and customize the output layout.

The interesting part is that the formatter does not re-parse the JSON stream. It operates as a streaming wrapper around file-like objects:

json.dump(obj, JSONFoldWriter(fp), indent=2)

That means that it can handle large documents with fixed memory usage and linear processing time. This approach works with most existing serializers. It also provides wrappers for json.dump(), json.dumps().

from jsonfold import dumps

data = {
    "a": {"b": {"c": "abc"}},
    "x": {"y": {"z": "xyz"}},
}

print(dumps(data))

Customization

The formatter allows controlling:

• maximum line width,
• folding depth,
• packing aggressiveness,
• array/object limits.

So you can choose between conservative formatting and more aggressive compaction.

Full Article:

Medium (no paywall): A Streaming JSON Formatter That Works With Existing Serializers

Minimal Usage

Pull jsonfold.py from GitHub project

import jsonfold
import sys
data = {
    "meta": {"version": 1, "ok": True},
    "ids": [1, 2, 3, 4, 5],
    "items": [{"id": 1, "name": "alpha"}, {"id": 2, "name": "beta"}],
}
# compact can be: default, low, med, high, max
jsonfold.dump(data, sys.stdout, compact="default")

GitHub Project

Repository: https://github.com/yairlenga/jsonfold

Python implementation is under python directory.

Future articles will cover other implementations: JavaScript, Java, C, ... - please watch the GitHub project, or follow the articles on Medium.