A JSON Encoder in Python, that puts small lists on single lines.

CompactJSONEncoder.py

#!/usr/bin/env python3
from __future__ import annotations

import json


class CompactJSONEncoder(json.JSONEncoder):
    """A JSON Encoder that puts small containers on single lines."""

    CONTAINER_TYPES = (list, tuple, dict)
    """Container datatypes include primitives or other containers."""

    MAX_WIDTH = 70
    """Maximum width of a container that might be put on a single line."""

    MAX_ITEMS = 10
    """Maximum number of items in container that might be put on single line."""

    def __init__(self, *args, **kwargs):
        # using this class without indentation is pointless
        if kwargs.get("indent") is None:
            kwargs["indent"] = 4
        super().__init__(*args, **kwargs)
        self.indentation_level = 0

    def encode(self, o):
        """Encode JSON object *o* with respect to single line lists."""
        if isinstance(o, (list, tuple)):
            return self._encode_list(o)
        if isinstance(o, dict):
            return self._encode_object(o)
        return json.dumps(
            o,
            skipkeys=self.skipkeys,
            ensure_ascii=self.ensure_ascii,
            check_circular=self.check_circular,
            allow_nan=self.allow_nan,
            sort_keys=self.sort_keys,
            indent=self.indent,
            separators=(self.item_separator, self.key_separator),
            default=self.default if hasattr(self, "default") else None,
        )

    def _encode_list(self, o):
        if self._put_on_single_line(o):
            return "[" + ", ".join(self.encode(el) for el in o) + "]"
        self.indentation_level += 1
        output = [self.indent_str + self.encode(el) for el in o]
        self.indentation_level -= 1
        return "[\n" + ",\n".join(output) + "\n" + self.indent_str + "]"

    def _encode_object(self, o):
        if not o:
            return "{}"

        # ensure keys are converted to strings
        o = {str(k) if k is not None else "null": v for k, v in o.items()}

        if self.sort_keys:
            o = dict(sorted(o.items(), key=lambda x: x[0]))

        if self._put_on_single_line(o):
            return (
                "{ "
                + ", ".join(
                    f"{self.encode(k)}: {self.encode(el)}" for k, el in o.items()
                )
                + " }"
            )

        self.indentation_level += 1
        output = [
            f"{self.indent_str}{self.encode(k)}: {self.encode(v)}" for k, v in o.items()
        ]
        self.indentation_level -= 1

        return "{\n" + ",\n".join(output) + "\n" + self.indent_str + "}"

    def iterencode(self, o, **kwargs):
        """Required to also work with `json.dump`."""
        return self.encode(o)

    def _put_on_single_line(self, o):
        return (
            self._primitives_only(o)
            and len(o) <= self.MAX_ITEMS
            and len(str(o)) - 2 <= self.MAX_WIDTH
        )

    def _primitives_only(self, o: list | tuple | dict):
        if isinstance(o, (list, tuple)):
            return not any(isinstance(el, self.CONTAINER_TYPES) for el in o)
        elif isinstance(o, dict):
            return not any(isinstance(el, self.CONTAINER_TYPES) for el in o.values())

    @property
    def indent_str(self) -> str:
        if isinstance(self.indent, int):
            return " " * (self.indentation_level * self.indent)
        elif isinstance(self.indent, str):
            return self.indentation_level * self.indent
        else:
            raise ValueError(
                f"indent must either be of type int or str (is: {type(self.indent)})"
            )


if __name__ == "__main__":
    import sys

    if "--example" in sys.argv:
        data = {
            "compact_object": {"first": "element", "second": 2},
            "compact_list": ["first", "second"],
            "long_list": [
                "this",
                "is",
                "a",
                "rather",
                "long\nlist",
                "and should be broken up because of its width",
            ],
            "non_ascii_汉": "汉语",
            1: 2,
        }
        json.dump(data, sys.stdout, cls=CompactJSONEncoder, ensure_ascii=False)
        exit()

    json.dump(json.load(sys.stdin), sys.stdout, cls=CompactJSONEncoder)

Little suggestion to also correctly process Numpy types:

    def encode(self, o):
        """Encode JSON object *o* with respect to single line lists."""
        if isinstance(o, (list, tuple)):
            return self._encode_list(o)
        elif isinstance(o, dict):
            return self._encode_object(o)
        if isinstance(o, float):  # Use scientific notation for floats
            return format(o, "g") 
        elif isinstance(o, np.integer):  # process numpy integers
            return self.encode(int(o))
        elif isinstance(o, np.floating): # process numpy floats
            return self.encode(float(o))
        elif isinstance(o, np.ndarray): # flatten numpy arrays as lists
            return self._encode_list(o.tolist())

        return json.dumps(
            o,
            skipkeys=self.skipkeys,
            ensure_ascii=self.ensure_ascii,
            check_circular=self.check_circular,
            allow_nan=self.allow_nan,
            sort_keys=self.sort_keys,
            indent=self.indent,
            separators=(self.item_separator, self.key_separator),
            default=self.default if hasattr(self, "default") else None,
        )

Also, personally, I would remove the format(o, "g") part, as it sometimes results in converting floats into exponential notation.

Awesome. I was going to implement something like this myself, but thought I'd check for an existing implementation first. This saved me so much time, and quite thorough job. Well done. 👏

@jannismain This saved me so much time, thanks a lot for the excellent work! 👏 And all you other guys for optimizing it ^^

Another minor bugfix:

f"{self.encode(k)}: {self.encode(el)}" for k, el in o.items() otherwise the key encoding is not correct (I had a non-ascii character in there)

Most json.dumps callas are missing at least ensure_ascii=self.ensure_ascii

This JSON data won't be formatted consistently:

{
  "data1": [
    27.135222728033753,
    114.02096846633076,
    26.074562556253937,
    24.57196064623251
  ],
  "data2": [
    27.135222,
    114.02094,
    26.074565,
    24.571964
  ]
}

@senyai
hi, to keep the number of digits after the decimal comma, remove the lines 32-33:

        if isinstance(o, float):  # Use scientific notation for floats
            return format(o, "g")

and to have both lists be printed on just one line, increase the value of MAX_WIDTH.

In this case, you will have the output:

{
  "data1": [27.135222728033753, 114.02096846633076, 26.074562556253937, 24.57196064623251],
  "data2": [27.135222, 114.02094, 26.074565, 24.571964]
}

Thanks @senyai @nikita-k0v and @chrismaes87, I have incorporated your suggestions.