Format specific JSON objects on one line

Question

Consider the following code:

>>> import json
>>> data = {
...     'x': [1, {'$special': 'a'}, 2],
...     'y': {'$special': 'b'},
...     'z': {'p': True, 'q': False}
... }
>>> print(json.dumps(data, indent=2))
{
  "y": {
    "$special": "b"
  },
  "z": {
    "q": false,
    "p": true
  },
  "x": [
    1,
    {
      "$special": "a"
    },
    2
  ]
}

I want to format JSON so that JSON objects that have only one property '$special'appear on the same line, as shown below.

{
  "y": {"$special": "b"},
  "z": {
    "q": false,
    "p": true
  },
  "x": [
    1,
    {"$special": "a"},
    2
  ]
}

I played with the implementation JSONEncoderand passed this in json.dumpsas an argument cls, but two methods JSONEncoderdo not have a problem:

• The method JSONEncoder defaultis called for each part data, but the return value is not a raw JSON string, so there seems to be no way to configure its formatting.

• The method JSONEncoder encodereturns the original JSON string, but it is called only once for datathe whole.

Is there any way to get JSONEncoderto do what I want?

Answer 1

, , hellip; .

import json
import re
data = {
    'x': [1, {'$special': 'a'}, 2],
    'y': {'$special': 'b'},
    'z': {'p': True, 'q': False}
}
text = json.dumps(data, indent=2)
pattern = re.compile(r"""
{
\s*
"\$special"
\s*
:
\s*
"
((?:[^"]|\\"))*  # Captures zero or more NotQuote or EscapedQuote
"
\s*
}
""", re.VERBOSE)
print(pattern.sub(r'{"$special": "\1"}', text))

.

{
  "x": [
    1,
    {"$special": "a"},
    2
  ],
  "y": {"$special": "b"},
  "z": {
    "q": false,
    "p": true
  }
}

Answer 2

json ; .

, json, , tokenize module:

import tokenize
from io import BytesIO


def inline_special(json_data):
    def adjust(t, ld,):
        """Adjust token line number by offset"""
        (sl, sc), (el, ec) = t.start, t.end
        return t._replace(start=(sl + ld, sc), end=(el + ld, ec))

    def transform():
        with BytesIO(json_data.encode('utf8')) as b:
            held = []  # to defer newline tokens
            lastend = None  # to track the end pos of the prev token
            loffset = 0     # line offset to adjust tokens by
            tokens = tokenize.tokenize(b.readline)
            for tok in tokens:
                if tok.type == tokenize.NL:
                    # hold newlines until we know there no special key coming
                    held.append(adjust(tok, loffset))
                elif (tok.type == tokenize.STRING and
                        tok.string == '"$special"'):
                    # special string, collate tokens until the next rbrace
                    # held newlines are discarded, adjust the line offset
                    loffset -= len(held)
                    held = []
                    text = [tok.string]
                    while tok.exact_type != tokenize.RBRACE:
                        tok = next(tokens)
                        if tok.type != tokenize.NL:
                            text.append(tok.string)
                            if tok.string in ':,':
                                text.append(' ')
                        else:
                            loffset -= 1  # following lines all shift
                    line, col = lastend
                    text = ''.join(text)
                    endcol = col + len(text)
                    yield tokenize.TokenInfo(
                        tokenize.STRING, text, (line, col), (line, endcol),
                        '')
                    # adjust any remaining tokens on this line
                    while tok.type != tokenize.NL:
                        tok = next(tokens)
                        yield tok._replace(
                            start=(line, endcol),
                            end=(line, endcol + len(tok.string)))
                        endcol += len(tok.string)
                else:
                    # uninteresting token, yield any held newlines
                    if held:
                        yield from held
                        held = []
                    # adjust and remember last position
                    tok = adjust(tok, loffset)
                    lastend = tok.end
                    yield tok

    return tokenize.untokenize(transform()).decode('utf8')

:

import json

data = {
    'x': [1, {'$special': 'a'}, 2],
    'y': {'$special': 'b'},
    'z': {'p': True, 'q': False}
}

>>> print(inline_special(json.dumps(data, indent=2)))
{
  "x": [
    1,
    {"$special": "a"},
    2
  ],
  "y": {"$special": "b"},
  "z": {
    "p": true,
    "q": false
  }
}

Answer 3

, / json.encoder, .

Basically, copy completely _make_iterencodefrom json.encoderand make changes so your special dictionary prints without indentation of a new line. Then monkeypatch json package to use your modified version, run json dump, and then cancel monkeypatch removal (if you want).

The function is _make_iterencodequite long, so I just placed the parts that need to be changed.

import json
import json.encoder

def _make_iterencode(markers, _default, _encoder, _indent, _floatstr,
    ...
    def _iterencode_dict(dct, _current_indent_level):
        ...
        if _indent is not None:
            _current_indent_level += 1
            if '$special' in dct:
                newline_indent = ''
                item_separator = _item_separator
            else:
                newline_indent = '\n' + (' ' * (_indent * _current_indent_level))
                item_separator = _item_separator + newline_indent
            yield newline_indent
        ...
        if newline_indent is not None:
            _current_indent_level -= 1
            if '$special' not in dct:
                yield '\n' + (' ' * (_indent * _current_indent_level))

def main():
    data = {
        'x': [1, {'$special': 'a'}, 2],
        'y': {'$special': 'b'},
        'z': {'p': True, 'q': False},
    }

    orig_make_iterencoder = json.encoder._make_iterencode
    json.encoder._make_iterencode = _make_iterencode
    print(json.dumps(data, indent=2))
    json.encoder._make_iterencode = orig_make_iterencoder