The second preset combines text that contains a Unicode escape and escaped backslashes with an empty object and an array of three numeric spellings. Four-space pretty printing is selected. The case demonstrates that formatting must distinguish structural characters from characters inside a string token. It also shows why a formatter should retain `-0` and `1e+09` rather than normalize them through an arithmetic type.
Escapes remain part of the original string token
The title uses `\u2603` rather than a literal snowman character. Both can represent related Unicode content, but they are different source spellings. The path contains escaped backslashes that must not become structure or disappear. JSONClear checks that each backslash introduces one permitted escape and that every `\u` is followed by four hexadecimal digits. It then emits the exact raw token in both pretty and minified output.
Empty objects and arrays stay compact
An empty object has no member to place on another line, so pretty output keeps `{}` together. The same rule applies to `[]`. Nonempty arrays place each value on its own line with one additional indentation level. This rule avoids decorative blank lines and gives identical output for every valid input with the chosen indentation. Member order and array order are never changed.
Negative zero and exponent notation remain visible
The values `0`, `-0` and `1e+09` all pass strict number grammar, but their spelling carries different evidence. JSONClear retains each token and reports the array and value counts without performing arithmetic. Minifying the output removes indentation and spaces only. A destination may interpret negative zero like zero or impose a range on the exponent; that is a semantic or schema decision outside the formatter and must be verified separately.