JSON Sort Keys

Reorders the keys of a JSON object in alphabetic order, with optional recursive descent into nested objects and arrays of objects. Arrays themselves are left in their original order — only object key order changes. Useful for deterministic diffs (committing a sorted package.json so reviewers see only meaningful changes), normalizing API fixtures before snapshot testing, or matching the expected order in a downstream tool that's strict about key order.

Leaving arrays alone is the correct behaviour rather than a missing feature. An object is an unordered collection by definition, so reordering its keys cannot change what it means, while a JSON array is ordered and sorting it would change the document. That is also why a sorted file and its original parse to equal values: the diff moves, the data does not.

The sort is on the string form of the key, which is not the same as an intuitive order. Uppercase letters sort before lowercase ones, so Name lands above age; digits sort before letters; and a numeric key sorts as text, which puts 10 before 2. A file with mixed-case keys will look shuffled rather than alphabetised the first time, and that is the ordering every other tool doing the same thing will produce too.

JSON permits duplicate keys and parsers handle them by keeping the last one, so a document that carries the same key twice loses a value here as it would anywhere else. Very large integers are the other quiet loss: values beyond the safe integer range change when they pass through a parse and re-serialise, which is why a 19-digit identifier should be a string.

One warning about the package.json case, which is the most common reason to reach for this. Sorting keys is safe, but the fields inside that file are not all order-free in practice: the exports map is matched in declaration order by Node's resolver, so sorting it can change which entry point a package hands out. Sort the top level and leave that one alone.