JSON ڈِف

Structure is compared, not text

The [msm_tool slug="json-diff"] walks both documents as trees rather than comparing them line by line as text, which means reformatting a file, reordering its keys, or changing its indentation produces no reported differences at all — a text-based comparison would flag the entire file as changed for exactly the same edit. Every genuine difference is reported by the path where it occurs in the tree.

Array order matters, until told otherwise

By default, an element that moved position within an array counts as a difference, which is correct wherever the order itself carries meaning — a sequence of steps, a list of instructions. Turning that setting off compares arrays as sets instead, which is the right mode for a list of tags, permissions or identifiers where membership matters and order does not.

What loose type comparison actually changes

With loose comparison enabled, the string `"1"` and the number `1` are treated as equal, which is useful specifically when one side of the comparison has passed through a system that turns every value into a string on the way. Left at the default strict comparison, the same pair is reported as a genuine difference.

A type change is reported even when the printed value looks identical

A field that used to hold a number and now holds a string differs structurally even if both happen to print the same digits, and that distinction is reported rather than smoothed over — it is frequently the exact kind of fault worth catching before it reaches something downstream that expected a number and receives text instead.

Why the same field can show up at many paths

Each occurrence of a value in the tree gets its own path in the report, so a field that repeats at many points in the structure produces one reported difference per occurrence rather than a single summary. That is what makes it possible to find every instance rather than fixing one and assuming the rest followed.

What it deliberately will not do

Whitespace, key order and indentation are invisible to this comparison by design, since none of them change what the data actually says. A text-based diff tool is the right choice when formatting itself is the thing being checked, rather than the values underneath it.