JSON Tree Diff

Two JSON documents can be byte-different but semantically identical: {"a":1,"b":2} and {"b":2,"a":1} have the same data, different text. A line-diff says "every line changed"; a structural diff says "no change". Conversely, two JSON docs with identical layout can differ semantically only at one deeply-nested value — a line-diff highlights one line, a structural diff names the JSON Pointer path.

For arrays the question of "moved" vs "changed" matters more. Some structural diff algorithms (RFC 6902 JSON Patch) treat arrays positionally — element at index 0 is different from element at index 1, even if it is the same object. Others (json-diff with --keys-only or moved-element detection) try to identify moves. This tool defaults to positional comparison; for "moved element" detection, normalize the arrays first.

• JSON Pointer paths — RFC 6901 syntax. /user/roles/2 = "the third element of the roles array inside user". Slashes escape characters: ~0 = literal ~, ~1 = literal /. Use these paths to navigate to the changed nodes in your source data.
• Path-by-path output — each change is one line: path + verb + value. Easy to grep, easy to programmatically post-process.
• Tree view — for visual review, collapses unchanged subtrees and highlights only the changed branches. Faster than reading text output for deeply nested data.
• Filtering — toggle additions / removals / modifications independently. Useful when reviewing a release: "show me all the new fields" without the noise of changed values.

• You are debugging "the API response changed and tests are failing" — diff the recorded fixture against the current response.
• You are migrating between two services that produce "the same" API and want to identify every behavioral difference.
• You inherited a JSON config file with hundreds of nested keys and need to compare staging vs production to find environment-specific overrides.
• You are validating that a JSON Patch (RFC 6902) actually transforms input A into expected output B — apply the patch, diff against expected.

• It will not detect renames automatically. A removed key + an added key with the same value will appear as two separate changes; you can interpret the pair as a rename, but no heuristic is universally correct.
• It will not align reordered arrays. {"items":[A,B,C]} vs {"items":[C,A,B]} appears as three changed elements. For order-insensitive comparison, sort arrays first by a stable key.
• It will not produce a JSON Patch. Output is human-readable. For producing a patch document (RFC 6902 operations), use a library like fast-json-patch.compare().