JSON Diff Online

Structure first, text second

A line-based text diff compares the characters you typed. A JSON diff compares what the characters mean. That difference matters the moment two files disagree only in whitespace, in key order, or in how a value was written. In the Compare view you get both: a text diff for the raw lines and a path diff that turns each document into a list of paths and values, then reports what was added, removed, changed or left the same. Compare is part of Pro; the other views in the app, and the Format and Validate tools, are free.

How paths are built

Every leaf value gets a path. Object keys are joined with dots, so a value inside user and then address and then city is written as user.address.city. Array positions go in square brackets, so the second item of orders is orders[1]. The path diff matches those strings between the two sides. A path on one side only is added or removed; a path on both sides with different values is changed; an identical value is listed as same, and you can hide those rows to keep the list short.

Key order and array order are not the same

Object members have no defined order in JSON, so { "a": 1, "b": 2 } and { "b": 2, "a": 1 } are the same data, and the path diff pairs them up by path anyway. The text diff, however, will mark them different, which is why it is worth putting the keys in the same order in both documents first if you intend to read the text view. The Ignore key order switch does this for you when you read the path diff.

Arrays are different, because position is part of the value. If a list gains one item at the front, every later path shifts by one, and the diff will report many changed rows even though only one element was inserted. When you see a wall of changes under one array, suspect a shift and look at the first row rather than the last.

Reading the result without being fooled

Decide which document is the baseline before you read a single row, and write it down. The view labels sides left and right, and the added and removed tags are relative to that. Then do a quick sanity check with a tiny pair of documents so you know which way the labels point in your session. Look out for type changes too: 5 and "5" print almost alike and are different values. A key set to null and a key that is missing are also different rows, even if your application treats them the same.

Remember that moved data looks like removal plus addition. The diff has no notion of a renamed key; it only sees one path vanish and another appear.

Making the diff small enough to read

Strip the noise before comparing: timestamps, request IDs and generated tokens change on every call. Compare the slice you care about, such as one record or one config section, rather than a whole export. You can filter rows by path text, copy a summary of the changes, or export them. If both sides must be valid before a path diff is possible, Validate each one first, since a path diff needs two parseable documents.