JSON Diff

Yes. Every view, every filter and all three export formats are free, with no account, no sign-up and no limit on how many documents you compare.

No. Both documents are parsed and diffed by JavaScript running in your browser. Nothing is sent to a server, nothing is stored and nothing is logged. You can disconnect from the network after the page loads and the tool keeps working — which is the point, because the documents people diff are usually API responses and config files they should not be pasting into a stranger's server.

They share the same comparison engine and answer the same question in two different shapes. JSON Compare puts the documents in two aligned columns, which suits reading two versions side by side. JSON Diff shows one merged document in the unified format that Git, GitHub and code review tools use, with removed lines marked − and added lines marked +. Diff also adds filters by change type, a changes-only list with both values, and exports as a .diff file or a standalone HTML report. Use Compare to read two documents; use Diff to produce something you will paste into a pull request or a ticket.

The format Git prints. Instead of two columns, you get one document with a marker in the left margin of each line: − for a line that only exists in the old version, + for one that only exists in the new, and a blank for lines that are the same in both. Two gutters number the lines in each version. It is the format almost every code review tool uses, so it is instantly readable to anyone who reviews code.

Both documents are parsed into data structures and walked in parallel, key by key and index by index. A key present on only one side is an addition or a removal; a key on both sides with different values is a modification. Because it compares parsed data rather than text, formatting, indentation and key order never appear as changes — only the data does.

No. Two objects with the same keys and values are identical whatever order they were written in, because JSON objects are unordered by definition. Reformatting a file, sorting its keys or switching from tabs to spaces produces no changes at all.

By default, yes — arrays are ordered, so moving an item is a real change. If you are diffing a list where the order carries no meaning, turn on 'Ignore array order' in the options and items are matched by content instead, so a reordered list reads as unchanged. On arrays beyond about 20,000 items this falls back to positional matching and the summary says so, because matching by content requires serialising every element.

They narrow the diff to the kind of change you are looking for. Turn off Added, Removed or Modified to hide that category, and use the Arrays or Objects limit to show only changes to list data or only changes to object properties. The counter in the result header always says how many of the total changes are currently shown, so a filter can never make you think there is less to review than there is.

Three ways. The JSON report is structured data — every change with its path and both values — for feeding into another program. The plain text export is a standard unified .diff file for a ticket or a commit message. The HTML report is a single self-contained file with the colours intact and no external dependencies, so it can be attached to an issue or emailed and will still render on a machine that has never seen this site.

Documents up to about 20 MB of text each. Parsing happens in a background worker so the page stays responsive, and the views only render the rows currently on screen, so a diff with tens of thousands of changes scrolls smoothly. Very large documents that differ almost everywhere stop at an internal ceiling and the summary says the result was truncated rather than quietly showing you a partial answer.

It is the address of a value inside the document, written the way you would reach it in code: $.users[2].email means the email property of the third item of the users array. Every change in the diff has one, and the copy button next to it puts it on your clipboard, so you can paste it straight into a bug report, a test assertion or a jq expression.

Because a diff between a document and a guess is not useful. If either side fails to parse, the status bar names the line, the column and what the parser expected, and clicking it jumps the cursor there. Unlike the formatter, this tool never auto-repairs your input: repairing one side would put a change on screen that exists in neither of your files.

The value exists on both sides but its JSON type is different — the string "1" became the number 1, say, or an object became an array. These are called out separately because they are the changes most likely to break something downstream, and they are easy to miss when you are only reading values.

Yes. Use the upload button on either panel, or drag a .json file straight onto it. The file is read locally with the browser's FileReader; it is never transmitted. The panel then shows the file's name so the exported report says which files were compared.

Ctrl+Enter (Cmd+Enter on a Mac) runs the diff immediately, Ctrl+Shift+F opens the search, Ctrl+Shift+S swaps the two sides, Ctrl+Shift+L loads a new sample pair and Ctrl+Shift+Delete clears both panels. Escape closes the search. The previous and next buttons in the result header step through the changes one at a time.

Once the page has loaded, yes. Everything runs locally, so you can disconnect and keep diffing. A connection is only needed to load the page the first time.