Text diff

How to use it

1. 1

   Paste both versions

   Original on the left, changed on the right. The comparison appears as soon as either box has something in it and redraws on every keystroke.

2. 2

   Read the summary line

   It gives added, removed and rewritten line counts, how much of the text is the same, and whether any block simply moved.

3. 3

   Step through the changes

   Previous and Next jump from one change to the next and outline it, so you never have to hunt through a long file.

4. 4

   Switch the view to suit the screen

   Side by side on a wide screen, one column on a phone. One column also matches the format Git and code review tools use.

5. 5

   Copy the result as a patch

   The patch is in standard unified format, with three lines of context around each change, so it pastes straight into a ticket or a code review.

How the comparison is worked out

The shared beginning and end of the two versions are matched first, so only the part that actually moved is compared properly. That middle section goes through the Myers difference algorithm, which finds the shortest set of insertions and deletions that turns one list into the other. It runs first over lines, then again inside any pair of lines that are similar enough to be a rewrite rather than a replacement, this time over words or characters. When the middle section is so large that a precise answer would stall the page, the tool says so in the summary line and shows the section as one block replacement instead of guessing.

Colour is never the only signal

Added lines carry a plus, removed lines carry a minus, and both have a coloured bar down the left edge as well as a tinted background. Inside a rewritten line, inserted words are underlined and deleted words are struck through. That means the whole comparison still reads correctly in black and white, on a projector, or for the roughly one in twelve men who cannot reliably separate red from green. It is the same approach the accessibility guidelines ask for and the same one code review tools settled on.

Word level and character level are different tools

Word level is the right default for prose. It marks whole words as inserted or deleted, which is how a person reads a change to a sentence. Character level is better for identifiers, reference numbers and prices, where one digit has changed inside a token and word level would just tell you the whole token is different. Switch between them in the advanced options.

Nothing is uploaded, and you can check

The comparison runs in JavaScript in this page. After the page loads it makes no network request, which you can confirm in your browser's network tab: type into either box and watch nothing happen. That matters here more than on most tools, because the two things people most often compare are contract drafts and configuration files.