Text diff checker: compare two texts online.

A diff checker compares two versions of a text and shows the smallest set of changes that turns the first into the second: which lines were removed, which were added and which stayed the same. The name comes from the Unix diff command, and the same idea powers version control, code review and the track changes view in word processors.

Finding that smallest set is equivalent to finding the longest common subsequence (LCS) of the two texts: the longest run of lines that appear in both, in the same order. Everything outside it is either a deletion from the original or an insertion in the new version.

This checker uses the algorithm Eugene Myers published in 1986, 'An O(ND) Difference Algorithm and Its Variations'. Picture a grid with the original's lines along one side and the new version's along the other. Moving right is a deletion, moving down is an insertion, and moving diagonally is free wherever two lines match. The algorithm searches outward one edit at a time and stops at the first path that reaches the far corner, which is guaranteed to use the fewest edits.

Its running time grows with N, the length of the texts, multiplied by D, the number of differences. That makes it very fast for the common case of two versions that are mostly the same, and it is why Myers is the default algorithm in git diff and the basis of GNU diff. Git also offers patience and histogram diffs, which use extra heuristics to align unique lines and often read better for moved code blocks; they do not always find the smallest edit, which Myers does.

The comparison runs first on whole lines. Where lines are removed and added at the same spot, the checker pairs them up as changed lines and then runs a second, word-level diff inside each pair, splitting text into words, spaces and punctuation. Only the words that actually changed are highlighted, so a one-word edit in a long paragraph is easy to spot.

To review a diff well, start with the counters to judge the size of the change, scan the overall shape of the edits, then read the word highlights line by line. In contracts and policies, look hardest at small words that flip meaning, such as may and must or a missing not, and at numbers, dates and names, which a quick read easily glosses over.

The side-by-side view puts the original and the new version in two columns with line numbers, which suits reviews on a large screen. The unified view shows one column, with removed lines marked - and added lines marked +, the same convention as git diff; it is the default on phones, where two columns do not fit. The counters above the result show how many lines were added, removed and left unchanged.

Ignore whitespace trims each line and collapses runs of spaces and tabs before comparing, so re-indented code, trailing spaces and double spaces after full stops no longer count as changes. Ignore case does the same for capitalisation. The original text is still displayed exactly as written; only the comparison changes.

Use them when you care about content rather than formatting, for example comparing a document that was reflowed, or configuration files that were re-indented. Leave them off when whitespace is meaningful, as in YAML, Python and Makefiles, where an indentation change can alter behaviour. Note that blank lines are still lines, so adding or removing one shows as a change either way.

Common jobs include reviewing what changed between two drafts of a contract, policy or proposal; comparing configuration or environment files between staging and production; checking an editor's changes to an article; comparing translations or subtitle files; and spotting the one value that differs between two long API responses or log extracts.

For JSON, format both sides first with the JSON formatter using the same indentation. Minified JSON is a single line, so any change marks the whole thing as different; pretty-printed JSON puts each value on its own line and the diff points straight at it. For Markdown documents, compare the source here and preview the result in the Markdown viewer.

An online diff tool is ideal for one-off comparisons of text that is not in a repository. For code and documentation that changes over time, version control such as Git records every change automatically, with author and reason, and shows diffs in every pull request as part of code review. git diff --no-index compares any two files on disk, and git diff --word-diff gives a word-level view similar to the highlights here.

Both texts are compared in your browser and never uploaded, so the checker is safe for contracts, credentials in config files and unreleased work. Texts of thousands of lines compare quickly. Because the work grows with the number of differences, two large and almost completely different texts can hit a built-in limit of 4,000 edits; at that point they are not really versions of each other and a diff is not useful.

A paragraph that was moved shows as a deletion in one place and an insertion in another, since a line diff does not track moves. Copy unified diff gives you the result as plain text with -, + and two-space prefixes, handy for sharing in chat or a ticket; it is not a patch file with @@ hunk headers, so use git diff when you need one to apply.