Paste an original on the left and a revision on the right. Removed lines come back red, added lines green, unchanged lines plain. Nothing is uploaded.
It splits both inputs into lines and computes the longest common subsequence between them. Lines present on both sides stay unmarked, lines only on the left are marked as removed, and lines only on the right are marked as added. The result is the same alignment a version control diff produces.
Line endings are normalised first, so a file saved on Windows and the same file saved on Linux compare as identical rather than as a hundred changed lines.
A red line exists only in the original. A green line exists only in the revision. Everything else matched. Blocks that appear as red-then-green are usually an edit: the old line replaced by a new one.
Because the comparison is by line, a change inside a long line shows the whole line as removed and re-added. For a character-level view you want a word diff, which this tool does not attempt.
Ignore whitespace collapses runs of spaces and tabs before comparing, so reindenting a block produces no diff. Ignore case folds both sides to lowercase, which catches a rename that only changed capitalisation as unchanged.
Turn both on when you want to know whether two files differ in substance. Turn both off when you are auditing exact bytes.
No. The comparison runs in the page.
Red is removed, green is added, plain is unchanged.
Only if you tick the box. Otherwise spaces and tabs count.
A few thousand lines. The algorithm is quadratic in line count, so bigger files get slower.