Diff Checker

The algorithm behind the highlights

Both texts are split into lines, and the tool finds the longest sequence of lines that appears in the same order in both versions (the longest common subsequence). Those lines are shown unchanged; every original line outside that sequence is a removal and every changed line outside it is an addition. This is the same idea used by version-control diffs, applied to whole lines with no character-level comparison.

With the sample poems, the first two lines match and form the common sequence, so the output is two plain lines, '− Sugar is sweet' and '+ And so are you', with the totals +1 and −1. Inserting a line keeps the rest aligned: comparing 'a, b, c' with 'a, x, b, c' shows only '+ x', not three changed lines.

What counts as a different line

Two lines are equal only if they are identical character for character. Anything else is a removal followed by an addition.

Reading the counters

• A line you edited counts once as removed and once as added; the totals do not report 'modified' lines separately.
• Where the algorithm could list a removal or an addition first, it shows the removal first, so a changed block reads as old lines then new lines.
• The + and − symbols in the left column are not selectable, and long lines wrap inside the panel instead of scrolling sideways.
• Two empty boxes show one blank unchanged line, because an empty text is still one (empty) line.

When everything shows as changed

If almost every line is both red and green, the versions usually differ in something invisible rather than in content. The common causes are Windows versus Unix line endings after one copy passed through a different editor, tabs replaced by spaces by a code formatter, trailing spaces added by an editor, and straight quotes turned into curly quotes by a word processor. Normalise both texts with the same editor, or paste them into the Text Case Converter or a plain-text field first, then compare again.

Size and performance

The comparison builds a table with one cell for every pair of lines, so the work grows with the product of the two line counts. Two texts of 1,000 lines need a million cells and finish instantly; 5,000 lines each need 25 million and take a noticeable moment; tens of thousands of lines on each side can stall the page or run out of memory. For very large files, compare one section at a time.