Vim's dot command: repeat any change

A change is a repeatable unit, and the dot key replays the last one anywhere, so shape the change to be repeatable. Vim keeps the keys of the last change: an operator with its motion (dw), or an insert from the key that started it to <Esc>, typed text included. . feeds those keys again at the cursor.

Transcript

One change, appending a semicolon with A, is recorded on the first line of a buffer, then the cursor moves down with j and the dot key replays the same change on the next two lines.

printf 'let a = 1\nlet bb = 22\nlet c = 3\n' > t.txt
vim -Nu NONE -n -es \
  -c 'exe "normal ggA;\<Esc>j.j."' -c wq t.txt
cat t.txt
# let a = 1;
# let bb = 22;
# let c = 3;

A;<Esc> is the change. j is a motion, not a change, so it is never recorded and . keeps meaning "append a semicolon".

Shape it to travel#

The change must carry its own position. $a;<Esc> puts the semicolon in the right place once, but $ is a separate motion, so . replays only a; wherever j left the cursor. Same file, line 2 is longer:

keys: gg$a;<Esc>j.j.
let a = 1;
let bb = 2;2
let c = 3;

A is $a fused into one command, so it travels. The same logic picks ciw over cw: ciw replaces the whole word from anywhere inside it. Pair it with * (find the word under the cursor) and n. to rename one hit at a time, skipping the ones you want to keep.

Counts#

After 2dw, a bare . deletes two more words. 3. deletes three: a count on dot replaces the original one.

The boundary#

• An arrow key in Insert mode ends the change. A1<Left>2<Esc> then . repeats only the 2.
• Ex commands are not changes. After :s/a/b/, . replays your last Normal mode change instead. Repeat a substitution with & on this line, or g& on every line.

Rule of thumb#

Before an edit, spend half a second asking whether . could do the next one. A instead of $a costs nothing now and pays every time you press dot. It is the cheapest automation you will ever write: one key, no YAML.