32.9 Undo

Most buffers have an undo list, which records all changes made to the buffer's text so that they can be undone. (The buffers that don't have one are usually special-purpose buffers for which Emacs assumes that undoing is not useful.) All the primitives that modify the text in the buffer automatically add elements to the front of the undo list, which is in the variable buffer-undo-list.

Variable: buffer-undo-list

This variable's value is the undo list of the current buffer. A value of t disables the recording of undo information.

Here are the kinds of elements an undo list can have:

position

This kind of element records a previous value of point; undoing this element moves point to position. Ordinary cursor motion does not make any sort of undo record, but deletion operations use these entries to record where point was before the command.

(beg . end)

This kind of element indicates how to delete text that was inserted. Upon insertion, the text occupied the range beg--end in the buffer.

(text . position)

This kind of element indicates how to reinsert text that was deleted. The deleted text itself is the string text. The place to reinsert it is (abs position).

(t high . low)

This kind of element indicates that an unmodified buffer became modified. The elements high and low are two integers, each recording 16 bits of the visited file's modification time as of when it was previously visited or saved. primitive-undo uses those values to determine whether to mark the buffer as unmodified once again; it does so only if the file's modification time matches those numbers.

(nil property value beg . end)

This kind of element records a change in a text property. Here's how you might undo the change:

(put-text-property beg end property value)

(marker . adjustment)

This kind of element records the fact that the marker marker was relocated due to deletion of surrounding text, and that it moved adjustment character positions. Undoing this element moves marker - adjustment characters.

nil

This element is a boundary. The elements between two boundaries are called a change group; normally, each change group corresponds to one keyboard command, and undo commands normally undo an entire group as a unit.

Function: undo-boundary

This function places a boundary element in the undo list. The undo command stops at such a boundary, and successive undo commands undo to earlier and earlier boundaries. This function returns nil.

The editor command loop automatically creates an undo boundary before each key sequence is executed. Thus, each undo normally undoes the effects of one command. Self-inserting input characters are an exception. The command loop makes a boundary for the first such character; the next 19 consecutive self-inserting input characters do not make boundaries, and then the 20th does, and so on as long as self-inserting characters continue.

All buffer modifications add a boundary whenever the previous undoable change was made in some other buffer. This is to ensure that each command makes a boundary in each buffer where it makes changes.

Calling this function explicitly is useful for splitting the effects of a command into more than one unit. For example, query-replace calls undo-boundary after each replacement, so that the user can undo individual replacements one by one.

Function: primitive-undo count list

This is the basic function for undoing elements of an undo list. It undoes the first count elements of list, returning the rest of list. You could write this function in Lisp, but it is convenient to have it in C.

primitive-undo adds elements to the buffer's undo list when it changes the buffer. Undo commands avoid confusion by saving the undo list value at the beginning of a sequence of undo operations. Then the undo operations use and update the saved value. The new elements added by undoing are not part of this saved value, so they don't interfere with continuing to undo.