This post explores an elegant method for implementing undo/redo functionality in text editors.
We wish to implement functionality that allows the user of a text editor to undo their last action. If they accidentally undo an action, we want to let them redo it.
For the purposes of this post, let's assume that a text editor makes edits to a
text object. This
text object is an array of string pointers1, with
each string representing a line of text2.
This technique for implementing undo is simple, but requires that the text
editor is written so that all edit commands are immutable. Immutability means
that the edit commands won't change the
text object, but will return a copy of
text, with the changes made to it. If we store each of these text objects
(let's call them 'states'), we can undo by reverting the text to the previous
Making a full copy of the
text object on every edit would be wasteful. We can
make things better by copying pointers, rather than full objects.
Consider a document which contains the text:
Let's assume this is stored in an array of pointers to strings. Our strings are stored at two locations in memory3:
0: "Hello" 1: "world"
Our array contains pointers to those memory locations:
When we make an edit, we create a new array of strings, where all pointers are the same, apart from the edited one:
0: "Hello" 1: "world" 2: "James"
The trick to implementing undo is to realise that our previous array, containing
"Hello\nworld" still exists. To undo our changes, we simply revert
to the previous array.
To implement infinite history, we store the states in a linked list. To undo, we move back a state. To redo, we move forward a state. Infinite undo history can take up an unbounded amount of memory, and with high enough usage will cause the program to crash.
To implement finite undo, we must store the states in a doubly linked list. We keep track of the length of the list by incrementing a counter whenever a new state is added to the list. We decrement this counter on undo and increment it again on redo. If the length exceeds some predefined limit, we delete the state from the tail end of the list.
When deleting the oldest state, we need to be careful to make sure we don't delete any strings which are reused by other states. In a garbage collected language, this is managed by the runtime.
We remove the state from our linked list. Because it is no longer used, it can be deleted. The strings that are only used in that state can also be deleted. Strings used in other states are kept.
This technique can be used in non-garbage collected languages, but this object management must be done manually.
This technique doesn't actually require a language with pointers; Pointers just offer a convenient notation for this example.
This isn't actually a good way to store text in a text editor. For a better solution, see a post I wrote about gap buffers.
Note that the memory addresses in this example have been simplified for the sake of clarity and don't make physical sense.