This project is an experiment to implement the ideas described here:
The application is a simple grid that the user can "edit". Edit operations include:
- Writing a character to a cell.
- Insert a row/column.
- Delete a row/column.
- Undo the last edit operation (using Ctrl-z).
- Redo the last undo (using Ctrl-y).
The user can move a cursor using the arrow keys. Cursor moves are not undoable.
The experiment aims to demonstrate a method for implementing undo/redo that is easy for users to understand and allows them to return to any previous state, regardless of the undo/redo actions taken. This is different from traditional undo methods, where it can be difficult to recover a previous state if an action is taken after undoing.
I tried two different approaches to undoing individual user operations. Hence there are two undo experiments (in separate GitHub repos). The two approaches are:
Take a copy of the current grid state before making any changes. To undo the operation, revert to the stored copy.
While this may seem like it would use a lot of memory, I can use immutable data and take advantage of "structural sharing" to avoid multiple copies of parts of the state that do not change.
For the purpose of this experiment, the data representation I use is not optimized for structural sharing, but it's fine as a proof of concept.
For each possible operation that changes the grid state, calculate the sequence of operations needed to restore the original grid state. This should be more efficient than approach 1 in terms of memory usage. However, it is more complicated to implement, and it's easy to make a mistake in undo calculation, so that undo does not restore the state correctly.
I am more familar with approach 2. I'm not sure how well approach 1 would scale in a "real" application.