Design for grouping undo steps when editing text

Question

Some undo for text editor editors handle text input differently.

• Every key press is a single undo step.
  Typing in a word N characters long, requires N undo steps.
• Undo steps use word boundaries
  Undoing will undo each word entered.
• Undo steps use timer-based boundaries
  If you stop typing N milliseconds - this adds an undo step.
• Undo steps add boundaries every N characters entered.

Given these different ways of handling undo for text input, are there strong reasons to pick one of these over another?

---

Edit, asking about this for text entry fields for a widget-toolkit.

Answer 1

Definitely. There are several reasons you might want to pick one over another.

For instance, if my inputs require a large number of characters, not case-sensitive or no strict adherence to pre-defined rules, I might most likely do with a time-based boundary undo step.

If I am inputting case sensitive characters, I might most definitely go for any other way rather than the above. The undo type I would now employ may now be dependent on some other factors like, how long each word is, how much attention to details do I really need etc.

It totally would be dependent on the general use of what the inputs are for.