respecting the active region

[ultronozm]

Hi, do you think it'd be feasible to adapt vundo so that when it is called with an active region, it shows and undoes only those changes that affect that region, similar to what the built-in undo does?

[casouri]

Sadly, no, vundo wouldn't be able to reconstruct the tree if you use "undo in region". Actually, even now, if you use "undo in region", that undo will be considered a buffer edit rather than an undo, because by doing a partial undo, you're not going back to a previous buffer state, you're essentially creating a new one.

[ultronozm]

Thanks for the quick response. I think my question might have been unclear:

If I understand correctly, you say that vundo won't reconstruct the tree after you've called region-based undo, because doing so creates new buffer states. I agree.

I'm interested in something different: using vundo to preview the effects of region-based undo.

As far as I can tell, there's no fundamental limitation -- it should amount to preprocessing the buffer-undo-list.

Does this clarification change your answer?

[casouri]

Sadly it's still not a simple feature to add. Because vundo doesn't directly call undo/redo commands when you move around the undo tree. We actually compute a series of entries taken from buffer-undo-list to feed to primitive-undo to modify the buffer.

Also, if you move back to a node in the tree using a limited region, it's not clear how to move back: there's no appropriate undo entries that we can feed directly to primitive-undo.

[jdtsmith]

I've thought of this too, would be useful. I think it should be possible, with some effort and help from upstream. The key difference with a region active: instead of buffer-undo-list, undo-in-region gets a newly created custom undo list from the "alternative universe" in which all edits had happened within the region BEG...END (see undo-make-selective-list). Since it's a new list with modified (but otherwise normal) undo records, none of the "equivalencies" that permit vundo to create a tree from the linear list pertain.

But I think we can do the following:

• Get an association map from "in region" nodes in the normal buffer-undo-list to nodes of the newly created undo-make-selective-list (tip of the list). This is a one-to-one mapping. We'd need help from upstream for this.
• Draw the normal vundo tree, using the normal algorithm on the normal buffer-undo-list. Nodes which are not "in region" get grayed out, and are skipped during navigation. We can even visually compress linear parts of the tree like ○──<12>──○ to show multiple out-of-region skipped nodes at once.
• Navigation is maybe not as tricky as it sounds: all region undos are actually new changes (no equivalencies), so you can move within the newly appended tail of in-region changes the same way you normally do. You just have to consult the extra association mapping for where to draw current node.
• None of these need changes actually show up in the tree, until the next time it is built.