Stephen Bach <[email protected]> writes: [Disclaimer: this reply only expresses my own opinion.] <snip>

Now, the single impacting regression I've run up against so far is with undo. Unlike Vim and my old version of Vimpulse, reverting a change with e.g. "cw" requires two undos instead of one. This is a pretty significant deviation, at least for my workflow.

I scanned through the code and saw vimpulse-connect-undos, currently disabled because of a dependency on undo-tree.el. A couple questions:

1. What makes undo-tree.el preferable to redo.el, for Vimpulse's purposes?

undo-tree.el provides undo tree functionality. Vim provides undo tree functionality. Vimpulse emulates Vim.

2. Is fixing "cw" and friends a priority, or is it something that users of Vimpulse have learned to live with?

I actually like Emacs' undo better, i.e. I like the current behaviour better than Vim's more extensive undo steps -- it's trivial to hit `u' twice instead of once, whereas it's non-trivial (impossible AFAIK) to only undo a part of an undo step. I personally would just get rid of the currently defunct vimpulse-connect-undos code entirely. OTOH, Vimpulse emulates Vim, so you can argue it should restore the bigger undo steps. Štěpán

On 2010-07-09 20:50, Stephen Bach wrote:

So, great work. :-)

Thanks!

A couple questions:

1. What makes undo-tree.el preferable to redo.el, for Vimpulse's purposes?

:undolist uses undo-tree.el.

2. Is fixing "cw" and friends a priority, or is it something that users of Vimpulse have learned to live with?

Both. :) I've fixed the undo in commit [99fb1aaa1e]. However, since some users have indeed grown used to the old behavior, I've made it contingent on the customization variable `vimpulse-want-change-undo'. For the old, fine-grained undo, add (setq vimpulse-want-change-undo nil) to .emacs. -- Vegard