3 ms·

I came across autojump a while back, but I have an aversion to 'learning' systems in that they can sometimes conflict with muscle-memory (you expect a certain t

by shabble 17y ago

I came across autojump a while back, but I have an aversion to 'learning' systems in that they can sometimes conflict with muscle-memory (you expect a certain thing to always perform a certain action, but through training it now does something different) I've just reinstalled it to check out, and I've noticed a couple of things that I dislike: 1) it doesn't seem to consistently tab-complete (j <tab><tab> returns a bunch of __1__/some/path entries) 2) It's not easy to cycle through/view multiple matches 3) I'm not sure how to modify the database (eg: when I complete a project and archive stuff, I probably don't want to jump there anymore, even though it might be quite highly rated) cdargs deals with most of these things in a better way, but doesn't have any of the learning features. I've used it now for a few months, and I'm in the process of hacking at it a bit to make it fit better into my workflow. One useful tip that applies probably equally to both is to add something like: "\C-f": "\C-a\C-kcdd \t\t" into ~/.inputrc for a binding style completion.

joelthelion 17y ago
About tab completion: The bunch of __1__ entries are the best matches, and you simply need to add the number you want and another tab to finish the completion. It's a bit hackish, but it works fine! Maybe I should document this somewhere though :) You can't directly edit the database, however autojump automatically removes directories that are not there anymore. Thus if you're done with a project, simply move the directory to archive/ or something, and autojump will forget about it. Otherwise the exponential decay works fast enough anyways. I understand learning systems are not for everybody, though. With autojump, I tried to make the behavior of the system as predictable as possible, but it's not and will never be fully deterministic. I accept this as a necessary bad counterpart to the nice features autojump has, but I understand some people will never like it.

by shabble 17y ago

joelthelion 17y ago

About tab completion: The bunch of __1__ entries are the best matches, and you simply need to add the number you want and another tab to finish the completion. It's a bit hackish, but it works fine! Maybe I should document this somewhere though :) You can't directly edit the database, however autojump automatically removes directories that are not there anymore. Thus if you're done with a project, simply move the directory to archive/ or something, and autojump will forget about it. Otherwise the exponential decay works fast enough anyways. I understand learning systems are not for everybody, though. With autojump, I tried to make the behavior of the system as predictable as possible, but it's not and will never be fully deterministic. I accept this as a necessary bad counterpart to the nice features autojump has, but I understand some people will never like it.