On 9/28/07, Beni Cherniavsky <[EMAIL PROTECTED]> wrote: > There was a lot of discussion about what is the best behaviour for > Alt+Backspace and Ctrl+W -- delete to last space / punctuation / slash > / start of argument... > > Here is a fresh idea: delete to the last point where pressing Tab > would show more than one option. E.g. if you have > "foo_bar_baz<cursor>" and the directory contains "foo_bar_baz" and > "foo_quux", it would leave "foo_". This should create a pleasant > experience of navigating the tree of possible completions, deleting a > lot in simple places and a little in complex ones. > > It might (or might not) be better to constrain the above definition to > boundaries between letters, digits and punctuation. This would be > less surprising to users (won't stop in middle of words), while still > doing the right thing in most cases.
Cool idea. > > It is of course not easy to implement when you consider programmable > completion. Right, that is definitely an issue. Long term, I'd ideally like to move out completion work to a separate process. That would make it possible to perform completions in the background and do new and cool things with them, like display suggestions in the background and lots of other stuff that won't work today because there is always a chance that running a completion could be too slow. This would be a bit of work, though. > Trying to run programmable completion at every possible position to > see whether it's unambiguous would probably be too slow. Binary > search should work in theory but might be too slow as well. > Perhaps just generating completions after the first letter and finding > the last ambiguous place statically by looking at all the possible > completions would work. Or not. > Anyway, I think it's worth trying to implement it. > > In any case, I think there should be a separate key (probably Ctrl+W) > that deletes the whole last argument. You mean that there should be two separate commands, delete-argument and delete-word? They seem to be a bit to close to each other to avoid confusion, but I don't know. Possibly. > > -- > Beni Cherniavsky <[EMAIL PROTECTED]> (I read email only on weekends) -- Axel ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Fish-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/fish-users
