Re: Completion style like Vim

["Benjamin R. Haskell" <zsh@xxxxxxxxxx>]

On Thu, 20 Jan 2011, Michael Treibton wrote:

> On 17 November 2010 22:43, Michael Treibton wrote:
> > Ben --
> >
> > [ Sorry about the off-list reply -- that's gmail being a pain.
> > Hopefully I've sorted that now. ]

No problem.


> > On 16 November 2010 21:21, Benjamin R. Haskell wrote:
> > > [...]
> > >
> > > In any case, the following seems to work like Vim's command-line 
> > > completion:
> > >
> > > bindkey "^[[A" history-beginning-search-backward
> > > bindkey "^[[B" history-beginning-search-forward
> > >
> > > ^[[A == up arrow, ^[[B == down.  Sounds like you want ^n and ^p, 
> > > too.
> >
> > Thanks, Ben.  This works exactly as I want, except for one thing -- 
> > the cursor position.
> >
> > Let's say I have this in my history:
> >
> > get_foo
> > get_bar
> >
> > ... and assume also that I've got your bindings configured (which I 
> > do) as:
> >
> > bindkey "^[[A" history-beginning-search-backward
> > bindkey "^[[B" history-beginning-search-forward
> >
> > If I type in (at the zsh prompt):
> >
> > get_
> >
> > and press the up-arrow key and down-arrow key, the cursor position 
> > remains after the underscore.  Ideally, I'd love it if the cursor, 
> > when pressed in either up or down direction could complete the line, 
> > *and* move to the end of the line, because now, I find myself having 
> > to do this:
> >
> > get_
> >
> > press "up-arrow", and then press "End", because actually I made a 
> > typo at the end of the line I need to correct and/or was missing some 
> > options from the command in the first place.  With my previous 
> > bindings for those arrow-keys, this cursor-position was not a 
> > problem.
> >
> > I have no doubt zsh can do this, and I am learning about it all the 
> > time -- and I appreciate your help and Matt's so far.  So any 
> > insights you might have are warmly appreciated.  :)
>
> Was there any ideas on this, may I ask?

Sorry.  I never saw this for some reason...

I'm not sure there's a nice way to do this.  I think 
history-beginning-search-* uses everything preceding the cursor as the 
search.  So, if you push the cursor to EOL, your search changes, which 
might not be desirable:

e.g. history contains:

abcdef 123456
abcghi
abcdef

abc_ + the search would end up with: abcdef_ (cursor at end of line). 
Then 'abcghi' would be skipped in a successive search, even though it 
matches the original (abc), because it doesn't start with the new search 
(abcdef).

Something more clever, that would maybe:

1. reset the 'search term' on accept-line
2. on first run, keep track of the 'search term'
3. run history-beginning-search-*

_might_ do the trick.

--
Best,
Ben