Experimental support for _message content as fzf header

[AndydeCleyre]

From discussion at #352, this draft PR is a place to explore possibilities for retaining messages from _message calls so that the user still sees them.

[AndydeCleyre]

Currently this doesn't touch -z4h-comp-files, and I don't know if it ought to.

[romkatv]

Currently this doesn't touch -z4h-comp-files, and I don't know if it ought to.

Let's ignore that for now. If needed, It'll be easy to handle -z4h-comp-files later, when we are happy with -z4h-comp-words.

[AndydeCleyre]

1. It's safe to consider extendedglob to be already set when -z4h-comp-words is evaluated, right?
2. My new "FIXUP" commit is in fact very wrong. With many-lined headers, even when they can fit with space to spare, the choices are invisible and there's no indication of the selected choice changing. And the headers get truncated I think.

I'll have another look at this in the evening I think.

[romkatv]

• It's safe to consider extendedglob to be already set when -z4h-comp-words is evaluated, right?

I don't see why it would be safe to assume that. What am I missing?

I'll have another look at this in the evening I think.

I'm here to review the change whenever you think it's ready. If I were doing it, I would run something like this with various arguments:

seq 3 | fzf --header=$'1\n2\n3\n4\n5\n'

Build a mental model of how exactly fzf displays the value --header, and then code the same login in zsh to compute the height.

[AndydeCleyre]

I don't see why it would be safe to assume that. What am I missing?

I guessed that just because it looks like the extended glob # operator is used in -z4h-main-complete without a local setopt (curcontext="${curcontext/:[^:]#:/:${_completer}:}", curcontext="${curcontext/:[^:]#:/::}"), and it looks like it's set in z4h.sh.

[romkatv]

-z4h-comp-words is called directly from z4h-fzf-complete. I don't think the latter enabled extended_glob.