Re: Preventing sorting in completers

[Bart Schaefer <schaefer@brasslantern.com>]

On Jul 15, 10:42am, Peter Stephenson wrote:
}
} It occurred to me we could test with the tag, then without, for
} compatibility; something similar is done in other places, though
} usually with some default tag.

That'd be reasonable, though it still requires the special case for
"menu" for _expand and therefore we can't remove code from _expand.
We might be able to remove it from _history.

} > OTOH, I wonder whether _description should NOT use the style when
} > an explicit -J or -V argument has been passed?
} 
} That's certainly arguable, but there's still something to be said
} for giving people control anyway, in the same way you can override
} the set of matched files.

Slightly revised suggestion:  Allow styles to override an explicit -J,
but do not allow overriding an explicit -V.  I did a quick grep, and
there are very few cases where the -V option is passed; but in none of
those would it be useful to apply a sort.

} I think the question is: are people likely to use a general enough
} context that it gives unexpected effects with sorting elsewhere?

Considering that in nearly all cases the style would be used to turn
sorting _off_ rather than _on_, that strikes me as unlikely.  However,
there are cases (e.g., _gdb) where nearly all contexts are sorted but
one specific one (baud rate values) is unsorted, so if the style _is_
used to enable sorting, the possibility does exist.