*$var*/* and recursive globbing

*$var*/* and recursive globbing

[Stephane Chazelas]

Hiya,

With zsh (but also ksh -G and bash -O globstar),

*"$var"*/* where $var is empty, expands like **/* while it
sounds to me like not the right thing to do. It can be worked
around by doing (*"$var"*)/*, but that requires extended_glob.

I would say that's a case where there's a fair chance to break a
perfectly correct POSIX script (considering that **/ is not
disabled in sh emulation).

I would be of the opinion that **/ (and ***/) should recurse only when
entered as is (or at least the two stars entered as is).

What do you think?
-- 
Stephane

Re: *$var*/* and recursive globbing

[Bart Schaefer]

On Jan 26, 10:25pm, Stephane Chazelas wrote:
} Subject: *$var*/* and recursive globbing
}
} *"$var"*/* where $var is empty, expands like **/*
} 
} I would be of the opinion that **/ (and ***/) should recurse only when
} entered as is (or at least the two stars entered as is).

I tend to agree, but I suspect it's impossible without making ** and ***
into lexical tokens, which might have other unintended consequences (not
to mention a significant ripple effect through the globbing code, turning
this into potentially a large change).

Re: *$var*/* and recursive globbing

[Peter Stephenson]

On Sat, 26 Jan 2013 14:59:08 -0800
Bart Schaefer <schaefer@brasslantern.com> wrote:
> On Jan 26, 10:25pm, Stephane Chazelas wrote:
> } Subject: *$var*/* and recursive globbing
> }
> } *"$var"*/* where $var is empty, expands like **/*
> } 
> } I would be of the opinion that **/ (and ***/) should recurse only when
> } entered as is (or at least the two stars entered as is).
> 
> I tend to agree, but I suspect it's impossible without making ** and ***
> into lexical tokens, which might have other unintended consequences (not
> to mention a significant ripple effect through the globbing code, turning
> this into potentially a large change).

Possibly, though it ought to be doable.

It might also be possible to delay remnulargs(), though it looks like
there are earlier forms of expansion that depend on it, so it doesn't
look very hopeful.

-- 
Peter Stephenson <p.w.stephenson@ntlworld.com>
Web page now at http://homepage.ntlworld.com/p.w.stephenson/