Empty directories

Empty directories

[Yuri D'Elia]

$ mkdir empty
$ cd empty<TAB>

will just stop at 'empty/'.

Is there a way to make directories/local-directories accept empty
directories as valid?

Re: Empty directories

[Yuri D'Elia]

On 10/01/2014 12:48 PM, Yuri D'Elia wrote:
> $ mkdir empty
> $ cd empty<TAB>
> 
> will just stop at 'empty/'.
> 
> Is there a way to make directories/local-directories accept empty
> directories as valid?

I should perhaps state that the obvious reasoning behind this is the
"rmdir" completion.

Re: Empty directories

[Peter Stephenson]

On Wed, 01 Oct 2014 12:50:54 +0200
Yuri D'Elia <wavexx@thregr.org> wrote:
> On 10/01/2014 12:48 PM, Yuri D'Elia wrote:
> > $ mkdir empty
> > $ cd empty<TAB>
> > 
> > will just stop at 'empty/'.
> > 
> > Is there a way to make directories/local-directories accept empty
> > directories as valid?
> 
> I should perhaps state that the obvious reasoning behind this is the
> "rmdir" completion.

I'm still not understanding.  Like cd completion, "rmdir empty<TAB>"
will also stop at "empty/" (with a / that disappears if you hit return
immediately). But what else could you want?

pws

Re: Empty directories

[Yuri D'Elia]

On 10/01/2014 01:00 PM, Peter Stephenson wrote:
> On Wed, 01 Oct 2014 12:50:54 +0200
> Yuri D'Elia <wavexx@thregr.org> wrote:
>> On 10/01/2014 12:48 PM, Yuri D'Elia wrote:
>>> $ mkdir empty
>>> $ cd empty<TAB>
>>>
>>> will just stop at 'empty/'.
>>>
>>> Is there a way to make directories/local-directories accept empty
>>> directories as valid?
>>
>> I should perhaps state that the obvious reasoning behind this is the
>> "rmdir" completion.
> 
> I'm still not understanding.  Like cd completion, "rmdir empty<TAB>"
> will also stop at "empty/" (with a / that disappears if you hit return
> immediately). But what else could you want?

In normal file completion, prefix<tab> will accept the entire path (if
unique) and insert a space after it.

$ touch pref1_file pref2_file
$ rm pref1<tab>

With an empty directory though the path is not accepted even though
there is nothing more to complete.

It's a minor difference, I know, but I wanted to know why it's there.

Re: Empty directories

[Peter Stephenson]

On Wed, 01 Oct 2014 14:47:48 +0200
Yuri D'Elia <wavexx@thregr.org> wrote:
> In normal file completion, prefix<tab> will accept the entire path (if
> unique) and insert a space after it.
> 
> $ touch pref1_file pref2_file
> $ rm pref1<tab>
> 
> With an empty directory though the path is not accepted even though
> there is nothing more to complete.
> 
> It's a minor difference, I know, but I wanted to know why it's there.

Oh.

When you hit tab the first time, it's just looking for something at the
current path level.  If it's a directory it just adds it.  It has no
idea whether or not it's an empty directory, that's the next level
down.

Then after it's added a / for that directory, you start completing
in that directory and it will now look for whether there's anything
there or not.

To do better, you need a special case in file completion to handle
directory *only* completion.  But it's tricky since in principle there
may be a mixture of directories that are empty and directories that
aren't which would need to be grouped differently.  There are already
way too many cases no one really understands in _path_files, so this
might need an intermediate level for use by cd, rmdir, etc.  This is a
lot of work for something this minor.

pws

Re: Empty directories

[Yuri D'Elia]

On 10/01/2014 03:25 PM, Peter Stephenson wrote:
> aren't which would need to be grouped differently.  There are already
> way too many cases no one really understands in _path_files, so this

This made me smile ;)

I tried to look into _path_files, and there's actually a special rule to
handle empty directories, though I got lost in the expansion loop.

Re: Empty directories

[Bart Schaefer]

On Oct 1,  4:47pm, Yuri D'Elia wrote:
}
} I tried to look into _path_files, and there's actually a special rule to
} handle empty directories, though I got lost in the expansion loop.

I'm not entirely sure (see "no one really understands") but I think that
rule is for handling partial path completion, that is, given "a/b/c" it
eliminates empty directories matching a* or a*/b* -- because it knows it
has to get as far as a*/b*/c*, so empty dirs along the way are useless.

Re: Empty directories

[Yuri D'Elia]

On 10/01/2014 05:47 PM, Bart Schaefer wrote:
> On Oct 1,  4:47pm, Yuri D'Elia wrote:
> }
> } I tried to look into _path_files, and there's actually a special rule to
> } handle empty directories, though I got lost in the expansion loop.
> 
> I'm not entirely sure (see "no one really understands") but I think that
> rule is for handling partial path completion, that is, given "a/b/c" it
> eliminates empty directories matching a* or a*/b* -- because it knows it
> has to get as far as a*/b*/c*, so empty dirs along the way are useless.

Ok, so maybe we can cut it short and avoid to append a slash on empty
directories directly on the first run?

This would stop path completion immediately.

Re: Empty directories

[Bart Schaefer]

[-- Attachment #1: Type: text/plain, Size: 668 bytes --]

On Oct 1, 2014 11:15 AM, "Yuri D'Elia" <wavexx@thregr.org> wrote:
>
> Ok, so maybe we can cut it short and avoid to append a slash on empty
> directories directly on the first run?

The trouble is there is no way to tell that a directory is empty except by
looking "inside" it.  If the directory is network-mounted or contains
thousands of files (or, worse, both) that could be an enormous performance
hit to save you typing one character.

We might be able to do something with the newly-added short-circuit glob
qualifier, but that doesn't help with the network mount issue, and it still
means manufacturing a path tail, testing it, and then throwing it away
again.

Re: Empty directories

[Yuri D'Elia]

On 10/01/2014 09:58 PM, Bart Schaefer wrote:
> On Oct 1, 2014 11:15 AM, "Yuri D'Elia" <wavexx@thregr.org> wrote:
>> Ok, so maybe we can cut it short and avoid to append a slash on empty
>> directories directly on the first run?
> 
> The trouble is there is no way to tell that a directory is empty except by
> looking "inside" it.  If the directory is network-mounted or contains
> thousands of files (or, worse, both) that could be an enormous performance
> hit to save you typing one character.
> 
> We might be able to do something with the newly-added short-circuit glob
> qualifier, but that doesn't help with the network mount issue, and it still
> means manufacturing a path tail, testing it, and then throwing it away
> again.

Very good point, yes.

Something "unexpected" on a same vein happens with glob_complete and
repeated slashes. Assuming "compinit" has been called, try with the
following for maximum enjoyment:

$ anything .//<TAB>

Re: Empty directories

[Ray Andrews]

On 10/01/2014 07:47 AM, Yuri D'Elia wrote:
> On 10/01/2014 03:25 PM, Peter Stephenson wrote:
>> aren't which would need to be grouped differently.  There are already
>> way too many cases no one really understands in _path_files, so this
> This made me smile ;)
>
> I tried to look into _path_files, and there's actually a special rule to
> handle empty directories, though I got lost in the expansion loop.
Yeah! Laugh now ;-)
I hope we're doing better than bash:

http://blog.erratasec.com/2014/09/the-shockingly-bad-code-of-bash.html

Re: Empty directories

[Bart Schaefer]

On Oct 1, 10:01am, Ray Andrews wrote:
}
} I hope we're doing better than bash:
} 
} http://blog.erratasec.com/2014/09/the-shockingly-bad-code-of-bash.html

We're not doing a whole lot better on the number of global variables,
I'm afraid, but we've at least modernized all the function signatures.

Re: Empty directories

[Vincent Lefevre]

On 2014-10-01 14:47:48 +0200, Yuri D'Elia wrote:
> On 10/01/2014 01:00 PM, Peter Stephenson wrote:
> > On Wed, 01 Oct 2014 12:50:54 +0200
> > Yuri D'Elia <wavexx@thregr.org> wrote:
> >> On 10/01/2014 12:48 PM, Yuri D'Elia wrote:
> >>> $ mkdir empty
> >>> $ cd empty<TAB>
> >>>
> >>> will just stop at 'empty/'.
> >>>
> >>> Is there a way to make directories/local-directories accept empty
> >>> directories as valid?
> >>
> >> I should perhaps state that the obvious reasoning behind this is the
> >> "rmdir" completion.
> > 
> > I'm still not understanding.  Like cd completion, "rmdir empty<TAB>"
> > will also stop at "empty/" (with a / that disappears if you hit return
> > immediately). But what else could you want?
> 
> In normal file completion, prefix<tab> will accept the entire path (if
> unique) and insert a space after it.
> 
> $ touch pref1_file pref2_file
> $ rm pref1<tab>
> 
> With an empty directory though the path is not accepted even though
> there is nothing more to complete.

Why do you think that it is not accepted? And what does "not accepted"
mean?

Perhaps you mean that the ending slash should not be there in case of
"leaf" directories, i.e. when there is nothing else to complete (no
other directories, or not enough permission to enter them)? The main
advantage I could see here is that you immediately get the information
that you have reached a leaf directory. Otherwise this is a bit
useless.

-- 
Vincent Lefèvre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)