Dired patch for `i' - bounce between subdir line and its inserted listing

Dired patch for `i' - bounce between subdir line and its inserted listing

[Drew Adams]

[-- Attachment #1: Type: text/plain, Size: 1760 bytes --]

Attached is a patch for Dired (dired.el and dired-aux.el).
It extends `i', giving it behavior on a non-directory line
and on subdir header lines, where it currently just raises a
wrong-line error.
 
This provides a handy go-back behavior, letting you bounce
back and forth between a subdirectory line and its inserted
listing.  `i' on a non-directory file within a subdir
listing acts the same as it does on the subdir's header
line, so you need not be on the header line to bounce back:
anywhere within the subdir listing will do.
 
More precisely -
 
Currently:
 
* If you hit `i' on a directory line, it goes to the
  corresponding subdirectory listing header line.  If the
  subdirectory listing has not yet been inserted, `i' first
  inserts it and then goes to it.
 
* If you hit `i' anywhere else, it raises an error: either
  "No file on this line" (for a header line) or "Attempt to
  insert a non-directory: c:/foobar".  This is not useful.
 
With the patch:
 
* `i' on a directory line does the same thing as before:
  insert and goto.
 
* `i' on a non-directory line in the top-level directory
  also does the same thing as before: nothing.
 
* (NEW) `i' on a subdir header line or a non-directory file
  in a subdir listing goes to the line for that subdir in
  the parent directory listing.

Please try the patch.  It should work in all contexts,
including with hidden subdirs.
 
The patch also fixes a bug that occurs in
`dired-get-filename' if you use the `ls' switch `F', as in
`ls -alF'.  In that case, the test for a dot file `.' or
`..' fails, because with switch `F' the names appear as `./'
and `../'.
 
BTW, I'm surprised that bug wasn't reported before.  Don't
you all use `F'?  And shouldn't `ls -alF' be the default
switches value?
 

[-- Attachment #2: dired-aux-2008-03-07.patch --]
[-- Type: application/octet-stream, Size: 5667 bytes --]

diff -c -w dired-aux-CVS-2008-03-07.el dired-aux-patched-2008-03-07.el
*** dired-aux-CVS-2008-03-07.el	Sat Mar  8 00:00:02 2008
--- dired-aux-patched-2008-03-07.el	Sat Mar  8 00:07:24 2008
***************
*** 1858,1895 ****
  ;;;###begin dired-ins.el
  
  ;;;###autoload
! (defun dired-maybe-insert-subdir (dirname &optional
! 					  switches no-error-if-not-dir-p)
!   "Insert this subdirectory into the same dired buffer.
! If it is already present, just move to it (type \\[dired-do-redisplay] to refresh),
!   else inserts it at its natural place (as `ls -lR' would have done).
! With a prefix arg, you may edit the ls switches used for this listing.
!   You can add `R' to the switches to expand the whole tree starting at
!   this subdirectory.
! This function takes some pains to conform to `ls -lR' output.
! 
! Dired remembers switches specified with a prefix arg, so that reverting
! the buffer will not reset them.  However, using `dired-undo' to re-insert
! or delete subdirectories can bypass this machinery.  Hence, you sometimes
! may have to reset some subdirectory switches after a `dired-undo'.
! You can reset all subdirectory switches to the default using
! \\<dired-mode-map>\\[dired-reset-subdir-switches].
! See Info node `(emacs)Subdir switches' for more details."
!   (interactive
!    (list (dired-get-filename)
! 	 (if current-prefix-arg
  	     (read-string "Switches for listing: "
  			  (or dired-subdir-switches dired-actual-switches)))))
!   (let ((opoint (point)))
      ;; We don't need a marker for opoint as the subdir is always
      ;; inserted *after* opoint.
      (setq dirname (file-name-as-directory dirname))
      (or (and (not switches)
  	     (dired-goto-subdir dirname))
  	(dired-insert-subdir dirname switches no-error-if-not-dir-p))
!     ;; Push mark so that it's easy to find back.  Do this after the
!     ;; insert message so that the user sees the `Mark set' message.
!     (push-mark opoint)))
  
  ;;;###autoload
  (defun dired-insert-subdir (dirname &optional switches no-error-if-not-dir-p)
--- 1858,1928 ----
  ;;;###begin dired-ins.el
  
  ;;;###autoload
! (defun dired-maybe-insert-subdir (dirname &optional switches no-error-if-not-dir-p)
!   "Move to Dired subdirectory line or subdirectory listing.
! This bounces you back and forth between a subdirectory line and its
! inserted listing header line.  Using it on a non-directory line in a
! subdirectory listing acts the same as using it on the subdirectory
! header line.
! 
! * If on a subdirectory line, then go to the subdirectory's listing,
!   creating it if not yet present.
! 
! * If on a subdirectory listing header line or a non-directory file in
!   a subdirectory listing, then go to the line for the subdirectory in
!   the parent directory listing.
! 
! * If on a non-directory file in the top Dired directory listing, do
!   nothing.
! 
! Subdirectories are listed in the same position as for `ls -lR' output.
! 
! With a prefix arg, you can edit the `ls' switches used for this
! listing.  Add `R' to the switches to expand the directory tree under a
! subdirectory.
! 
! Dired remembers the switches you specify with a prefix arg, so
! reverting the buffer does not reset them.  However, you might
! sometimes need to reset some subdirectory switches after a
! `dired-undo'.  You can reset all subdirectory switches to the
! default value using \\<dired-mode-map>\\[dired-reset-subdir-switches].  See \
! Info node
! `(emacs)Subdir switches' for more details."
!   (interactive (list (dired-this-subdir)
!                      (and current-prefix-arg
                            (read-string "Switches for listing: "
                                         (or dired-subdir-switches dired-actual-switches)))))
!   (let ((opoint (point))
!         (filename dirname))
!     (if (consp filename)            ; Subdir header line or non-directory file.
!         (progn (setq filename (car filename))
!                (if (assoc filename dired-subdir-alist)
!                    (dired-goto-file filename) ;  Subdir header line.
!                  (dired-insert-subdir
!                   (substring (file-name-directory filename) 0 -1))))
        ;; We don't need a marker for opoint as the subdir is always
        ;; inserted *after* opoint.
        (setq dirname (file-name-as-directory dirname))
        (or (and (not switches)
                 (dired-goto-subdir dirname))
            (dired-insert-subdir dirname switches no-error-if-not-dir-p))
!       ;; Push mark so that it's easy to go back.  Do this after the
!       ;; insertion message so that the user sees the `Mark set' message.
!       (push-mark opoint))))
! 
! (defun dired-this-subdir ()
!   "This line's filename, if directory, or `dired-current-directory' list.
! If on a directory line, then return the directory name.
! Else return a singleton list of a directory name, which is as follows:
!   If on a subdirectory header line (either of the two lines), then use
!   that subdirectory name.  Else use the parent directory name."
!   (or (let ((file (dired-get-filename nil t)))
!         (and file (file-directory-p file)
!              (not (member (file-relative-name file (file-name-directory
!                                                     (directory-file-name file)))
!                           '("." ".." "./" "../")))
!              file))
!       (list (dired-current-directory))))
  
  ;;;###autoload
  (defun dired-insert-subdir (dirname &optional switches no-error-if-not-dir-p)

Diff finished.  Sat Mar 08 00:15:25 2008

[-- Attachment #3: dired-2008-03-07.patch --]
[-- Type: application/octet-stream, Size: 6282 bytes --]

diff -c -w dired-CVS-2008-03-07.el dired-patched-2008-03-07.el
*** dired-CVS-2008-03-07.el	Fri Mar  7 23:59:44 2008
--- dired-patched-2008-03-07.el	Sat Mar  8 00:13:40 2008
***************
*** 1823,1837 ****
        (if (setq p1 (dired-move-to-filename (not no-error-if-not-filep)))
  	  (setq p2 (dired-move-to-end-of-filename no-error-if-not-filep))))
      ;; nil if no file on this line, but no-error-if-not-filep is t:
!     (if (setq file (and p1 p2 (buffer-substring p1 p2)))
! 	(progn
  	  ;; Get rid of the mouse-face property that file names have.
  	  (set-text-properties 0 (length file) nil file)
  	  ;; Unquote names quoted by ls or by dired-insert-directory.
  	  ;; Using read to unquote is much faster than substituting
  	  ;; \007 (4 chars) -> ^G  (1 char) etc. in a lisp loop.
! 	  (setq file
! 		(read
  		 (concat "\""
  			 ;; Some ls -b don't escape quotes, argh!
  			 ;; This is not needed for GNU ls, though.
--- 1823,1835 ----
        (if (setq p1 (dired-move-to-filename (not no-error-if-not-filep)))
            (setq p2 (dired-move-to-end-of-filename no-error-if-not-filep))))
      ;; nil if no file on this line, but no-error-if-not-filep is t:
!     (when (setq file (and p1 p2 (buffer-substring p1 p2)))
        ;; Get rid of the mouse-face property that file names have.
        (set-text-properties 0 (length file) nil file)
        ;; Unquote names quoted by ls or by dired-insert-directory.
        ;; Using read to unquote is much faster than substituting
        ;; \007 (4 chars) -> ^G  (1 char) etc. in a lisp loop.
!       (setq file (read
                    (concat "\""
                            ;; Some ls -b don't escape quotes, argh!
                            ;; This is not needed for GNU ls, though.
***************
*** 1841,1861 ****
  			 "\"")))
  	  ;; The above `read' will return a unibyte string if FILE
  	  ;; contains eight-bit-control/graphic characters.
! 	  (if (and enable-multibyte-characters
  		   (not (multibyte-string-p file)))
! 	      (setq file (string-to-multibyte file)))))
      (and file (file-name-absolute-p file)
  	 ;; A relative file name can start with ~.
  	 ;; Don't treat it as absolute in this context.
  	 (not (eq (aref file 0) ?~))
  	 (setq already-absolute t))
!     (cond
!      ((null file)
!       nil)
!      ((eq localp 'verbatim)
!       file)
!      ((and (not no-error-if-not-filep)
! 	   (member file '("." "..")))
        (error "Cannot operate on `.' or `..'"))
       ((and (eq localp 'no-dir) already-absolute)
        (file-name-nondirectory file))
--- 1839,1855 ----
                            "\"")))
        ;; The above `read' will return a unibyte string if FILE
        ;; contains eight-bit-control/graphic characters.
!       (when (and enable-multibyte-characters
                   (not (multibyte-string-p file)))
!         (setq file (string-to-multibyte file))))
      (and file (file-name-absolute-p file)
           ;; A relative file name can start with ~.
           ;; Don't treat it as absolute in this context.
           (not (eq (aref file 0) ?~))
           (setq already-absolute t))
!     (cond ((null file) nil)
!           ((eq localp 'verbatim) file)
!           ((and (not no-error-if-not-filep) (member file '("." ".." "./" "../")))
             (error "Cannot operate on `.' or `..'"))
            ((and (eq localp 'no-dir) already-absolute)
             (file-name-nondirectory file))
***************
*** 1867,1874 ****
  	(if (and handler (not (get handler 'safe-magic)))
  	    (concat "/:" file)
  	  file)))
!      ((eq localp 'no-dir)
!       file)
       ((equal (dired-current-directory) "/")
        (setq file (concat (dired-current-directory localp) file))
        (let ((handler (find-file-name-handler file nil)))
--- 1861,1867 ----
               (if (and handler (not (get handler 'safe-magic)))
                   (concat "/:" file)
                 file)))
!           ((eq localp 'no-dir) file)
            ((equal (dired-current-directory) "/")
             (setq file (concat (dired-current-directory localp) file))
             (let ((handler (find-file-name-handler file nil)))
***************
*** 1878,1885 ****
  	(if (and handler (not (get handler 'safe-magic)))
  	    (concat "/:" file)
  	  file)))
!      (t
!       (concat (dired-current-directory localp) file)))))
  
  (defun dired-string-replace-match (regexp string newtext
                                     &optional literal global)
--- 1871,1877 ----
               (if (and handler (not (get handler 'safe-magic)))
                   (concat "/:" file)
                 file)))
!           (t (concat (dired-current-directory localp) file)))))
  
  (defun dired-string-replace-match (regexp string newtext
                                     &optional literal global)
***************
*** 2324,2333 ****
  			;; the search faster (e.g. for the filename "-"!).
   			(search-forward (concat " " search-string)
  					boundary 'move))
  	      ;; Match could have BASE just as initial substring or
  	      ;; or in permission bits or date or
  	      ;; not be a proper filename at all:
! 	      (if (equal base (dired-get-filename 'no-dir t))
  		    ;; Must move to filename since an (actually
  		    ;; correct) match could have been elsewhere on the
  		    ;; ;; line (e.g. "-" would match somewhere in the
--- 2316,2326 ----
                          ;; the search faster (e.g. for the filename "-"!).
                          (search-forward (concat " " search-string)
                                          boundary 'move))
+               ;; Remove / from filename, then compare with BASE.
                ;; Match could have BASE just as initial substring or
                ;; or in permission bits or date or
                ;; not be a proper filename at all:
!               (if (equal base (directory-file-name (dired-get-filename 'no-dir t)))
                    ;; Must move to filename since an (actually
                    ;; correct) match could have been elsewhere on the
                    ;; ;; line (e.g. "-" would match somewhere in the

Diff finished.  Sat Mar 08 00:14:02 2008

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

>  This provides a handy go-back behavior, letting you bounce
>  back and forth between a subdirectory line and its inserted
>  listing.  `i' on a non-directory file within a subdir
>  listing acts the same as it does on the subdir's header
>  line, so you need not be on the header line to bounce back:
>  anywhere within the subdir listing will do.

I would like this. Sometimes I manually call `i' on specific dirs and
get tired of needing to do C-u C-SPC to get back. Double tapping `i'
would be much more convenient. Maybe, maybe maybe maybe, should it
also step one line down from the original dir line when jumping back,
but that depends on how you use it I guess, it would fit my way of
working with dirs in Dired. It has the disadvantage however that you
cannot go back to the inserted dir again by tapping `i' for a third
time.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Juri Linkov]

>>  This provides a handy go-back behavior, letting you bounce
>>  back and forth between a subdirectory line and its inserted
>>  listing.  `i' on a non-directory file within a subdir
>>  listing acts the same as it does on the subdir's header
>>  line, so you need not be on the header line to bounce back:
>>  anywhere within the subdir listing will do.
>
> I would like this. Sometimes I manually call `i' on specific dirs and
> get tired of needing to do C-u C-SPC to get back.

You already can do this easily with `^'.  It have advantage of being
context-independent: it works everywhere in any part of the subdirectory
listing.  OTOH, overloading a clear meaning of `i' ("insert a subdir")
with more meanings will add inconsistencies.

> Double tapping `i' would be much more convenient. Maybe, maybe maybe
> maybe, should it also step one line down from the original dir line
> when jumping back, but that depends on how you use it I guess, it
> would fit my way of working with dirs in Dired. It has the
> disadvantage however that you cannot go back to the inserted dir again
> by tapping `i' for a third time.

It seems what you are trying to do here is to insert a group of
subdirectories.  Maybe, it would be much easier just to mark them
with e.g. `* /' (`dired-mark-directories') and type `i' to insert
them all at once into the same dired buffer.

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

>  You already can do this easily with `^'.

Oh? I use that to "back up" all the time but never thought of using it
in that context. Cool!

>  It have advantage of being
>  context-independent: it works everywhere in any part of the subdirectory
>  listing.  OTOH, overloading a clear meaning of `i' ("insert a subdir")
>  with more meanings will add inconsistencies.

I just found a problem with this, maybe that is what you find
inconsistent: sometimes when inserting a dir you end up on another dir
and sometimes a normal file. This means that double tapping `i' would
not always take me back to where I came from because some of the times
it would insert a sub dir instead. We should not mess up the user's
ability to get a muscle memory of an operation like this (i.e. he
cannot rely on double tapping `i' doing the same thing each time, he
needs to know the content of the dir to insert to know what is going
to happen).

I am therefore removing my vote for this change.

>  It seems what you are trying to do here is to insert a group of
>  subdirectories.  Maybe, it would be much easier just to mark them
>  with e.g. `* /' (`dired-mark-directories') and type `i' to insert
>  them all at once into the same dired buffer.

Yes, if `i' could do that I would probably use that, but it does not
work that way now.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Juri Linkov]

>>  It have advantage of being
>>  context-independent: it works everywhere in any part of the subdirectory
>>  listing.  OTOH, overloading a clear meaning of `i' ("insert a subdir")
>>  with more meanings will add inconsistencies.
>
> I just found a problem with this, maybe that is what you find
> inconsistent: sometimes when inserting a dir you end up on another dir
> and sometimes a normal file. This means that double tapping `i' would
> not always take me back to where I came from because some of the times
> it would insert a sub dir instead. We should not mess up the user's
> ability to get a muscle memory of an operation like this (i.e. he
> cannot rely on double tapping `i' doing the same thing each time, he
> needs to know the content of the dir to insert to know what is going
> to happen).

Yes, this is what I meant.

>>  It seems what you are trying to do here is to insert a group of
>>  subdirectories.  Maybe, it would be much easier just to mark them
>>  with e.g. `* /' (`dired-mark-directories') and type `i' to insert
>>  them all at once into the same dired buffer.
>
> Yes, if `i' could do that I would probably use that, but it does not
> work that way now.

But maybe you can just use the option `-lR' described in
(info "(emacs) Subdirectories in Dired")?

Oh, and I noticed that this node suggests using only `C-u C-SPC'
to go back, but not `^'.  Maybe we should mention `^' as well?

-- 
Juri Linkov
http://www.jurta.org/emacs/

RE: Dired patch for `i' - bounce between subdir line and its inserted listing

[Drew Adams]

> [With the proposed patch] sometimes when inserting a dir you end up on
> another dir and sometimes a normal file. This means that double tapping
> `i' would not always take me back to where I came from because some of
> the times it would insert a sub dir instead.

It's not about double-tapping. That was never described as the purpose of the
patch. Similarly, always taking you back where you came from was never described
as the purpose of the patch.

With or without the patch, `i' always inserts a directory when used on a
directory line. That's a feature (the original behavior of `i'), not a bug.

The question is what `i' should do when it is _not_ on a directory line. In
vanilla Emacs without the proposed patch it just raises an error: "Attempt to
insert a non-directory: c:/foo/bar/toto.el". With the patch, it goes to the
parent directory line, if there is one, in the same Dired buffer. Which of these
behaviors is more useful? That's the only question, because that's the only
change the patch makes.

`^' always goes up. If you are on a file or directory line in the top-level
Dired listing (that is, not within an inserted subdir listing), then `^' leaves
the current directory listing altogether. (It is true that you can always get
back by hitting `RET' in that parent directory listing.) With the patch, `i' in
that same context leaves you where you are. The behaviors are intentionally
different in that context. `i' (with the patch) is for moving around within a
given Dired buffer, among its subdir listings.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Juri Linkov]

> The question is what `i' should do when it is _not_ on a directory line. In
> vanilla Emacs without the proposed patch it just raises an error: "Attempt to
> insert a non-directory: c:/foo/bar/toto.el". With the patch, it goes to the
> parent directory line, if there is one, in the same Dired buffer. Which of these
> behaviors is more useful? That's the only question, because that's the only
> change the patch makes.

I think `i' on a file line should be reserved for something more useful.
`i' means "insert", so it makes sense to insert a file into the dired
buffer, as some file managers already do.

-- 
Juri Linkov
http://www.jurta.org/emacs/

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

>  It's not about double-tapping. That was never described as the purpose of the
>  patch. Similarly, always taking you back where you came from was never described
>  as the purpose of the patch.

Sorry for reading my thoughts into your proposal.

>  With or without the patch, `i' always inserts a directory when used on a
>  directory line. That's a feature (the original behavior of `i'), not a bug.

Yes, it is not a bug.

>  The question is what `i' should do when it is _not_ on a directory line. In
>  vanilla Emacs without the proposed patch it just raises an error: "Attempt to
>  insert a non-directory: c:/foo/bar/toto.el". With the patch, it goes to the
>  parent directory line, if there is one, in the same Dired buffer. Which of these
>  behaviors is more useful? That's the only question, because that's the only
>  change the patch makes.

I can only describe how I work and I would not use the patched `i' in
that way. I would have liked it for the double tap thingy but now that
I realized that have bad side effects I would not like it anymore :) I
would never use `i' on a non-dir line outside the context of inserting
the same dir. That's not saying other would not like it, of course.
Now that I know about `^' doing what I want I will probably start to
use that. It is unlikely that I do this on the top level by mistake so
it will work for me I think.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

>  >>  subdirectories.  Maybe, it would be much easier just to mark them
>  >>  with e.g. `* /' (`dired-mark-directories') and type `i' to insert
>  >>  them all at once into the same dired buffer.
>  >
>  > Yes, if `i' could do that I would probably use that, but it does not
>  > work that way now.
>
>  But maybe you can just use the option `-lR' described in
>  (info "(emacs) Subdirectories in Dired")?

I use that sometimes when I want to insert a whole dir tree but it
isn't a replacement for multiply marked dirs and `i'.

>  Oh, and I noticed that this node suggests using only `C-u C-SPC'
>  to go back, but not `^'.  Maybe we should mention `^' as well?

I think that would be good.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

>  I think `i' on a file line should be reserved for something more useful.
>  `i' means "insert", so it makes sense to insert a file into the dired
>  buffer, as some file managers already do.

Can you explain how that would work, inserting a file into the dired
buffer? Would it be like a copy, i.e. you execute `i', get a prompt
for a file, and then that file is copied there?

However, using the same key for different things in the same buffer,
depending on context, might not be a good idea. Depending on if you
use font lock, what colors etc, a user might not necessarily notice he
is not on a dir and might be surprised if it works very differently.
The current error message at least makes him aware of the fact that he
is not on a dir line.

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Richard Stallman]

Is double-tapping anything like double-dipping?

Re: Dired patch for `i' - bounce between subdir line and its inserted listing

[Mathias Dahl]

    Is double-tapping anything like double-dipping?

Someone here has watched too much "Seinfeld" lately :)