Passing Emacs variables to minibuffer shell commands - emacs

I can run a shell command quickly by hitting M-!. One thing I'd like to do is perform shell quick operations on the current file. An example would be checking the file out through perforce:
M-! p4 edit buffer-file-name RET
(Yes there are perforce integrations, but I'm more interested in the minishell/variable problem rather than a specific workflow)
Of course, the buffer-file-name variable is not evaluated before the command is sent to the shell.
Is there an easy on-the-fly way to do this? Or will I have to roll a custom elisp function?

It seems current Emacs has something built-in to achieve the desired result, after M-! (shell-command) press <down>, you will get the file name you are currently visiting on the prompt. Now you can edit it to add the command you want to run on it.
In dired-mode it will give you the file your cursor is currently on.

Indeed using C-u M-: is almost right. I'm not so sure about using shell-quote-argument in eval-to-shell-argument since it only works on strings making it impossible to use eval-to-shell-argument to insert a number or a symbol. You could try something like:
(defun sm-minibuffer-insert-val (exp)
(interactive
(list (let ((enable-recursive-minibuffers t))
(read-from-minibuffer "Insert: "
nil read-expression-map t
'read-expression-history))))
(let ((val (with-selected-window (minibuffer-selected-window)
(eval exp)))
(standard-output (current-buffer)))
(prin1 val)))
and then bind this function in your minibuffer with (define-key minibuffer-local-map [?\M-:] 'sm-minibuffer-insert-val).
Of course, if the only thing you ever want to insert is the buffer-file-name, then your execute-shell-command-on-buffer is simpler.

I did roll my own elisp function, and it looks like this:
(defun execute-shell-command-on-buffer (shell-command-text)
(interactive "MShell command:")
(shell-command (format shell-command-text (shell-quote-argument buffer-file-name)))
)
https://gist.github.com/2367513
I bound it to M-", so now my example can be completed with:
M-"p4 edit %sRET
I won't accept this as the answer, because I did ask for solutions that don't require a function.

You can use C-u M-: (eval-expression with a universal prefix argument) to evaluate any Lisp expression and insert its value at point in the current buffer (including minibuffers, as long as you have enable-recursive-minibuffers set to a non-nil value).
In your example: C-u M-: buffer-file-name RET.
Note that the result of the expression is printed in Lisp form: that is, quoted in such a way that a subsequent call to read would construct an equal Lisp value. For strings, this means enclosing in double quotes, which will probably be interpreted as you expect by the inferior shell. However, you may run into problems with strings that contain special characters, which need different escaping by Elisp and the shell.
The more correct way uses shell-quote-argument, as in phils' solution. Here's a quick defun that reads a Lisp expression and inserts its value at point as a properly quoted shell word:
(defun eval-to-shell-argument (form)
(interactive "XEval: ")
(insert (shell-quote-argument form)))
The read-and-evaluate step happens automatically by using an "X" as the argument to interactive.
Edited to add: As #tenpn notes, the above solution doesn't work for inserting buffer-local variables like buffer-file-name in a minibuffer like the one M-! pops up (more precisely, it inserts the buffer-local value of the minibuffer, which is unlikely to be useful). Here is a revised version which seems to work. If the minibuffer is active, it makes the buffer of the previously-selected window temporarily active while reading and evaluating an expression.
Final edit: From #Stefan's answer I see that I should have used (minibuffer-selected-window) to find the previously-selected window. I've also added a (format "%s" ..) to allow inserting non-string values, while still quoting special characters in strings. Here's the final version:
(defun eval-to-shell-argument ()
(interactive)
(let* ((buffer
(if (minibufferp)
(window-buffer (minibuffer-selected-window))
(current-buffer)))
(result
(with-current-buffer buffer
(eval-minibuffer "Eval: "))))
(insert (shell-quote-argument (format "%s" result)))))

You can't do that with M-!, but you can evaluate arbitrary elisp from the minibuffer, so writing a function isn't strictly necessary:
M-: (shell-command (format "p4 edit %s" (shell-quote-argument buffer-file-name))) RET
In this case however, I think eshell is what you want to use:
M-x eshell-command RET p4 edit (eval buffer-file-name) RET
Edit: Except unfortunately that doesn't work, as the *eshell cmd* buffer is selected when that is evaluated. One solution would be:
M-x eshell-command RET p4 edit (eval buffer-file-name (other-buffer nil t)) RET
(Not quite as elegant, sorry.)

Everyone seems to be rolling their own version, so here's mine -- it will substitue the current filename or marked dired-files or current dired file wherever a % is in the shell command. It follows the same conventions as M-! so I bind it to that.
(defun my-shell-command (command &optional output-buffer error-buffer)
"Run a shell command with the current file (or marked dired files).
In the shell command, the file(s) will be substituted wherever a '%' is."
(interactive (list (read-from-minibuffer "Shell command: "
nil nil nil 'shell-command-history)
current-prefix-arg
shell-command-default-error-buffer))
(cond ((buffer-file-name)
(setq command (replace-regexp-in-string "%" (buffer-file-name) command nil t)))
((and (equal major-mode 'dired-mode) (save-excursion (dired-move-to-filename)))
(setq command (replace-regexp-in-string "%" (mapconcat 'identity (dired-get-marked-files) " ") command nil t))))
(shell-command command output-buffer error-buffer))

Related

M-x shell open shell in other windows

In previous emacs, "M-x shell" new a shell buffer in current windows.
but recently, I update my emacs to GNU Emacs 26.0.50.2. "M-x shell" new a shell buffer in the other windows. I search google and could not find answer.
does anybody know how to prevent this behavior.
Add this line in your .emacs file:
(push (cons "\\*shell\\*" display-buffer--same-window-action) display-buffer-alist)
This fixed it for me. I'm using eMacs 25.2 (9.0) on a Mac, and was getting really annoyed by the shell opening up in another frame, or even a new frame when there was only one.
This is the source where I got this answer from.
Unless you're customizing the names of your shell buffers, this should be all you need:
(add-to-list 'display-buffer-alist
`(,(regexp-quote "*shell") display-buffer-same-window))
To handle all shell buffers, whatever their names, you can use this advice:
(defun shell-same-window-advice (orig-fn &optional buffer)
"Advice to make `shell' reuse the current window.
Intended as :around advice."
(let* ((buffer-regexp
(regexp-quote
(cond ((bufferp buffer) (buffer-name buffer))
((stringp buffer) buffer)
(:else "*shell*"))))
(display-buffer-alist
(cons `(,buffer-regexp display-buffer-same-window)
display-buffer-alist)))
(funcall orig-fn buffer)))
(advice-add 'shell :around #'shell-same-window-advice)
The reason is that shell uses (pop-to-buffer buffer) instead of (switch-to-buffer buffer). I don't know how to advise the function, so I can't give you an appropriate answer. However if you only want shell to work the way you want you can simply add the whole function in your config.
Maybe somebody else can replace the function with an advice. I would be interested in that solution.
(defun shell (&optional buffer)
"Run an inferior shell, with I/O through BUFFER (which defaults to `*shell*').
Interactively, a prefix arg means to prompt for BUFFER.
If `default-directory' is a remote file name, it is also prompted
to change if called with a prefix arg.
If BUFFER exists but shell process is not running, make new shell.
If BUFFER exists and shell process is running, just switch to BUFFER.
Program used comes from variable `explicit-shell-file-name',
or (if that is nil) from the ESHELL environment variable,
or (if that is nil) from `shell-file-name'.
If a file `~/.emacs_SHELLNAME' exists, or `~/.emacs.d/init_SHELLNAME.sh',
it is given as initial input (but this may be lost, due to a timing
error, if the shell discards input when it starts up).
The buffer is put in Shell mode, giving commands for sending input
and controlling the subjobs of the shell. See `shell-mode'.
See also the variable `shell-prompt-pattern'.
To specify a coding system for converting non-ASCII characters
in the input and output to the shell, use \\[universal-coding-system-argument]
before \\[shell]. You can also specify this with \\[set-buffer-process-coding-system]
in the shell buffer, after you start the shell.
The default comes from `process-coding-system-alist' and
`default-process-coding-system'.
The shell file name (sans directories) is used to make a symbol name
such as `explicit-csh-args'. If that symbol is a variable,
its value is used as a list of arguments when invoking the shell.
Otherwise, one argument `-i' is passed to the shell.
\(Type \\[describe-mode] in the shell buffer for a list of commands.)"
(interactive
(list
(and current-prefix-arg
(prog1
(read-buffer "Shell buffer: "
;; If the current buffer is an inactive
;; shell buffer, use it as the default.
(if (and (eq major-mode 'shell-mode)
(null (get-buffer-process (current-buffer))))
(buffer-name)
(generate-new-buffer-name "*shell*")))
(if (file-remote-p default-directory)
;; It must be possible to declare a local default-directory.
;; FIXME: This can't be right: it changes the default-directory
;; of the current-buffer rather than of the *shell* buffer.
(setq default-directory
(expand-file-name
(read-directory-name
"Default directory: " default-directory default-directory
t nil))))))))
(setq buffer (if (or buffer (not (derived-mode-p 'shell-mode))
(comint-check-proc (current-buffer)))
(get-buffer-create (or buffer "*shell*"))
;; If the current buffer is a dead shell buffer, use it.
(current-buffer)))
;; On remote hosts, the local `shell-file-name' might be useless.
(if (and (called-interactively-p 'any)
(file-remote-p default-directory)
(null explicit-shell-file-name)
(null (getenv "ESHELL")))
(with-current-buffer buffer
(set (make-local-variable 'explicit-shell-file-name)
(file-remote-p
(expand-file-name
(read-file-name
"Remote shell path: " default-directory shell-file-name
t shell-file-name))
'localname))))
;; The buffer's window must be correctly set when we call comint (so
;; that comint sets the COLUMNS env var properly).
(switch-to-buffer buffer)
(unless (comint-check-proc buffer)
(let* ((prog (or explicit-shell-file-name
(getenv "ESHELL") shell-file-name))
(name (file-name-nondirectory prog))
(startfile (concat "~/.emacs_" name))
(xargs-name (intern-soft (concat "explicit-" name "-args"))))
(unless (file-exists-p startfile)
(setq startfile (concat user-emacs-directory "init_" name ".sh")))
(apply 'make-comint-in-buffer "shell" buffer prog
(if (file-exists-p startfile) startfile)
(if (and xargs-name (boundp xargs-name))
(symbol-value xargs-name)
'("-i")))
(shell-mode)))
buffer)

Emacs AucTeX; How to set C-c C-c default command?

I have set this in my .emacs file:
(add-hook 'TeX-mode-hook
(lambda ()
(setq TeX-command-default "LaTeX"))
(add-hook 'LaTeX-mode-hook
(lambda ()
(setq TeX-command-default "LaTeX"))
I see that C-c C-c is bound to TeX-command-master, which calls TeX-command-query. But since my (TeX-master-file) is "<none>", I expect the default command to be called, but keeps wanting to invoke "View" instead of "LaTeX".
If you check the source for TeX-command-query you'll find that it checks the modification date of the tex (lines 4-9) and bbl (lines 10-19) files involved in your document. Unless those files are more recent than the output file and there is no known next command to be performed (lines 20-22) it will use the "View" command as default (line 23).
This behaviour is of course sensible because normally you don't want to recompile unless there are changes (modified tex files). Apart from "patching" the command [posted below, would not really recommend to use because it will not receive automatic updates ;-) ] there isn't really anything you can do.
If you decide to use the patched command, just put is somewhere in your init file after the original command has been loaded. You could for example wrap it into (replace ;; BODY by code)
(eval-after-load "tex-buf"
'(progn
;; BODY
))
Here comes the patched command:
(defun TeX-command-query (name)
"Query the user for what TeX command to use."
(let* ((default
(cond ((if (string-equal name TeX-region)
(TeX-check-files (concat name "." (TeX-output-extension))
(list name)
TeX-file-extensions)
(TeX-save-document (TeX-master-file)))
TeX-command-default)
((and (memq major-mode '(doctex-mode latex-mode))
;; Want to know if bib file is newer than .bbl
;; We don't care whether the bib files are open in emacs
(TeX-check-files (concat name ".bbl")
(mapcar 'car
(LaTeX-bibliography-list))
(append BibTeX-file-extensions
TeX-Biber-file-extensions)))
;; We should check for bst files here as well.
(if LaTeX-using-Biber TeX-command-Biber TeX-command-BibTeX))
((TeX-process-get-variable name
'TeX-command-next
;; HERE COMES THE PATCH
;; was TeX-command-View
TeX-command-default))
;; END OF PATCH
(TeX-command-Show)))
(completion-ignore-case t)
(answer (or TeX-command-force
(completing-read
(concat "Command: (default " default ") ")
(TeX-mode-specific-command-list major-mode) nil t
nil 'TeX-command-history))))
;; If the answer is "latex" it will not be expanded to "LaTeX"
(setq answer (car-safe (TeX-assoc answer TeX-command-list)))
(if (and answer
(not (string-equal answer "")))
answer
default)))

emacs shell: type once, run everywhere

Within emacs, I want to have multiple shells open, type a command once, and have it run in each shell -- similar to the way multixterm ( http://freecode.com/projects/multixterm ) does.
With some minimal testing, this will do:
(defun send-to-all-shells ()
(interactive)
(let ((command (read-from-minibuffer "Command: ")))
(mapcar #'(lambda (x) (comint-send-string x (concat "\n" command "\n")))
(remove-if-not
#'(lambda (x)
(string= "/bin/bash"
(car (process-command x))))
(process-list)))))
To run, just M-x send-to-all-shells, enter the command you want, and it will be sent to all open shells. This assumes your shell is found in /bin/bash. If not, change that bit accordingly.
If you do this a lot, you'll want to bind that to your favourite key combo. It would be possible to borrow and modify the code in comint-send-input such that you could just enter the command you want at the prompt of one shell, hit your key and have that command send to all shells simultaneously. I'm short on time, so I'll leave that as an exercise for the reader.

Emacs custom command line argument

From the documentation I can see I can access command line arguments (command-line-args).
I'd like to add my own arguments but Emacs complains at start up that it doesn't recognize them.
E.g.
emacs -my_argument
I get:
command-line-1: Unknown option `-my_argument'
What's a proper way to define my custom arguments and provide information to my Emacs session?
Is there a way to pop an argument from a command line?
Add something like this to your ~/.emacs, ~/.emacs.el, or ~/.emacs.d/init.el file:
(defun my-argument-fn (switch)
(message "i was passed -my_argument"))
(add-to-list 'command-switch-alist '("-my_argument" . my-argument-fn))
Then you can execute emacs -my_argument and it should print i was passed -my_argument to the minibuffer. You can find more information in the GNU elisp reference.
As stated in another post you can add your custom switches to command-switch-alist and emacs will call the handler function for any matching switch passed in on the command line. However, this operation is done after your .emacs file has been evaluated. This is fine for most cases but you may wish for a command line argument to alter the execution path or behaviour of your .emacs evaluation; I often do this to enable/disable configuration chunks (mainly for debugging).
To achieve this you can read command-line-args and check for your switch manually and then delete it from the list, this will stop emacs complaining about an unknown argument.
(setq my-switch-found (member "-myswitch" command-line-args))
(setq command-line-args (delete "-myswitch" command-line-args))
Which can alter your .emacs evaluation like so:
(unless my-switch-found
(message "Didn't find inhibit switch, loading some config.")
...)
And you could build this into a single step:
;; This was written in SO text-box, not been tested.
(defun found-custom-arg (switch)
(let ((found-switch (member switch command-line-args)))
(setq command-line-args (delete switch command-line-args))
found-switch))
(unless (found-custom-arg "-myswitch")
(message "Loading config...")
...)
For those who are interested, here is a code snip to show how to process custom arguments in Emacs lisp. In this case, I am processing an argument --suffix / -S to variable _suffix.
I pulled the idea from a BSD-Lite Script Emacs script.
(setq _suffix nil)
;; Process cli args
(while command-line-args-left
(setq k (car command-line-args-left))
(setq command-line-args-left (cdr command-line-args-left))
(setq command-line-args (delete k command-line-args))
(cond
(or (string-equal k "--cs-suffix")
(string-equal k "-S"))
(setq _suffix (intern (car command-line-args-left)))
(setq command-line-args-left (cdr command-line-args-left))
(setq command-line-args (delete _suffix command-line-args))
)))
This will roll through command-line-args-left and remove them all from command-line-args which will prevent Emacs from complaining.

How do I change read/write mode for a file using Emacs?

If a file is set to read only mode, how do I change it to write mode and vice versa from within Emacs?
M-x read-only-mode
in very old versions of Emacs, the command was:
M-x toggle-read-only
On my Windows box, that amounts to Alt-x to bring up the meta prompt and typing "read-only-mode" to call the correct elisp function.
If you are using the default keyboard bindings,
C-x C-q
(which you read aloud as "Control-X Control-Q") will have the same effect. Remember, however, given that emacs is essentially infinitely re-configurable, your mileage may vary.
Following up from the commentary: you should note that the writeable status of the buffer does not change the writeable permission of the file. If you try to write out to a read only file, you'll see a confirmation message. However, if you own the file, you can write out your changes without changing the permissions on the file.
This is very convenient if you'd like to make a quick change to a file without having to go through the multiple steps of add write permission, write out changes, remove write permission. I tend to forget that last step, leaving potentially critical files open for accidental changes later on.
Be sure you're not confusing 'file' with 'buffer'. You can set buffers to read-only and back again with C-x C-q (toggle-read-only). If you have permission to read, but not write, a file, the buffer you get when you visit the file (C-x C-f or find-file) will be put in read-only mode automatically. If you want to change the permissions on a file in the file system, perhaps start with dired on the directory that contains the file. Documentation for dired can be found in info; C-h i (emacs)dired RET.
What I found is M-x set-file-modes filename mode
It worked at my Windows Vista box.
For example: M-x set-file-modes <RET> ReadOnlyFile.txt <RET> 0666
As mentioned up there by somebody else: M-x toggle-read-only would work.
However, this is now deprecated and M-x read-only-mode is the current way to do it, that it is set to C-x C-q keybinding.
CTRL + X + CTRL + Q
If only the buffer (and not the file) is read-only, you can use toggle-read-only, which is usually bound to C-x C-q.
If the file itself is read-only, however, you may find the following function useful:
(defun set-buffer-file-writable ()
"Make the file shown in the current buffer writable.
Make the buffer writable as well."
(interactive)
(unix-output "chmod" "+w" (buffer-file-name))
(toggle-read-only nil)
(message (trim-right '(?\n) (unix-output "ls" "-l" (buffer-file-name)))))
The function depends on unix-output and trim-right:
(defun unix-output (command &rest args)
"Run a unix command and, if it returns 0, return the output as a string.
Otherwise, signal an error. The error message is the first line of the output."
(let ((output-buffer (generate-new-buffer "*stdout*")))
(unwind-protect
(let ((return-value (apply 'call-process command nil
output-buffer nil args)))
(set-buffer output-buffer)
(save-excursion
(unless (= return-value 0)
(goto-char (point-min))
(end-of-line)
(if (= (point-min) (point))
(error "Command failed: %s%s" command
(with-output-to-string
(dolist (arg args)
(princ " ")
(princ arg))))
(error "%s" (buffer-substring-no-properties (point-min)
(point)))))
(buffer-substring-no-properties (point-min) (point-max))))
(kill-buffer output-buffer))))
(defun trim-right (bag string &optional start end)
(setq bag (if (eq bag t) '(?\ ?\n ?\t ?\v ?\r ?\f) bag)
start (or start 0)
end (or end (length string)))
(while (and (> end 0)
(member (aref string (1- end)) bag))
(decf end))
(substring string start end))
Place the functions in your ~/.emacs.el, evaluate them (or restart emacs). You can then make the file in the current buffer writable with M-x set-buffer-file-writable.
If you are looking at a directory of files (dired), then you can use Shift + M on a filename and enter the modespec, the same attributes used in the chmod command.
M modespec <RET>
See the other useful commands on files in a directory listing at
http://www.gnu.org/s/libtool/manual/emacs/Operating-on-Files.html
I tried out Vebjorn Ljosa's solution, and it turned out that at least in my Emacs (22.3.1) there isn't such function as 'trim-right', which is used for removing an useless newline at the end of chmod output.
Removing the call to 'trim-right' helped, but made the status row "bounce" because of the extra newline.
C-x C-q is useless. Because your also need the permission to save a file.
I use Spacemacs. It gives me a convenient function to solve this question. The code is following.
(defun spacemacs/sudo-edit (&optional arg)
(interactive "p")
(if (or arg (not buffer-file-name))
(find-file (concat "/sudo:root#localhost:" (ido-read-file-name "File: ")))
(find-alternate-file (concat "/sudo:root#localhost:" buffer-file-name))))
I call spacemacs/sudo-edit to open a file in emacs and input my password, I can change the file without read-only mode.
You can write a new function like spacemacs/sudo-edit.