Is there an emacs extension that takes periodic snapshots (once a minute, once every x keystrokes, whatever) of a file while it is being edited similar to the change history in Eclipse or the edit history in Google Docs and other programs?
I'm hoping for something that'll let me easy navigate through the changes I've made from day to day - Is there anything like this already written?
Edit
I should be more specific - I'm not looking for a VCS. I'm looking for a minor-mode or something similar I can just switch on and have hard copies of the revisions on disk.
There's a built-in feature called autosave that saves after N keystrokes (and maybe after M seconds, I'm not sure). I generally use this if Emacs crashes, not for looking at what edits I've made; undo is better for that. Here's my config:
(setq autosave-dir (concat user-emacs-directory "autosaves/")
auto-save-list-file-prefix (concat emacs-persistence-directory
"autosave-list"))
(if (not (file-exists-p autosave-dir))
(make-directory autosave-dir t))
(add-to-list 'auto-save-file-name-transforms
`("\\`/?\\([^/]*/\\)*\\([^/]*\\)\\'" ,(concat autosave-dir "\\2") t))
;; tramp autosaves
(setq tramp-auto-save-directory (concat user-emacs-directory "tramp-autosaves/"))
(if (not (file-exists-p tramp-auto-save-directory))
(make-directory tramp-auto-save-directory))
There's also a backup system that creates a copy after every save (not autosave). I use this for what I think you're asking for - looking at history since my last VCS commit. Here's my config:
(setq make-backup-files t
vc-make-backup-files t
version-control t
kept-new-versions 256
kept-old-versions 0
delete-old-versions t
backup-by-copying t)
(setq backup-dir (concat user-emacs-directory "backup/"))
(if (not (file-exists-p backup-dir))
(make-directory backup-dir))
(add-to-list 'backup-directory-alist
`(".*" . ,backup-dir))
(defun force-backup-of-buffer ()
(setq buffer-backed-up nil))
(add-hook 'before-save-hook 'force-backup-of-buffer)
;; this is what tramp uses
(setq tramp-backup-directory-alist backup-directory-alist)
(add-to-path "backup-walker")
(autoload 'backup-walker-start "backup-walker"
"start walking with the latest backup" t)
I use the excellent backup-walker to navigate through the backups.
The best solution I have found for this is undo-tree, which is the Emacs equivalent to gundo for vim: it lets you visualise the tree of undo/redo, navigate through the changes and go back and forth between different versions.
undo-tree can be installed using ELPA; once it is installed, add the following to .emacs:
(require 'undo-tree)
(global-undo-tree-mode)
The undo tree can then be visualised using Ctrl-x u (undo-tree-visualize). The different “versions” can be navigated in a very intuitive manner, using arrow keys.
The history can also be made persistent using undo-tree-save-history.
If you really value keeping your changes, I would highly recommend starting to use git. It will work for your windows and linux coding environments. I use it to port changes from code back and forth. It does a great job of fixing all the little line endings and merging changes.
If you really just want it to keep old versions, then emacs can create a new backup every time that you save. It just creates another file right next to your current one. That way you can control how often it makes a new backup (every time you save).
Here's a good page that talks about the options:
ftp://ftp.gnu.org/pub/old-gnu/Manuals/emacs-20.7/html_chapter/emacs_18.html#SEC109
Try http://www.emacswiki.org/emacs/BackupEachSave
It has saved me a number of times allowing me to backtrack.
Related
I am involved in python project where tabs are used, however i am not using them in every other code i write, it is vital to use them in that particular project. Projects are located in one directory under specific directories. I.E:
\main_folder
\project1
\project2
\project3
...etc
I have couple functions/hooks on file open and save that untabify and tabify whole buffer i work on.
;; My Functions
(defun untabify-buffer ()
"Untabify current buffer"
(interactive)
(untabify (point-min) (point-max)))
(defun tabify-buffer ()
"Tabify current buffer"
(interactive)
(tabify (point-min) (point-max)))
;; HOOKS
; untabify buffer on open
(add-hook 'find-file-hook 'untabify-buffer)
; tabify on save
(add-hook 'before-save-hook 'tabify-buffer)
If i put it in .emacs file it is run on every .py file i open which is not what i want. What i`d like to have is to have these hooks used only in one particular folder with respective subfolders. Tried .dir_locals but it works only for properties not hooks. I can not use hooks in specific modes (i.e. python-mode) as almost all projects are written in python. To be honest i tried writing elisp conditional save but failed.
A very easy solution is to just add a configuration variable that can be used to disable the hooks. For example:
(defvar tweak-tabs t)
(add-hook 'find-file-hook
(lambda () (when tweak-tabs (untabify (point-min) (point-max)))))
(add-hook 'before-save-hook
(lambda () (when tweak-tabs (tabify (point-min) (point-max)))))
Now you can add a .dir-locals.el file in the relevant directories, setting tweak-tabs to nil, disabling this feature there.
(But another problem is that this is a pretty bad way to deal with tabs. For example, after you save a file you do see the tabs in it.)
Just for the record, to answer the literal question in the title (as I reached this question via a web search): one way to add a hook that depends on file location is to make it a function that checks for buffer-file-name. (Idea from this answer.)
For example, for the exact same problem (turn on tabs only in a particular directory, and leave tabs turned off elsewhere), I'm currently doing something like (after having installed the package smart-tabs-mode with M-x package-install):
(smart-tabs-insinuate 'python) ; This screws up all Python files (inserts tabs)
(add-hook 'python-mode-hook ; So we need to un-screw most of them
(lambda ()
(unless (and (stringp buffer-file-name)
(string-match "specialproject" buffer-file-name))
(setq smart-tabs-mode nil)))
t) ; Add this hook to end of the list
(This is a bit inverted, as smart-tabs-insinuate itself modifies python-mode-hook and then we're modifying it back, but it should do as an example.)
I am upgrading to emacs23. I find that my emacs.el loads much more slowly.
It's my own fault really... I have a lot of stuff in there.
So I am also trying to autoload everything possible that is currently "required" by my emacs.el.
I have a module that exposes 12 entry points - interactive functions I can call.
Is the correct approach to have 12 calls to autoload in order to insure that the module is loaded regardless of which function I call? Are there any problems with this approach? Will it present performance issues?
If not that approach, then what?
What you really want is to get the autoloads generated for you automatically, so that your .emacs file remains pristine. Most packages have the ;;;###autoload lines in them already, and if not, you can easily add them.
To manage this, you can put all the packages in a directory, say ~/emacs/lisp, and in there have a file named update-auto-loads.el which contains:
;; put this path into the load-path automatically
;;;###autoload
(progn
(setq load-path (cons (file-name-directory load-file-name) load-path)))
;;;###autoload
(defun update-autoloads-in-package-area (&optional file)
"Update autoloads for files in the diretory containing this file."
(interactive)
(let ((base (file-truename
(file-name-directory
(symbol-file 'update-autoloads-in-package-area 'defun)))))
(require 'autoload) ;ironic, i know
(let ((generated-autoload-file (concat base "loaddefs.el")))
(when (not (file-exists-p generated-autoload-file))
(with-current-buffer (find-file-noselect generated-autoload-file)
(insert ";;") ;; create the file with non-zero size to appease autoload
(save-buffer)))
(cd base)
(if file
(update-file-autoloads file)
(update-autoloads-from-directories base)))))
;;;###autoload
(defun update-autoloads-for-file-in-package-area (file)
(interactive "f")
(update-autoloads-in-package-area file))
If you add 'update-autoloads-in-package-area to your kill-emacs-hook, then the loaddefs.el will automatically be updated every time you exit Emacs.
And, to tie it all together, add this to your .emacs:
(load-file "~/emacs/lisp/loaddefs.el")
Now, when you download a new package, just save it in the ~/emacs/lisp directory, update the loaddefs via M-x update-autoloads-in-package-area (or exit emacs), and it'll be available the next time you run Emacs. No more changes to your .emacs to load things.
See this question for other alternatives to speeding up Emacs startup: How can I make Emacs start-up faster?
Well, who cares how slowly it starts?
Fire it up via emacs --daemon & and then connect using either one of
emacsclient -c /some/file.ext, or
emacsclient -nw
I created aliases for both these as emx and emt, respectively. Continuing once editing session is so much saner...
Ideally you shouldn't have any load or require in your .emacs file.
You should be using autoload instead...
e.g.
(autoload 'slime-selector "slime" t)
You will need to use eval-after-load to do any library specific config, but the upshot is that you won't need to wait for all this to load up front, or cause errors on versions of Emacs that don't have the same functionality. (e.g. Terminal based, or a different platform etc.)
While this may not affect you right now, chances are, in future you will want to use the same config on all machines / environments where you use Emacs, so it's a very good thing to have your config ready to fly.
Also use (start-server) and open external files into Emacs using emacsclient - So you avoid restarting Emacs.
I would like to make a gradual switch from GNU Emacs to Xemacs. Are there tricks I can use to have the two play well?
Currently, I see the following issues:
xemacs alters .emacs
The two do not like each others .elc files.
Thanks!
Interesting, most appear to be moving in the other direction as I believe XEmacs to be fairly dormant (based on activity of the xemacs-announce list). Simple packages can co-exist, but many folks have given up making their packages work in both XEmacs and Emacs.
But, in answer to your question, to get your .emacs to work in both, I'd start writing a some routines to do function translation between the two. For example, at one point I needed this to get my .emacs to work in XEmacs:
(if (not (fboundp 'tags-table-files))
(defun tags-table-files ()
(tag-table-files tags-file-name)))
Other things were triggered on the Emacs variant, which I stored in a variable GNU:
(setq GNU (not (string-match "XEmacs\\|Lucid" (emacs-version))))
(if GNU
(do-emacs-thing)
(do-xemacs-thing))
I was keeping compiled .emacs files and did this:
(setq compiled-dot-emacs-name (format ".emacs-%d%s" emacs-major-version
(if GNU "" "X")))
Regarding compiled packages, I'd probably store all the .el files in one directory (say emacs-lisp), but have an xemacs variant (xemacs-lisp) with symlinks to the .el files. And then you just byte compile each directory from the appropriate Emacs variant, and make sure to have your load-path point to the right one.
The Emacs wiki has a page on Emacs versus XEmacs which might be a good starting point to figure out other tips to make them cohabitate. Specifically, there's a page for customizing both.
I don't use xemacs, but newer GNU emacs will check for .emacs.d/init.el as well, so maybe moving .emacs stuff to init.el makes sense. Additionally you can link it to .xemacs/init.el if you manage to keep your customization applicable for both.
There is also a discussion on what emacs to prefer on emacswiki.
I started a slow move from Xemacs to Emacs a while ago. I now use both on a daily basis. To make the transition smoother (one set of init files), I stole the following .emacs file from http://xemacs.seanm.ca/_emacs (but the link is now dead).
(setq user-init-file
(expand-file-name "init.el"
(expand-file-name ".xemacs" "~")))
(setq custom-file
(expand-file-name "custom.el"
(expand-file-name ".xemacs" "~")))
(if (file-exists-p user-init-file)
(load-file user-init-file))
(if (file-exists-p custom-file)
(load-file custom-file))
My ~/.xemacs/init.el starts off with:
(unless (boundp 'running-xemacs)
(defvar running-xemacs nil))
(setq load-path (cons "~/.elisp" load-path)) ; packages for both emacsen
(if running-xemacs
(setq load-path (cons "~/.elisp/xemacs" load-path)) ; packages for Xemacs only
(setq load-path (cons "~/.elisp/gnuemacs" load-path))) ; packages for Gnuemacs only
From then on it is pretty obvious what I have (the occasional (if running-xemacs) ...). I also deleted all the .elc files from ~/.elisp, but I presume that Trey Jackson's suggestion will work.
on modern systems i do not see the need for precompiled elisp files anymore. The benefit in looking and on the fly changing the .el-files is much more higher.
.emacs: put your own defines in .emacs_startup (or which name you prefere) and put all your gnu-enmacs stuff there and put a conditional load in your .emacs
When I call find-file to open a new file, it generally happens that the file I'm looking for is in one of the directories I've already loaded from.
Ideally, I'd like to scroll through the history using the up/down arrows.
The problem with this is that if I've already loaded 10 files from a directory, I first have to pass through those ten files, all in the same directory, before I see a new directory where my file might be.
In the end, I often just type in the directory again or cut/paste it from an xterm.
in the find-file command, can I change the behavior of the up/down arrows to iterate over directories instead of files.
Alternatively, can I change the file order to match the order of buffers most recently visited instead of when I loaded the file?
My first answer suffered from the behavior that TAB completion no longer worked as expected in 'find-file. But the technique still seems useful (and if you like it, preferable).
However, this solution has the same history behavior, but maintains the TAB completion as expected inside 'find-file.
I'd be interested to know if there were a way to avoid advising find-file, but I couldn't find any introspection that gave me the knowledge that 'find-file was called.
(defadvice find-file (around find-file-set-trigger-variable protect activate)
"bind a variable so that history command can do special behavior for find-file"
(interactive (let (inside-find-file-command) (find-file-read-args "Find file: " nil)))
ad-do-it)
(defadvice next-history-element (around next-history-element-special-behavior-for-find-file protect activate)
"when doing history for find-file, use the buffer-list as history"
(if (boundp 'inside-find-file-command)
(let ((find-file-history (delq nil (mapcar 'buffer-file-name (buffer-list))))
(minibuffer-history-variable 'find-file-history))
ad-do-it)
ad-do-it))
I suggest IDO. You can search in the buffer list or in find-file. When searching in find-file and it has no matches in the current folder it searches through history.
not what you want, but
have you tried (electric-buffer-list) Ctrl-x Ctrl-b?
or (dired)?
This will use the buffer-list's order for the history, which is what you want.
(setq read-file-name-function 'my-read-file-name)
(defun my-read-file-name (prompt dir default-filename mustmatch initial predicate)
(let ((default-directory dir)
(files (directory-files dir))
(history (delq nil (mapcar 'buffer-file-name (buffer-list)))))
(completing-read prompt files predicate mustmatch initial 'history)))
Hmmm... This changes the behavior of find-file's TAB completion because the TAB completes over the history (already opened files), and not over the files available in the directory you're specifying.
Getting both to work at the same time is a bit trickier...
With Icicles you can cycle among candidate file names, and you can sort them in many ways. You can access them in order of last use etc.
http://www.emacswiki.org/emacs/Icicles_-_History_Enhancements
http://www.emacswiki.org/emacs/Icicles_-_Sorting_Candidates
http://www.emacswiki.org/emacs/Icicles_-_File-Name_Input
I know about M-x dire, but would like to customize it. I would like to hit one key (for example F2) and get dire buffer open. When I navigate across the directory hierarchy it shouldn't open new buffers.
And when I finally open the file it also shouldn't open new buffer for it (not strictly necessary, but strongly preferred).
Of course this behavior can be global, i.e. for all dire buffers/invocations.
Check out dired-single, which pretty much does what you want (except that last bit, where it reuses the dired buffer for the newly visted file).
Caveat Lector: I wrote it, so I'm biased towards its usefulness.
Some alternatives - EmacsWiki: DiredReuseDirectoryBuffer, and this short snippet from an awkwardly-formatted blog-entry.
caveat: haven't tried them, myself.
I know this is very old but All you have to do is press 'a' on a dir or file to get this functionality. It's already there.
Here's what I finally used:
(require 'dired)
(global-set-key [(f2)] 'my-dired)
(defun my-dired ()
(interactive)
(dired (file-name-directory (buffer-file-name))))
(defadvice dired-advertised-find-file (around dired-subst-directory activate)
"Replace current buffer if file is a directory."
(interactive)
(let ((orig (current-buffer)) (filename (dired-get-filename :no-error-if-not-filep t)))
ad-do-it
(when (not (eq (current-buffer) orig)) (kill-buffer orig))))