This package adds the command dired-rsync which allows the user to
copy marked files in a dired buffer via rsync. This is useful,
especially for large files, because the copy happens in the background
and doesn’t lock up Emacs. It is also more efficient than using tramp’s
own encoding methods for moving data between systems.
Configuration is simple as you only need to bind the dired-rsync
command to your preferred dired binding.
(use-package dired-rsync
:bind (:map dired-mode-map
("C-c C-r" . dired-rsync))A second package in this repository provides a transient version
called dired-rsync-transient. This wraps the command in a `magit`
like transient interface allowing you to tweaks the parameters for
your call.
(use-package dired-rsync-transient
:bind (:map dired-mode-map
("C-c C-x" . dired-rsync-transient)))A variable called dired-rsync-modeline-status is provided for mode
lines that will report the number of active rsync operations in
progress. The variable dired-rsync-job-count contains the number of
currently active dired-rsync jobs currently running.
For those that don’t like the way dired-rsync unmarks transferred
files when it completes please see dired-rsync-unmark-on-completion.
If you want to tweak the way marks are collected you can customise
dired-rsync-source-files and provide your own function to return a
list of fully qualified names to include in the command.
From time to time the call to rsync may fail. dired-rsync keeps the
process buffer around for debugging and reports to the console. You
can customise dired-rsync-failed-hook with your own hook function or
select the more aggressive dired-rsync--pop-to-rsync-failed-buf to
pop straight to the buffer. Also you can customise dired-rsync-success-hook
with your own hook function to provide notification for compleated transfers.
The options dired-rsync-command and dired-rsync-options are there
to modify the call to rsync but a user is unlikely to need to tweak
these.
While you can use rsync to copy files locally the main use case is copying files to/from a remote system. The rsync tool is always run locally as rsync needs working SSH authentication to work. If you can access a remote machine via tramp/ssh without having to enter a password (because ssh-agent is working) then rsync should work fine.
You can also do a remote-to-remote copy although this does involve setting up a port forward so the first remote can access the second remote. Also as dired-rsync needs to reverse ssh from your source location it is best if you fully qualify the location in tramp (i.e. user@remote:/dest/path) so it can extract the correct username to login with. It’s ugly but it works.
The test used by dired-rsync is tramp-tramp-file-p which will mean
any tramp path will attempt to be converted to an rsync path with
optional ssh transport. Obviously if your remote target doesn’t also
have ssh running and remote rsync binary this will fail. The reporting
could be improved.
There have been several attempts at doing this but I found them wanting in usability. This attempts to clean up the ideas from:
- https://truongtx.me/tmtxt-dired-async.html
- https://oremacs.com/2016/02/24/dired-rsync/
- https://vxlabs.com/2018/03/30/asynchronous-rsync-with-emacs-dired-and-tramp/
in a hopefully cleaner and more idiomatic way.