Emacs Parallel is yet another library to simulate parallel computations in Emacs (because it lacks threads support in Elisp).
You can execute a simple function a retrive the result like this:
(parallel-get-result (parallel-start (lambda () (* 42 42))))
⇒ 1764
Though you won't benefit from the parallelism because
parallel-get-result is blocking, that is it waits for the function
to be executed.
So you can use define a callback to be called when the function is finished:
(parallel-start (lambda () (sleep-for 4.2) "Hello World")
:post-exec (lambda (results _status)
(message (first results))))
⊣ Hello World
Here, why (first results) and not result? Because you can send
data from the remote instance while it's running with
parallel-remote-send:
(parallel-start (lambda ()
(parallel-remote-send "Hello")
(sleep-for 4.2)
"World")
:post-exec (lambda (results _status)
(message "%s"
(mapconcat #'identity (reverse results) " "))))
⊣ Hello World
As you may have noticed the results are pushed in a list, so the first element is the result returned by the function called, the second is the last piece of data send, and so on…
And of course you can execute some code when you receive data from the remote instance:
(parallel-start (lambda ()
(parallel-remote-send 42)
(sleep-for 4.2) ; heavy computation to compute PI
pi)
:on-event (lambda (data)
(message "Received %S" data)))
⊣ Received 42
⊣ Received 3.141592653589793
Because the function is executed in another Emacs instance (in Batch
Mode by default), the environment isn't the same. However you can
send some data with the env parameter:
(let ((a 42)
(b 12))
(parallel-get-result (parallel-start (lambda (a b) (+ a b))
:env (list a b))))
⇒ 54
By default, the remote Emacs instance is exited when the function is
executed, but you can keep it running with the
:continue-when-executed option and send new code to be executed
with parellel-send.
(let ((task (parallel-start (lambda () 42)
:continue-when-executed t)))
(sleep-for 4.2)
(parallel-send task (lambda () (setq parallel-continue-when-executed nil) 12))
(parallel-get-results task))
⇒ (12 42)
As you can see, to stop the remote instance you have to set the
variable parallel-continue-when-executed to nil.
Emacs XWidget is an experimental branch which permits to embed GTK+ widget inside Emacs buffers. For instance, it is possible to use it to render an HTML page using the webkit engine within an Emacs buffer.
With this module, you can configure your "main" Emacs to use another one to render web pages.
Let's assume that you've cloned the Emacs XWidget repository in
$HOME/src/emacs-xwidget/. Once you've compiled it, an Emacs
executable is available $HOME/src/emacs-xwidget/src/emacs.
Configure parallel-xwidget to use it:
(setq parallel-xwidget-config (list :emacs-path
(concat (getenv "HOME")
"/src/emacs-xwidget/src/emacs")))
Then configure your current Emacs to use it:
(setq browse-url-browser-function 'parallel-xwidget-browse-url)
You can check it out with M-x browse-url RET google.com RET.
If your windows manager is smart enough (like StumpwWM) you can use it to move graphical windows (Emacs frames) in another desktop.
For example, I use this to move Emacs frames (with the title "emacs-debug") to the group (aka desktop) 9:
(define-frame-preference "9"
(0 nil t :title "emacs-debug"))
And this to specify the title of the frame:
(parallel-start (lambda () 42)
:no-batch t
:emacs-args '("-T" "emacs-debug"))
You can only send data to the remote (with the env parameter) or
from the remote (with parallel-send and parallel-remote-send)
that have a printed representation (see <elisp#Printed>
Representation).
So you can pass around numbers, symbols, strings, lists, vectors, hash-table but you can't pass buffers, windows, frames…
It lacks documentation, tests and probably a clean API, but I'm working on it!