Lacking Natural Simplicity

Random musings on books, code, and tabletop games.

Arrow Macros in Emacs Lisp

Are you familiar with arrow macros? (They have nothing to do with arrows in Haskell.) They started out in Clojure, I think, and have been ported to Common Lisp and probably other Lisps, though I didn't find one for Emacs Lisp. I think the arrows package is the most up to date in Quicklisp for Common Lisp, but I found the cl-arrows package first so that's the one I converted to Emacs Lisp today.

There were two slightly tricky parts to the port

  • I needed to use lexical binding for the source file

  • cl-arrows's test suite used fractions, but Emacs Lisp doesn't have them.

Anyway, I think they're very useful. If you have a value and you want it to be processed by several functions in a chain you can use an arrow macro instead of using lots of nested function calls. For instance,

(replace-regexp-in-string "-+$" "" (replace-regexp-in-string "^-+" "" (replace-regexp-in-string "^-+" "" (replace-regexp-in-string "\\.+" "." (replace-regexp-in-string "\\.-" "." (replace-regexp-in-string "-\\." "." (replace-regexp-in-string "-+" "-" (replace-regexp-in-string "[^-.a-z0-9]+" "-" string))))))))

is annoying and hard to read on one line. If you split it into multiple lines

(replace-regexp-in-string
 "-+$" ""
 (replace-regexp-in-string
  "^-+" ""
  (replace-regexp-in-string
   "^-+" ""
   (replace-regexp-in-string
    "\\.+" "."
    (replace-regexp-in-string
     "\\.-" "."
     (replace-regexp-in-string
      "-\\." "."
      (replace-regexp-in-string
       "-+" "-"
       (replace-regexp-in-string "[^-.a-z0-9]+" "-" string))))))))

that's also pretty horrible. To read it you have to skip to the end, figure out that function call, then skip back to the next function call, and so forth, until you get back to the last function call, reading it from inside out. It looks much better if you use the ->> arrow macro, which substitutes the value you are passing along as the last argument to the functions you are calling:

(->> string
     (replace-regexp-in-string "[^-.a-z0-9]+" "-")
     (replace-regexp-in-string "-+" "-")
     (replace-regexp-in-string "-\\." ".")
     (replace-regexp-in-string "\\.-" ".")
     (replace-regexp-in-string "\\.+" ".")
     (replace-regexp-in-string "^-+" "")
     (replace-regexp-in-string "-+$" ""))

I also think it is more understandable than the highly nested calls, since you can follow along in from start to finish, rather than from inside out: you've got the value, you pass it to the first function, then to the second, and so forth.

There are also macros to insert the value at the front of the argument list (->) and to insert the value where a <> symbol is found (-<>).

I wrote some Emacs Lisp code with the last expression above today, and I think it makes a good example of why they are useful.

Here's the use case that inspired me to port them to Emacs Lisp. When I save things from the internet I often want to create a directory to save them in based on the filename or some text on a web page or something, but usually there are spaces or special characters in that text, so I have to sanitize the text to make it a safe and simple filename. So I need to replace all special characters with dashes. But I don't want dashes repeated multiple times, just a single dash, and I don't want dashes followed or preceded by periods, only the period.

Here' s some code to do that:

(defun tkb-sanitize-for-filename (string)
  "Clean up characters in STRING that aren't good for
filenames."
  (->> string
       (replace-regexp-in-string "[^-.a-z0-9]+" "-")
       (replace-regexp-in-string "-+" "-")
       (replace-regexp-in-string "-\\." ".")
       (replace-regexp-in-string "\\.-" ".")
       (replace-regexp-in-string "\\.+" ".")
       (replace-regexp-in-string "^-+" "")
       (replace-regexp-in-string "-+$" "")))

(defun tkb-sanitize-kill-for-filename (string)
  "Clean up characters that aren't good for filenames in the
top of the kill
ring and put the result on the top of the kill ring."
  ;; gets the kill or the system clipboard if it is new.
  (interactive (list (current-kill 0 t)))
  (let ((new-string (tkb-sanitize-for-filename string)))
    (message "Old string: %s\nNew String: %s" string new-string)
    (kill-new new-string)))

(global-set-key (kbd "C-c k S") 'tkb-sanitize-kill-for-filename)

Note that the function tkb-sanitize-kill-for-filename that I use interactively gets uses (current-kill 0 t) because it either returns the top of the kill ring, or if the system clipboard is newer it puts that on top of the kill ring and returns it. (The docstring of current-kill makes that a little difficult to figure out.)

So if I have the name of a product as "***Flashy####Product%%%%Name!!!!" I copy that to the clipboard with ⌘-C, switch to emacs, do C-c k S, switch back to whatever program is saving the file, and use the generated, sanitized name Flashy-Product-Name with ⌘-V. Nifty!

(And if you've ever wondered what the Unicode character for the Mac keyboard Command key (⌘) is, it is the PLACE OF INTEREST SIGN, and you can enter it by C-x 8 RETURN PLACE OF INTEREST SIGN. The symbol they use for for the key labeled alt and option (⌥) is the Unicode OPTION KEY, and the symbol they use for the Shift Key (⇧) is UPWARDS WHITE ARROW. I'm surprised they don't have a special symbol for Control, but it looks like they just use the plain ASCII ^, known in Unicode as CIRCUMFLEX ACCENT.)

Anyway, here's a ZIP file with the Emacs Lisp source for the arrow macros so you can use them if you've been convinced of their usefulness by this post. It also has some minimal documentation in the README.md that came from the README.md from cl-arrows, which seems to repeat the information from the docstrings for the macros.

Print Friendly and PDF

Comments

Comments powered by Disqus