Essential Elisp Libraries31 March 2013
Recently, I’ve been writing a lot of Emacs Lisp. I’ve been trying to write as many of my own tools as possible in elisp, to see what the language works well at and what it doesn’t do well.
Turns out, elisp does many things well. However, there are several libraries that make life much easier.
(Disclaimer: I’ve written some of them.)
Firstly, dash.el, which describes itself as “a modern list library for Emacs”. Dash.el offers
-group-by, and many other list operations that are very common in other languages.
Dash.el has modernised elisp. To my knowledge, stock Emacs doesn’t even include a traditional
(map 'some-function some-list) operation – there’s only
cl-map, but it requires another argument and using
'cl at runtime is frowned upon.
Edit: Several people have pointed out that stock Emacs has
mapcar, which does exactly this.
The documentation is excellent (it’s generated automatically from the test suite), but my favourite feature is the anaphoric macros. Rather than writing:
(-filter (lambda (x) (> x 10)) (list 8 9 10 11 12))
You can simply write an s-expression that is evaluated for each item:
(--filter (> it 10) (list 8 9 10 11 12))
s.el, “the long lost Emacs string manipulation library”, is another library that is too valuable to miss. Emacs functions often revolve around text editing (surprise!) and often need a versatile set of string manipulation functions.
As with dash.el, s.el introduces a more contemporary style to elisp. All s.el functions are pure, returning a new string, despite elisp supporting mutable strings. Some common string operations include
s-join. It’s not that these operations are impossible in elisp – you can replace literal strings using
regexp-quote, but s.el makes this much easier.
Hash tables are a rarely used datastructure in elisp. It’s more common to see lists of pairs used, despite their slower performance. This is mostly because lists are easier to work with: creating a list is trivial, but
make-hash-table has an intimidating docstring with five different keyword arguments. Common operations (can you see a theme here?) for hash tables, such as iterating over keys, are not as easy as they should be.
ht.el solves this.
ht-create has only one argument, and it can be omitted. ht.el also provides common operations like
ht-values as well as convenience functions for converting from and to alists or plists.
With a good list package like dash.el, I find that an explicit loop is rarely necessary. However, sometimes you’re iterating for side effects and you need something more powerful. Elisp itself only includes
dolist and dash.el has
-each, but they’re all low-level.
loop.el “friendly imperative loop structures” offers a larger set of loop structures, including:
loop-for-each. More interestingly, you can call
loop-continue in any loop.
Now that Marmalade and MELPA are maturing, it’s easy to depend on these libraries. Previously, elisp package authors would try to make their code self-contained. This is no longer necessary, you can simply put the following in the header of your package:
;; Package-Requires: ((dash "1.1.0") (loop "1.1"))
(Note that the version specifier is simply the minimum version.)
Since these packages are available on both Marmalade and MELPA, people who install your package will get these utility libraries automatically, if they don’t have them already.
All these utility libraries demonstrate the remarkable flexibility of elisp. You can take familiar ideas from other, newer languages and use them in elisp. I have deliberately not described any of the functions I’ve mentioned here (they have good docstrings anyway) since their names will be familiar to programmers with mainstream programming language experience.
These libraries also help with a classic problem in the Emacs world: discoverability. Their purposes are well-defined, and each library mentioned has a README that lists all the functionality available to the elisp programmer.
I don’t believe this is the last word in utility libraries. There are still several areas of elisp that would benefit from a new API. Regular expressions in elisp modify global state rather than returning multiple values (see for example
save-match-data). Platform-independent path manipulation is awkward (use
concat for joining paths), and I’m not aware of any elegant way of ignoring
.. in directory listings.
So there’s plenty of room for new interesting solutions. Some will even (or at least should) get pushed into Emacs proper. Happy hacking!