groupby: A higher order function named groupby as known from Scala.
A short implementation of Scalas groupBy functionality for Common Lisp. The function comes in its own package to avoid naming conflicts. Actually it should be part of a utility library.
MIT License. See included LICENSE file.
Let's think of the task "find the first unique character" in a given string.
(defparameter *str* "lisp rulez")
The naive approach in Lisp is as simple as elegant:
(char (remove-if #'(lambda (x) (> (count x *str*) 1)) *str*) 0)
For simplicity let's ignore the point that it will throw an error when there is no uniqe char in str. Unfortunately the performance of the function is quite bad, it has quadratic runtime.
A quicker solution would look like the following lines. While this is fast, it is not as intuitive as the first solution.
(defun first-uniq-char (str) (let* ((size (length str)) (cnt (make-hash-table :size size))) (do ((i 0 (incf i)) ) ((<= size i)) (incf (gethash (char str i) cnt 0))) (do ((i 0 (incf i)) (c nil)) ((<= size i) c))))
Compare both functions with a long string:
(defparameter *strlong* (coerce (loop for i to 10000 collect (code-char (+ (char-code #\a) (random 24)))) 'string)) (setf (char *strlong* (- (length *strlong*) 2)) #\z) (time (char (remove-if #'(lambda (x) (> (count x *strlong*) 1)) *strlong*) 0)) (time (first-uniq-char *strlong*))
A factor of something around than 600 on my machine.
Now we make use of our new groupby function:
(time (remove-if #'(lambda (x) (> (second x) 1)) (mapcar #'(lambda (x) (list (first x) (length (second x)))) (gb:groupby #'(lambda (x) x) *strlong*))))
This seems to be only a little bit slower than the second, quick function above, while being an intuitive and quick solution again. First group the data by its identity, then count each of the groups and remove the groups with more than one elements.
As you can see, with the groupby higher order function it is possible to write very comprehensive code, that takes the standard lisp solutions to the next level. Its drawback seems to be, that it is quite not simple to identify the good use of this feature to proper solve problems.
The most simple and recommended way to install cl-groupby is by using Quicklisp. If you installed Quicklisp a simple
will download the package and acually load it. You only need to do this once per machine. Later a
will be enough.
You may get the code with:
git clone git://github.com/wlbr/cl-groupby.git
Either you add this to your asdf repository, then you will only need
to do a
(require :groupby) in your source.
Or, you may put the source in a subdirectory of your project and add
groupby.asd with its full path to your own asdf definition.
Or, you may put the source in a subdirectory of your project and load
the file "groupby.asd" directly. After that a
should be sufficient.
Or, as a kind of worst case, but definitely the most simple way, as this project consists of one function only: simply do a direct:
(load "package.lisp") (load "groupby.lisp")
Well, the whole thing is really just one function. Probably it is best to simply copy the few lines into your own code.
None except for asdf.
xlunit for the unit tests only (the tests are not included in the asdf system).
Tested with SBCL and CCL. No rocket science required, should run in any environment.
A set of unit tests is included in tests.lisp.
If you run into any trouble or find bugs, please report them via the Github issue tracker.
Written by Michael Wolber.