4779eac75e799bc912ecd0ce11d064866114ee74
[frak.git] / README.md
1 # frak
2
3 frak transforms collections of strings into regular expressions for
4 matching those strings. The primary goal of this library is to
5 generate regular expressions from a known set of inputs which avoid
6 backtracking as much as possible.
7
8 ## "Installation"
9
10 Add frak as a dependency to your `project.clj` file.
11
12 ```clojure
13 [frak "0.1.1"]
14 ```
15
16 ## Usage
17
18 ```clojure
19 user> (require 'frak)
20 nil
21 user> (frak/pattern ["foo" "bar" "baz" "quux"])
22 #"(?:ba(?:r|z)|foo|quux)"
23 user> (frak/pattern ["Clojure" "Clojars" "ClojureScript"])
24 #"Cloj(?:ure(?:Script)?|ars)"
25 ```
26
27 ## How?
28
29 A frak pattern is constructed from a very stupid trie of characters.
30 As characters are added to it, meta data is stored in it's branches.
31 The meta data contains information such as which branches are terminal
32 and a record of characters which have "visited" the branch.
33
34 During the rendering process frak will prefer branch characters that
35 have "visited" the most. In the example above, you will notice the
36 `ba(?:r|z)` branch takes precedence over `foo` even though `"foo"` was
37 the first to enter the trie. This is because the character `\b` has
38 frequented the branch more than `\f` and `\q`. The example below
39 illustrates this behavior on the second character of each input.
40
41 ```clojure
42 user> (frak/pattern ["bit" "bat" "ban" "bot" "bar" "box"])
43 #"b(?:a(?:t|n|r)|o(?:t|x)|it)"
44 ```
45
46 ## Why?
47
48 [Here's](https://github.com/guns/vim-clojure-static/blob/249328ee659190babe2b14cd119f972b21b80538/syntax/clojure.vim#L91-L92)
49 why. Also because.
50
51 ## And now for something completely different
52
53 Let's build a regular expression for matching any word in
54 `/usr/share/dict/words`.
55
56 ```clojure
57 user> (require '[clojure.java.io :as io])
58 nil
59 user> (def words
60 (-> (io/file "/usr/share/dict/words")
61 io/reader
62 line-seq))
63 #'user/words
64 user> (def word-re (frak/pattern words))
65 #'user/word-re
66 user> (every? #(re-matches word-re %) words)
67 true
68 ```
69
70 You can view the full expression
71 [here](https://gist.github.com/noprompt/6106573/raw/fcb683834bb2e171618ca91bf0b234014b5b957d/word-re.clj)
72 (it's approximately `1.5M`!).