git clone ''

(ql:quickload :blancas.kern)


Kern is a library of parser combinators for Clojure. It is useful for implementing recursive-descent parsers based on predictive LL(1) grammars with on-demand, unlimited look-ahead. The inspiration for Kern comes from Parsec, a Haskell library written by Daan Leijen, and from work by Graham Hutton, Erik Meijer, and William Burge.


The library is intended for parsing text of all kinds. In addition, there is support for parsing programming languages and data formats with language-like notations like JSON and EDN. The lexer combinators may be configured for various language features like handling of whitespace, block and line comments, case sensitivity, and patterns for identifiers, among others.

Parser Combinators

A parser is a high-order function whose job is to match an input string against a specific pattern. A combinator is a parsing function whose pattern is determined by parsers or by other combinators for the purposes of providing repetition, choice, filtering, or to enforce a certain order in the input. We also use the term parser in its more general sense, as a logical module of parsing functions for a particular task.

The key feature of this technique is the ability to devise a composition and sequencing of combinators for parsing a grammar, and to do so in such a way that the resulting code resembles the structure of said grammar. Furthermore, the resulting parser may itsef be composed, like any other, custom or predefined, to form more elaborate combinators.

Thus parser combinators offer means of abstraction and composition that result in a powerful and pleasant way to code parsing modules. Kern provides a rich set of parsers for higher productivity and not having to start from scratch. The next section illustrates the above by defining a custom parser for JSON data.

Sample Usage

Parsing JSON data.

pair    ::=  string ':' json
array   ::=  '[' (json (',' json)*)* ']'
object  ::=  '{' (pair (',' pair)*)* '}'
json    ::=  string | number | object | array | true | false | null
(use 'blancas.kern.core

(declare json)

(def pair (bind [f string-lit _ colon v json]
            (return [f v])))

(def array (brackets (comma-sep (fwd json))))

(def object (braces
              (bind [fields (comma-sep pair)]
                (return (apply hash-map (reduce concat [] fields))))))

(def json (<|> string-lit dec-lit float-lit object array bool-lit nil-lit))

Evaluate the json parser:

(run json "{\"fst\": \"Joe\", \"lst\": \"Hacks\",\"id\":1122}")
;; {"fst" "Joe", "lst" "Hacks", "id" 1122}
(run json "{\"id\":1122,\"scores\":[400,125,999],\"top\":true}")
;; {"scores" [400 125 999], "top" true, "id" 1122}


Kern's design isn't well-suited for achieving very high performance. It is based on pure functions and the systematic composition of its core parsers. My main goals are high productivity and usability. I'll make, however, every effort to improve the efficiency of the library.

Just to give a ballpark figure for performance at the REPL, on my MacBook Pro, without any JVM warm up, Kern will parse 350K worth of JSON data per second using the above definitions, with variations of some 10% either way. In contrast, data.json will parse 4MB in a quarter of a second. (I realize times can vary a lot, but after a few runs they converge somewhat; YMMV.)

I've written Kern as my preferred way of working on domain-specific languages, stand-alone and for program extensions. For these jobs I've found its performance quite satisfactory, with the big pay-off being the boost in productivity.



[org.blancas/kern "1.1.0"]



Changes for release 1.1.0:


Browse the whole change log.

Kern is documented in the Wiki.

Browse the Codox Kern v1.1.0 API.


YourKit is kindly supporting open source projects with its full-featured Java Profiler.

YourKit, LLC is the creator of innovative and intelligent tools for profiling Java and .NET applications. Take a look at YourKit's leading software products:


Copyright © 2013 Armando Blancas.

Licensed under the Eclipse Public License.