Created: 2008-11-27 14:35
Updated: 2016-05-08 17:57




Voorhees is a yet another Lua library to parse JSON documents. It is implemented in ANSI C and have no dependencies other than Lua itself.

It is based on the Pushdown Automaton implemented in the JSON_checker and is faster than any other JSON parser for Lua that I've found (do correct me if I'm wrong).

Also the encoding of the input JSON text is independent from the strings it returns. That is, the encoding of the input is auto detected whereas you specify the encoding of the strings it returns.

Voorhees also enables you to parse big JSON documents by supplying a generator function instead of having to load the full JSON text in a string.

Any parser that uses a machete has my vote


If you have set up luarocks just type luarocks make in the source directory.

In a *nix-like environment you can also do

make PREFIX=/usr install

This will install voorhees.so in /usr/lib/lua/5.1. Have a look at the Makefile if this isn't right for your system.


Here is a sample script:

-- import the module
voorhees = require 'voorhees'

-- parse a JSON string
data = voorhees.parse('[{ "key" : "string" }, { "answer" : 42 }]')

-- open a file to parse
file, err = io.open('bigfile.json', 'r')
if not file then
   print('Error opening '..err)

myNull = {}

-- parse the file in pieces to avoid storing the whole file in memory,
-- return latin-1 encoded strings, set the maximum stack size to 30,
-- and use myNull as the JSON null value
data = voorhees.parse(function() return file:read(1024) end,
                    'latin1', 30, myNull)


The default encoding of strings is UTF-8 ("utf8"), but also little endian UTF-16 ("utf16" or "utf16le") is implemented. There is also a latin-1 ("latin1") mode, which writes code points less than 256 and maps all other Unicode to '?' characters.

The default maximum stack size is 20. That is you can nest 20 levels of arrays and objects.

The default value to return in place of the JSON null keyword is stored in voorhees.null. It is defined as

local function null()
   return null

so you can reference it as both voorhees.null and voorhees.null().

The generator function

Instead of passing the whole JSON document in one string to the parser it is possible to feed the text piece by piece by supplying a generator function.

This function is called without arguments whenever the parser needs a new piece of the text and must return the next piece as a string (or a number which is then converted to a string). Voorhees does not care how long the strings are as long as they're not empty.

The JSON document is deemed finished when the function returns an empty string, nil or something else not a string, or raises an error. Also the generator function mustn't yield.


Illegal arguments results in an error being raised whereas syntax errors and encoding errors makes voorhees.parse() return nil followed by a string with the error message.


Voorhees is free software. It is distributed under the terms of the MIT license


Please send bug reports, patches, feature requests, praise and general gossip to me, Emil Renner Berthing esmil@mailme.dk.

Cookies help us deliver our services. By using our services, you agree to our use of cookies Learn more