Node bindings for hfst-ospell

This is a simple work-in-progress library which aims to offer spell checking using Hfst-ospell in node.js.

Note on dictionary files

You can find several dictionaries on divvun.no. Many are under the GPL, for some there is no license specified, though.

Assuming you accept the license terms, you can e.g. use

$ mkdir etc
$ curl http://divvun.no/static_files/zhfsts/se.zhfst > etc/se.zhfst

to download the dictionary for North Sámi.

Install

To install this module, you first need to make sure you have

• a C++ compiler,
• Python (for node-gyp),
• and libarchive (using e.g. apt-get install libarchive-dev or brew install libarchive && brew link libarchive --force)

installed.

You can then use npm install divvun/hfst-ospell-js to install the node module directly from GitHub (it is not yet published to npm).

Note: NPM versions older than 3.7.0 do not resolve git submodules when installing, so you may need to clone this repository manually and use npm install with a local path.

Usage

The API is pretty simple:

var hfstospell = require("hfst-ospell-js");
var path_to_dictionary = "etc/se.zhfst";
var spellchecker = new hfstospell.SpellChecker(path_to_dictionary);

// .suggestions(string) returns a Promise
spellchecker.suggestions("akkusativa")
.then((suggestions) => console.log(suggestions))
.catch((error) => console.error(suggestions));
// => ['akkusatiivva', 'akkusatiiva', 'akkusatiivan']

// But you can also use it with a callback
spellchecker.suggestions("akkusativa", (error, suggestions) =>
  console.log(error, suggestions));
// => null, ['akkusatiivva', 'akkusatiiva', 'akkusatiivan']

Development

After you cloned this repository, make sure to also fetch the hfst-ospell sources using git submodule update --init --recursive.

You can build the library using node-gyp configure build. (Feel free to ignore any warnings on lib/* files.)

Use npm test to verify the library works on the node side. Please note that this requires a dictionary file and tries to read etc/se.zhfst by default.

TODO

• Make it compile!
• Async with a small JS wrapper for Promise support
• Investigate thread safety of hfst-ospell (we are currently using Mutexes to only ever search for one suggestion at a time)
• Use a constructor for the wrapped C++ object so we can re-use the loaded spell checker
• Everything with TODO and FIXME in code!
• Compile tinyxml2 ourselves (it's just one file)
• Compile libarchive ourselves (it's a truckload of stuff)