Skip to content

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
A native node extension to load and execute .Net libraries
C++ C# JavaScript C
branch: master

This branch is 16 commits ahead, 1 commit behind saary:master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Failed to load latest commit information.
Sharp
SharpTime
StubClass
bin
.gitignore
.npmignore
LICENSE
README.md
bindings.gyp
package.json

README.md

Loading .Net libraries in Node Js

This is a native nodejs extension which allows us to load dot-net assemblies, create objects and call methods on them.

Based upon

saary's node.net and v8sharp

The interface

Instantiate the module, load a dll and create a class. Currently it searches the active directory of the node process for assemblies.

var sharp = require('./Sharp').Sharp();
var stub = sharp.load('StubClass.dll');
var oc = stub.new('StubClass.Output');

Then the class exposes all methods for synchronous and asynchronous callbacks.

var list = oc.call('ReturnList');
// list = [17, 6, 81]  // some random set of integers
oc.async('ReturnObject', function(err, data){ 
    console.dir(data); 
});

There is also a .methods() call to return a list of callable method names.

You can call constructors and access static methods now too.

var constructed = stub.new('StubClass.Constructor', 1, "string");
var st = stub.static('StubClass.Static');
st.methods(); // list all static methods
var val = st.call('GetString');

What is it for?

Some of the linked resources from node.net, and research on the subject will tell you that C++/CLI has a performance hit in node over native javascript. This add-on uses System.Reflection for most operations, and additional marshaling into managed types which further add to the cost. So in short - your production system probably shouldn't rely on this path. It is intended for prototyping, when you already have classes to do what you want in C#. And the same advice as native modules apply - if your calls are long and async, then the costs of transforming the data are a smaller fraction.

Things that work

  • Listing dlls in the current directory
  • Loading managed assemblies (doesn't have to be .dll, loading Sharp.node has all the fun generated internals)
  • Call zero argument constructors
  • Calling methods where types can be converted

Speaking of - type conversion (mostly inherited from v8sharp)

  • string
  • int
  • date -> datetime
  • arrays -> typed arrays and lists
  • object -> dictionary of string and objects, and objects with public properties.
  • returned objects are returned

What's next?

From the previous list, there are many things which could be added to make the process a little more seamless.

  • access to fields
  • overloaded methods (honestly, this is hard when you're not a compiler)
  • also playing with node-gyp and npm to get the build install to be more streamlined.
Something went wrong with that request. Please try again.