Message Boards Message Boards


Easier way to write LibraryLink code (code & design review request)

Posted 8 years ago
1 Reply
1 Total Likes

I am working on a package to make it easier to write LibraryCode. I started this out of personal need, but I would like to make the package general enough to be of use to a broader audience. I found the barrier to writing LibraryLink extensions too high, especially when managed library expressions were involved. I tried to lower this by generating all the boilerplate code based on "templates" that describe the library interface.

This post is to request feedback on the package design. The packages is far from done yet, but it's complete enough to demonstrate the idea and to ask for guidance on how to proceed.

It can be downloaded from

Any feedback and constructive criticism is most welcome!

The following example shows how it can be used. While incomplete, I hope it conveys the main idea well.

Load package


Write the code as a C++ class. For this short example I'm going to Export the C++ source directly form Mathematica, but normally this would be written separately. The following class is for computing the mean and variance of an array in one go.


code = "
  class MeanVariance {
    double m, v;

    void compute(mma::RealTensorRef vec) {
     double sum = 0.0, sum2 = 0.0;
     for (mint i=0; i < vec.length(); ++i) {
         sum  += vec[i];
         sum2 += vec[i]*vec[i];
     m = sum / vec.length();
     v = sum2 / vec.length() - m*m;

    double mean() { return m; }
    double variance() { return v; }

Export["MeanVariance.h", code, "String"]

Define the interface to the class through a "template" definition:

template =
    LFun["compute", {{Real, _, "Constant"}}, "Void"],
    LFun["mean", {}, Real],
    LFun["variance", {}, Real]

Compile & load:


Now the class can be instantiated as a managed library expressions using Make. When an expression is no longer referenced from Mathematica, it is automatically freed.

meanVariance[arr_] :=
 Module[{obj = Make[MeanVariance]},
  {obj@"mean"[], obj@"variance"[]}

list = RandomReal[1, 100];


{0.521816, 0.0888919}

This also provides a solution to the multiple return problem with LibraryLink (without having to resort to MathLink).

POSTED BY: Szabolcs Horvát

LTemplate version 0.2 is available now. It has received many improvements to allow it to support IGraph/M. The biggest new feature is the ability to privately embed LTemplate into any Mathematica applications that you might build with it.

The API is still not stable but any future changes will hopefully me minor.

POSTED BY: Szabolcs Horvát
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
or Discard

Group Abstract Group Abstract