DynamicallyLoadedEmbedding.jl

Embed Julia with dynamical loading of libjulia at runtime.
Author GunnarFarneback
Popularity
9 Stars
Updated Last
1 Year Ago
Started In
October 2020

Dynamically Loaded Julia Embedding

This package serves two purposes.

  • Make Julia embedding more accessible and easier to use.

  • Provide a regression test (via PkgEval) so that changes in new versions of Julia that break this form of embedding do not pass unnoticed.

With "Dynamically Loaded" it is understood that libjulia is not linked into the embedding application but is loaded at runtime with dlopen/LoadLibrary. This comes with both advantages and disadvantages as discussed at some length in the background section. To summarize the main advantages are:

  • Not tied to a specific version of Julia.

  • Julia is not needed in the build environment.

  • The application can run without having Julia bundled or installed, for users who do not need the Julia functionality.

  • All interaction is made via cfunctions - simpler interface than the "regular" embedding.

The main disadvantages are:

  • More difficult setup of the embedding environment.

  • All interaction is made via cfunctions - less powerful interface than the "regular" embedding.

The idea of this package is to take care of most of the difficulties of this approach.

Theoretical Background

The concepts and steps involved in a dynamically loaded embedding are discussed in a separate theory section.

Practical Use

The c directory contains a minimal example C implementation. Some of the files can be reused and some need to be adapted to a specific application. How to go about this is discussed in the separate practice section.

Licensing

This package is released under the MIT license.

Acknowledgements

This package is sponsored by ContextVision, where the the same approach is used to embed Julia in an in-house application.