juliainterop / octcall.jl

calling GNU Octave functions from the Julia language

OctCall: Calling GNU Octave from Julia

This package allows you to call and interact with GNU Octave, a mostly Matlab-compatible free-software numerical-computing language, from Julia. It works by directly accessing the GNU Octave C++ libraries using Cxx.jl, and hence should have performance comparable to calling Octave functions from within Octave.

Currently, communication of basic types such as numeric/boolean scalars, matrices/vectors, and strings are supported. Support for more types will be added in the future.

Currently, OctCall is a pre-release preview.

Installation

Before installing OctCall, make sure that GNU Octave is installed on your machine and that mkoctfile is in your PATH.

You should also add ENV["JULIA_CXX_RTTI"]=1 to your ~/.julia/config/startup.jl file.

octcall.jl's Issues

Consider adding "nout" as in Oc2Py

and if, then latest addition here, from 6 days ago:
blink1073/oct2py@8970a66

I'm not sure this applies to Julia, as prety standard to return many there (and I thought in Python?), and in MATLAB:

https://stackoverflow.com/questions/43394393/oct2py-only-returning-the-first-output-argument

Removed inferred nout for Octave function calls; it must be explicitly given if not 1. The old behavior was too surprising and relied on internal logic of the CPython interpreter.

I'm just filing so I do not forget, maybe this is not needed, but if a good idea, then also at MATLAB.jl?

FYI: Single-threading may be needed

For Oct2Py at least I got "Session died" (for inv function) and tracked it down to needing:

ENV["OPENBLAS_NUM_THREADS"]="1"

https://github.com/xianyi/OpenBLAS/wiki/faq#multi-threaded

OpenBLAS is ok with threading, but Octave doesn't support the combination, it seems, at least I fixed this way. This needs probably not be reported at Oct2Py, as I think Python doesn't enable threading by default, but Julia does, so for the combination of the three languages is a problem, or I assume would be also for you, this package.