Csound-numeric is a C11 library/plugin for Csound 7 that brings numpy-style N-dimensional array operations to Csound. 

The project includes real and complex arrays, broadcasting, indexing, masking, reductions, statistics, linear algebra, interpolation, resampling, FFT/STFT tools, set operations, persistence, and k-rate-aware versioning to avoid unnecessary recomputation.

You can build csn directly from the official repository https://github.com/PasqualeMainolfi/csound-numeric. 

csn is still under active development, so testing is especially welcome. Bug reports, crashes, numerical mismatches, unexpected NumPy-like behavior, shape/broadcasting issues, complex-number edge cases, realtime/k-rate problems, portability issues, and performance observations are all very useful. If you try it, feedback and reproducible test cases are greatly appreciated.

… a small example of the basis workflow, using an FFT:

instr 1 

sig:a = oscili(0.5, 440)

frame:CsnArr, new:k = csnsnap(sig, 1024, 256)

    spectrum:CsnArr = csnrfft(frame, 1024, -1, new)

    magnitude:CsnArr = csnabs(spectrum, new)

    out:a, ready:k  = csnstream(frame, 256)

endin

The goal has been to preserve the essence of Csound while bringing some of the simplicity and composability of NumPy to operations that are especially important in our workflows.

I am currently focusing on optimizing the deeper linear algebra layer, including operations such as determinants, matrix inversion, linear-system solving, and related routines, with particular attention to numerical correctness, memory reuse, and performance.