Error building in Alpine Linux 3.21.2

[chentz78]

I'm trying to use the fathon in a Alpine Linux version 3.21.2.
There is no version available in pip and I trying to build it from source.

When I compile it and I get these error message:

> python3 setup.py

Compiling fathon/dfa.pyx because it changed.
Compiling fathon/dcca.pyx because it changed.
Compiling fathon/mfdfa.pyx because it changed.
Compiling fathon/mfdcca.pyx because it changed.
Compiling fathon/ht.pyx because it changed.
[1/5] Cythonizing fathon/dcca.pyx

Error compiling Cython file:
------------------------------------------------------------
...
            print('At least two points are required.')

    @cython.boundscheck(False)
    @cython.wraparound(False)
    @cython.nonecheck(False)
    cpdef multiFitFlucVec(self, np.ndarray[np.int_t, ndim=2, mode='c'] limitsList, float logBase=np.e, bint verbose=False):
                                              ^
------------------------------------------------------------

fathon/dcca.pyx:264:46: Invalid type.
Traceback (most recent call last):
  File "/root/Dev/fathon/setup.py", line 109, in <module>
    ext_modules=cythonize(extensions),
                ^^^^^^^^^^^^^^^^^^^^^
  File "/usr/lib/python3.12/site-packages/Cython/Build/Dependencies.py", line 1154, in cythonize
    cythonize_one(*args)
  File "/usr/lib/python3.12/site-packages/Cython/Build/Dependencies.py", line 1321, in cythonize_one
    raise CompileError(None, pyx_file)
Cython.Compiler.Errors.CompileError: fathon/dcca.pyx

My env is:

• LXC Container with Alpine Linux 3.21.2
• Python 3.12.8
• PIP 24.3.1
• numpy 2.1.3
• cython 3.0.10

[stfbnc]

Hi @chentz78, I'll take a look at it in the next days and try to provide a compatible wheel on pip.

[settwi]

Hi @stfbnc I have run into the same issue when building using numpy 2.3.4. It seems like their types or annotation syntax may have changed? I've never used cython myself so I'm unsure what the error message means. for now going to numpy 1.26.4 seems to work

[settwi]

@stfbnc i updated the types in the .pyx files to just use the cstdint int64_t and double instead of the Numpy float64_t. that should work with old/new versions of Numpy as well. i will put in a PR shortly.

one thing is that the return type of MFDFA.fitFlucVec is a little wonky for the HT example, so i had to modify that as well to get it to work. please feel free to change that detail if you want it done differently.

thanks