upcoef - TypeError: No matching signature found

[djessop]

I've come across this error when trying to reconstruct data from the coefficients given by dwt2 in a project to port some matlab code to python. An example of the data and code that I'm using is given below.

data = np.array([[ 1.47728539, 1.59481049, 1.59481049, 1.46021461, 1.44889259], [ 1.48014069, 1.48872185, 1.67795181, 1.38459396, 1.32822609], [ 1.41849327, 1.57595444, 1.50884819, 1.44937897, 1.57595444], [ 1.46079826, 1.53956032, 1.54245186, 1.53089523, 1.49732971], [ 1.49732971, 1.4323349 , 1.44937897, 1.49158859, 1.48872185]])

ca, (_, _, _) = pywt.dwt2(data, 'db1')

reconstructed = pywt.upcoef('a', ca, 'db1')

`---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
in ()
----> 1 pywt.upcoef('a', ca, 'db1')

/usr/local/lib/python3.4/dist-packages/pywt/_dwt.py in upcoef(part, coeffs, wavelet, level, take)
341 if part not in 'ad':
342 raise ValueError("Argument 1 must be 'a' or 'd', not '%s'." % part)
--> 343 return np.asarray(_upcoef(part == 'a', coeffs, wavelet, level, take))

/usr/local/lib/python3.4/dist-packages/pywt/_extensions/_dwt.cpython-34m.so in pywt._extensions._dwt.__pyx_fused_cpdef (pywt/_extensions/_dwt.c:7860)()

TypeError: No matching signature found`

However, using idwt2 works and gives approximately the same answers compared to the matlab code that uses the native upcoef2 function.

coeffs = (ca, (None, None, None))

reconstructed = pywt.idwt2(coeffs, 'db1')

I think that this kind of TypeError is usually raised when there is a conflict between type definitions, but everything is of type numpy.ndarray and dtype float64.

[grlee77]

The problem in your use case is that pywt.upcoef only supports 1D data (The underlying Cython code for upcoef is expecting a contiguous 1D array).

We do not currently have a 2D version of upcoef that operates like Matlab's upcoef2.

[grlee77]

I opened #339 to provide a more useful error message in this case.