dfm
str of Matern32Kernels
To string version of Matern32Kernels must be improved in order to allow copy and paste
On code:
kernels.Matern32Kernel(metric=18.9, ndim=3) + kernels.Matern32Kernel(metric=1.2, ndim=3)
When printing kernel name:
Matern32Kernel(metric=**Metric**(1.6, ndim=3, axes=array([0, 1, 2]), bounds=[(None, None)]), block=None) + Matern32Kernel(metric=**Metric**(1.2, ndim=3, axes=array([0, 1, 2]), bounds=[(None, None)]), block=None)
Bold emphasis on wrong printing
MWE
from george import kernels
k = kernels.Matern32Kernel(metric=18.9, ndim=3) + kernels.Matern32Kernel(metric=1.2, ndim=3)
print(str(k))
ERROR:
from george import kernels
k1 = kernels.Matern32Kernel(metric=Metric(1.6, ndim=3, axes=array([0, 1, 2]), bounds=[(None, None)]), block=None)
NameError: name 'Metric' is not defined
While it's true that there is an issue here, this report is not it. Yes, you need to import Metric if you want to copy and paste the __repr__, just numpy arrays and everything else!
The real issue is somewhere deep in the modeling protocol:
from george import kernels
k = kernels.Matern32Kernel(metric=18.9, ndim=3)
print(str(k))
# prints:
# >>> Matern32Kernel(metric=Metric(18.9, ndim=3, axes=array([0, 1, 2]), bounds=[(None, None)]), block=None)
Then:
from george.kernels import Matern32Kernel
from george.metrics import Metric
from numpy import array
Matern32Kernel(metric=Metric(18.9, ndim=3, axes=array([0, 1, 2]), bounds=[(None, None)]), block=None)
which should work, but fails with the error:
AttributeError: 'Metric' object has no attribute 'log_M_0_0'
It's not high on my priority list to fix this, but I'm happy to keep it open in case someone else wants to take a look!
