[BUG]: grid.transform has unnecessary checks for non-periodic dims

[dhruvbalwada]

Issue: grid. transform is checking for that even dimensions not being transformed along are not periodic. Peridocity should only checked in the direction that is being transformed (e.g why would a transformation to density grid from depth care if my longitude is periodic).

import xarray as xr 
import xgcm
import numpy as np
import matplotlib.pyplot as plt

# Make the coords
dx = 0.5
L = 20
x_left = np.arange(0,L-dx, dx)
x = np.arange(dx, L, dx)

dz = 10
L = 2000
z_left = np.arange(0,L-dz, dz)
z = np.arange(dz, L, dz)

da_z = xr.DataArray(z, dims=['z'])
da_x = xr.DataArray(x, dims=['x'])

# Make a fake density like profile
da_ones = xr.DataArray(np.ones( (len(x), len(z)) ), dims=['x','z'])
da_rho = da_ones*np.exp(da_z/-1000)

ds = xr.Dataset(coords={'x':da_x, 'xg':x_left, 
                          'z':da_z, 'zg':z_left})
ds['rho'] = da_rho

# Define grid
grid = xgcm.Grid(ds, 
                 coords={"X": {"center": "x", "left": "xg"},
                         "Z": {"center": "z", "left": "zg"}},
                periodic='X')
#                 periodic=False) To make it work

# Rho axis to be used for transformation
rho_ax = np.linspace(0,1, 50)

# do the transform
Z_transformed = grid.transform(ds.z.broadcast_like(ds.rho), 'Z', rho_ax, 
                               target_data=ds.rho)

The above code fails if we say that 'X' is periodic, even though we are transforming along 'Z'.

xgcm.__version__ = '0.8.1'

[hdrake]

The periodic argument seems to not longer be working properly, now that it is in process of being deprecated in favor of boundary (and I think eventually both in favor of padding?). I had the same issue as you and the following fixed it for me:

grid = xgcm.Grid(ds, coords=coords, boundary={'X':'periodic', 'Y':'periodic', 'Z':'extend'})

Also fixes #604