Closed AnsleyManke closed 4 years ago
this is fixed as suggested above.
1) Allow single-letter I, J, K, L to indicate axis direction (but not M and N as those aren't as standard). That lets us use the example file as-is.
2) If the file initialization stage can't create consistent grids for the variables and dimensions in the file, stop with a note suggesting AXIS attributes for setting axis directions.
On opening some datasets we get warning messages like this (Wei Cheng recently reported this)
Possibly in conjunction with notes about unrecognized axis units. A file like this (shortening the ncdump header for clarity) fails to open correctly:
The file opens with a bunch of NOTE's and then has pretty wacky grids that can't be fixed up with Ferret commands:
but another similar file opens up fine, even with the same grid setup:
Both have 2-D variables longitude and latitude with axes (i,j) or (x,y) and also "vertex" or "bounds" variables that have dimensions (j, i, vertices) or (vertex, x, y)
Ferret/PyFerret doesn't determine the first file has axis i is in the X direction and axis j is in the Y direction. But in the second file it decides that x is in the X direction and y in the Y direction using the axis attribute.
I think we have changes to be made here:
1) Go ahead and decide that an axis named i is in the X direction. We already allow axes named X, Y, Z, or T to be in those directions; could also recognize I, J, K, or L as being in those directions. This doesn't solve things in general but would let this set of CMIP6 model outputs open correctly.
2) Change the behavior if any axis direction is still undetermined and we're unable to form consistent grids. Report the problem and stop with an error message rather than trying to continue.
**ERROR: inconsistent data grids: File needs axis attributes on coordinate variables indicating correct directions