I just realized that I incorrectly defined the mate pairs for the new coordinates XU, YU and XV, YV. Mate is an attribute that is defined for variables at velocity points in dataset with complex topology. For example, U and V are mates (pair), an information that is required when data across facets (faces) have different topology (this is a consequence of the logical X dimension increasing with -Latitude in rotated faces).
The incorrect definition is that right now I set XU and YU as mate pairs, and XV and YV as another mate pair. This is incorrect. The correct pair is:
XU, YV
and
XV, YU
The definition of these mate variables happens in llc_rearrange.py ( line 639-682). The result of incorrect assignment leads to a transformed dataset with those variables (XU, YU, XV, YV) having incorrect transformed dimensions. Since the transformation is all done lazily, there is no error that flags it.
Until I fix this later today with the other PR that closes #324, a simple workaround this is to drop the grid variables with wrong dimensions for the cut_od dataset after the transformation. This is:
Description
I just realized that I incorrectly defined the mate pairs for the new coordinates
XU, YU
andXV, YV
.Mate
is an attribute that is defined for variables at velocity points in dataset with complex topology. For example,U
andV
are mates (pair), an information that is required when data across facets (faces) have different topology (this is a consequence of the logicalX
dimension increasing with -Latitude in rotated faces).The incorrect definition is that right now I set
XU
andYU
as mate pairs, andXV
andYV
as another mate pair. This is incorrect. The correct pair is:and
The definition of these mate variables happens in
llc_rearrange.py
( line 639-682). The result of incorrect assignment leads to a transformed dataset with those variables (XU, YU, XV, YV
) having incorrect transformed dimensions. Since the transformation is all done lazily, there is no error that flags it.Until I fix this later today with the other PR that closes #324, a simple workaround this is to drop the grid variables with wrong dimensions for the
cut_od dataset
after the transformation. This is:Expect a PR later today fixing this.