DefaultDataAccess class crashes in case that two files of same vertical coordinate type have different number of levels.

[Joern Ungermann]

The consistency check of the data access class tries to compare the given vertical levels of all files and skips a file if they are not consistent. Sadly, the comparison did not take into account a differing number of levels. The comparison for different values thus crashes in case that a differing number of levels is encountered, This is obviously not intended as the desired behaviour here is also an error message and a skipping of the file.

[Reimar Bauer]

crashing is bad, bad also we should have some method to sort off wrong files. May be some descriptive meta file based on yaml or json to setup a minimum requirements. And everything what did not match becomes ignored.

[Joern Ungermann]

Indeed, the crash is the error, the desired behaviour is skipping of the file PLUS error in the log. As I see your comment, this is not clear. I'll update the issue description.

With respect to the descriptive meta file:

One gives the access class a "tag" that is used for identifying files. Files not matching this tag are not used. This is a very simple mechanism, with which one may have multiple associated filetypes in one directory. For example, it allowed the pressure-level and theta-level files of ClAMS to be in two different data groups that allow for differing init_times for the groups.

[Reimar Bauer]

Properly treat model files with differing vertical levels.

The data access class checks for consistency between discovered model data files. In case of a different number of vertical levels in two files, MSS Crashes as the comparison by np.allclose does not likle to arrays of different shape. A second checks before now checks for different shapes.

Fix issue #282

→ <<cset f0dedc60bcdd>>