issue_comments: 473416359

Just as note for me, to not have to reiterate:

• It seems, that variables are handled in creation order. Means that concat reads them, handles them (even if the variable remains unchanged) and writes them to the new dataset in that order.
• This does not happen for coordinate for some reason. There only the changed dimension is handled and written after the variables. The unaffected coordinates are written at the beginning before the variables.

Example (dst concat over 'x'): The ordering of the src_dim1 is because the dimensions in the variables/coordinates are x,y,z in that order. The ordering of the dst_dim1 is because the dimensions in the variables/coordinates are z, y, x. netcdf src_dim1 { dimensions: x = 2 ; y = 3 ; z = 4 ; variables: double test2(x, y) ; test2:_FillValue = NaN ; double test3(x, z) ; test3:_FillValue = NaN ; double test1(y, z) ; test1:_FillValue = NaN ; int64 z(z) ; int64 y(y) ; int64 x(x) ; netcdf dst_dim1 { dimensions: z = 4 ; y = 3 ; x = 4 ; variables: int64 z(z) ; int64 y(y) ; double test2(x, y) ; test2:_FillValue = NaN ; double test3(x, z) ; test3:_FillValue = NaN ; double test1(x, y, z) ; test1:_FillValue = NaN ; int64 x(x) ;

It seems, that the two coordinates (z and y) are written first, then the variables, and then the changed coordinate. Now trying to find, where this happens. If the two coordinates would be written in the same way as the variables (and after them), then the ordering would be x,y,z as in the source.