Generic derived variables

[ecomodeller]

This PR adds the possibility to create derived variables without reading the entire file into memory.

The proposed transform function operates on data from a single timestep.

import mikeio
from mikeio.generic import transform, DerivedItem
# calc current speed from u,v
derived = [
      DerivedItem(
          item=ItemInfo("Current Speed", mikeio.EUMType.Current_Speed),
          func=lambda x: np.sqrt(x["U velocity"] ** 2 + x["V velocity"] ** 2),
      ),
 # add more items, e.g. direction
]
# include existing items except Total Water Depth
infilename = "HD.dfsu"
dfs1 = mikeio.Dfsu2DH(infilename)
sel_items = [
      DerivedItem(item=item)
      for item in dfs1.items
      if item.name != "Total Water Depth"
]
sel_items.extend(derived)

# Transform
transform(infilename, "transformed.dfsu", vars=sel_items)

The default is to append derived items to the list of existing items, controlled by the keep_existing_items flag.

[ecomodeller]

@buchmann-dhi we are trying to work out a simple syntax for creating derived items with the ambition of being flexible but still a terse and readable syntax. Any suggestions?

[ryan-kipawa]

Looks like an easy to use syntax, nice documentation, and great typing. My only comment is in regards to whether files with many timesteps are performant. That could also be an improvement later on (if it's an issue)...

[ryan-kipawa]

Have you checked performance on files with many time steps?

[ecomodeller]

Nope. But this is the same type of loop we have for all dfs files except dfs0. I.e. I would not use use this function to transform dfs0 files, which generally fits in memory.