Tests that need to be written

[hjabbot]

The existing tests are old an need updating. This issue is meant to be a place to host a list of tests we should implement. We should also prune through the existing tests to see what's necessary or redundant.

We can also create a bash script to automatically figure out what tests to run, and then run them, without us having to figure it out manually

[hjabbot]

Merging meshes

• Corner to corner
• Side to side
• Corner to side
• Side to corner
• Side to larger side
• Side to smaller side

[hjabbot]

Antimeridian

• Correctly aggregates data
• Correctly creates neighbour graph
• Also check global mesh as special case

[hjabbot]

Vector splitting

• Doesn't split open water
• Splits areas of high curl
• Doesn't split "visually curly" areas that have zero curl mathematically

[hjabbot]

Dataloaders

• Validation of meshed representation of that data using @SamuelHall700's mesh validator
• Individual dataloader tests for real data in a private repo to avoid licensing issues
• LUT dataloader splitting

[hjabbot]

Adding bugs / areas for improvement that I'm finding along the way here as a reminder to fix them after setting up testing:

cellbox.py:

• deallocate_cellbox() does not remove instance of self, unclear what intended purpose is

---

aggregated_cellbox.py: (All fixed in PR #70)

• load_from_json() includes 'id' as part of agg_data
• inconsistent naming between set_boundary and get_bounds methods
• contains_point() does not account for antimeridian

---

neighbour_graph.py:

• remove_node_from_neighbours() docstring says int values passed to method, but method fails when cellbox_indx is an int. Unsure how this method is actually meant to work, I think there's an error with the casting between int and str somewhere
• naming of methods is very confusing, e.g. update_neighbour, update_corner_neighbours, update_neighbours all exist and do different things. update_corner_neighbours appears most similar to update_neighbour, and not update_neighbours
• get_neighbour_case(), get_neighbour_case_bounds(), get_global_mesh_neighbour_case(), initialise_map() all take self as argument but have no call to self, should be staticmethods instead
• get_global_mesh_neighbour_case() doesn't return 0 when cellboxes aren't connected. This hasn't cropped up before because the method is only called within get_neighbour_case() after doing a check to see if they're connected
• update_neighbours() docstring says new_neighbour_indx is an int, but references it in code as if it is a list of at least length 2. Unclear what intended input is.

---

cli.py

• export_mesh_cli() does not support changing output filename
• merge_mesh does not change config of first provided mesh, so config params do not match the output mesh