[FEATURE] Environment Support

[Eliza-Diggins]

✨ What would you like to see?

A comprehensive environment for containing fields, their derivations, etc.

💭 Why is this important?

Would provide a unified environment for maximizing the ability to reuse derived quantities like metrics and inverse metrics.

🧩 Additional Details

The idea would be to have some form of FieldContainer class which contains

1. A single grid backed by a single coordinate system.
2. A dictionary of Field objects.

Depending on the type of FieldContainer, it could just be a dictionary or it could be entirely HDF5 backed. It would also facilitate the creation of new fields by anticipating certain needs. For example, and HDF5 FieldContainer would know that new buffers are HDF5Buffer classes.

[Eliza-Diggins]

Development Plan

To implement this feature, we'll include a FieldManager base class with the following responsibilities.

• Shares a common grid and coordinate_system.
• Tracks and caches geometric quantities like metric tensor fields and metric densities.
• Supports named access and dynamic creation of Field subclasses within the manager.
• Optionally integrates specific storage backends to allow for entire FieldManagers to be written self-consistently to disk.

The standard FieldManager class would simply store the fields in a generic dictionary. Subclasses could alter this behavior
to ensure a specific storage backend.

API

Foundationally, the FieldManager class would have two dictionary like access points:

• The .fields and the .geometric_fields containers.

By default, access via index is to the .fields container:

FM['velocity'] # Fetch the velocity field.
FM['density'] = DensityField # set the density field.
...

In specialized backends, one might insist that FM['density'] = ... forces the given field to be copied to the desired disk backend and then reinitialized.

The .geometric_fields are stored as tuples (expression, axes). They are not stored by default, but may be stored as desired.

We'll have methods in the class like .gradient(field_name, ...) which will automatically seek the specific field and any of the relevant geometric quantities in the cache. Similarly, we might have .compute_expression(name,cache=False) to compute expressions.

[Eliza-Diggins]

Now under development in branch env_support.