Attributes

Attributes store metadata on datasets and individual variables. All attribute values must be JSON-serializable.

Dataset Attributes

with cfdb.open_dataset(file_path, flag='w') as ds:
    # Set
    ds.attrs['title'] = 'Example dataset'
    ds.attrs['institution'] = 'My University'

    # Get
    print(ds.attrs['title'])

    # Iterate
    for key, value in ds.attrs.items():
        print(key, value)

Variable Attributes

with cfdb.open_dataset(file_path, flag='w') as ds:
    temp = ds['temperature']
    temp.attrs['units'] = 'degC'
    temp.attrs['long_name'] = 'Air Temperature'
    temp.attrs['valid_range'] = [-80, 60]

Dict-Like Interface

The Attributes class supports standard dict operations:

Operation	Example
Set	attrs['key'] = value or attrs.set('key', value)
Get	attrs['key'] or attrs.get('key')
Delete	del attrs['key']
Contains	'key' in attrs
Keys	attrs.keys()
Values	attrs.values()
Items	attrs.items()
Pop	attrs.pop('key')
Update	attrs.update({'key': 'val'})
Clear	attrs.clear()
All data	attrs.data (returns a deep copy)

JSON Requirement

All values must be JSON-serializable. This includes strings, numbers, booleans, lists, and dicts (with string keys). Numpy arrays and other complex objects are not allowed:

# OK
ds.attrs['scale'] = 0.1
ds.attrs['tags'] = ['climate', 'surface']
ds.attrs['config'] = {'version': 2, 'enabled': True}

# Raises ValueError
ds.attrs['data'] = np.array([1, 2, 3])  # not JSON-serializable

Read-Only Access

Attributes on a read-only dataset cannot be modified:

with cfdb.open_dataset(file_path, flag='r') as ds:
    print(ds.attrs['title'])      # OK
    ds.attrs['title'] = 'new'     # Raises ValueError

Template Attributes

When using coordinate or data variable templates (e.g., ds.create.coord.lat()), standard CF attributes like long_name, standard_name, and units are set automatically from cfdb-vars.