Make DataArray a descriptor to enforce Data.data type invariant

[Copilot]

Data.data was enforced as a DataArray instance only via a hand-rolled @property setter. Converting DataArray itself into a Python data descriptor moves that responsibility into the type, making the invariant self-enforcing for all subclasses.

Changes

• Stoner/core/array.py — Added the descriptor protocol to DataArray:

  • __set_name__: records _attr_name = 'data' and _private_name = '_data' when the descriptor is installed on a class
  • __get__: returns np.atleast_2d(obj._data) via object.__getattribute__ (avoids allocating a default on every access)
  • __set__: contains the normalization logic previously in the property setter — 1-D → 2-D column, scalar → 2-D, >2-D raises ValueError, preserves setas/column headers when column count is unchanged

• Stoner/core/property.py — Replaces @property data / @data.setter in DataFilePropertyMixin with a single class-level descriptor assignment:

  data = DataArray([])

• Stoner/tools/tests.py — isproperty() now recognises data descriptors (any class attribute with both __get__ and __set__) in addition to property instances, since DataArray is no longer a property.

Behaviour

d = Data()
d.data = np.ones((10, 3))          # plain ndarray → automatically wrapped
assert isinstance(d.data, DataArray)  # always True

d.data = np.arange(5)              # 1-D → stored as (5, 1) column
assert d.data.shape == (5, 1)

d.data = np.ones((3, 3, 3))        # raises ValueError (>2-D)

The private _data attribute and all direct obj._data._setas accesses throughout the codebase are unaffected.

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.