v0.6.5

This version brings three enhancements:

• Add support for annex files, which allow storing large variales (esp. arrays) into a separate file (#2e610dd, #6cb4072)
• Add support for xarray objects (#a726209, #6a0b393, #29878bc, #1164fe4)
• Add support for autogenerated Data containers (#39afda4, #9d69e69, #f6cf5ad, #bb2b11b)
  I.e. for simply cases where fields are simply stored, the nested Data class which specifies how
  a dataclass should be serialized is autogenerated by inspecting its fields.

As well as a few fixes:

• Fix issue with NumPy casting introduced by NEP 50 (#c05a880)

• Serialization fixes for

  • Pydantic dataclass (#f3a636c)
  • nested generic types (#e1ced3f)
  • NumPy values (#aba1ff7)

• Fix typos (#f8ff2ea, #39b7eb8)

• Fix readthedocs config (#b24611b)

• Update/cleanup docs (#f15cffd, #e01738c)

Full Changelog: v0.6.4...v0.6.5

v0.6.5-rc.2

[GH Actions] Regfix: publish-on-testPyPI

v0.6.5-rc.1

New functionality

• Autogenerated Data containers
• Support for xarray.DataArray
• Support for annex files: For data which is too large to store inlined into the JSON, an annex file can be used. A configuration option specifies where those annex files are located.

Full Changelog: v0.6.4...v0.6.5-rc.1

v0.6.4

Full Changelog: v0.6.3...v0.6.4

Patch for non-hashable but serializable functions

Add fallback to `base_encoder` when serializing functions

Things like callable Pydantic models may not be serializable with the
function serializer but still serializable with the base encoder.

v0.6.2

[pyproject.toml] Fix classifier "Typing :: Typed"