bgrimstad/splinter

Simplify save/load functionality

bgrimstad opened this issue · 2 comments

bgrimstad commented

Consider removing serialization to shed some code and simplify maintenance. The binary format is not version proof and may depend on endianness.

It should be sufficient for most applications to save and load from JSON files. Currently, only BSpline objects can be saved and loaded from JSON files. It is not implemented for DataTable, for exampel, but this is probably a feature of minor importance.

Suggestion: Remove serialization, but keep save/load from JSON files. Rename save and load function to BSpline.to_json and BSpline.from_json to mimick the syntax of the Pandas library.

bgrimstad commented

Binary serialization was removed in 9952d80.

We should consider adding JSON serialization to DataTable.

bgrimstad commented

Completed in ab060ac.

Notable changes:

  • JSON for Modern C++ was updated to v3.1.
  • JSON serialization added to DataTable (for C++ and C, Python users can use Pandas). The format is verbose and not optimized for very large tables.
  • Interface changed: saving and loading from JSON files is now done via to_json and from_json for BSpline and DataTable (similar to Pandas)