Data, saving and interchange
The right saved representation depends on the computation you want to do after loading it. Points, a finite encoding, a feature table and a drawing contain different information. A drawing lets someone inspect a result; a finite encoding keeps the spaces, maps and assignment of parameters to labels available for further questions. Keeping a selected mathematical view follows this choice through an encoding and the parameter queries we want to recover.
For a small example, save and reopen an interval view keeps a persistence interval together with its representative. A saved selection identifies which interval we were examining. Retaining the mathematical object as well lets us resume the supported inspector and ask another question. The recipe makes this distinction visible by saving and loading a ring whose hole has a known birth and death.
Sometimes the intended result is the figure itself. Composing and sharing mathematical figures develops that route: combine the relevant views and preserve the coordinates, labels and annotations that make them readable. A feature table serves another purpose, carrying numerical values into a subsequent analysis. Optional integrations explains how table, file and rendering adapters connect these choices to the surrounding Julia environment.
Recovering a finite encoding means more than recovering its dimensions. We also need its maps and the assignment from original parameters to finite labels. Even when those data describe the same module, a reconstructed basis may give a different matrix for the same map. The account of serialization and interchange explains what the stored formats preserve and how their checks relate to the mathematical object. It is the next place to look when choosing between a saved result, exact matrix coordinates and an object that supports new queries.