Data types

What a column can hold: the supported Parquet/Arrow types and their Fortran counterparts, the three temporal element types, and the compact variable-length string container.

• Supported data types — the Fortran-kind-to-MAML type table, how nulls work and how a column is declared null-free, reading a column into a different numeric kind, the size limits, and what another producer's file can contain.
• Date, time and timestamp columns — parquet_date/parquet_time/ parquet_timestamp: building a value from civil fields, ISO-8601 parsing and formatting, comparison and difference/offset arithmetic, Unix-time and MJD/JD interop, declared units, and nulls that live in the element itself.
• Compact string columns with parquet_string_column — an offsets-plus-payload string store that needs no fixed width and no padding, plus the lightweight parquet_string handle that refers to one of its elements; searching, bulk row-set operations, and threading inside a single column.