Definition of _array_data.data seems incorrect

[vaitkus]

The DDLm definition of _array_data.data defines it as an array of Complex numbers while it should most likely be a Text string (code type in the DDL2 version).

[jamesrhester]

DDL2 envisions only CIF and CBF files as targets, and thus the DDL2 types describe the construction of textual values. _array_data.data in DDL2 is binary, to indicate a departure from this DDL2 "pure-text" philosophy. In a CBF file, binary means actual binary (i.e. sequence of arbitrary bytes) after a MIME header. In a CIF file, binary means a MIME header followed by text-encoded binary e.g. BASE64. In doing this, the DDL2 type is choosing to abandon any description of the interpretation of the contents (unlike e.g. 'yy-mm-dd' or 'float'). Because everything has a binary representation, the binary type in DDL2 is really only serving to partially describe how to encode arbitrary bytes in CBF/CIF and giving no information on interpretation.

DDLm is format agnostic, so our _type.contents provides information about the interpretation (and potentially the internal format in text-based files). It needs to be able to provide information in an HDF5 context as well, for example. In an HDF5 context, 'binary' is truly useless as everything is binary.

We know that the intention behind binary in _array_data.data is that the contents should be interpreted as an array of numbers, which could be integers or reals. I chose 'complex' as a placeholder because it's the richest number type so it covers all bases. We could downgrade that to Real and if anybody ever wants to encode complex numbers in an array we would define a different type.

I hope that makes sense. This is certainly open to discussion and a better solution.

[vaitkus]

Sorry, I somehow misread the DDL2 dictionary. I could swear the type was code, but it is indeed binary.