Re: Restructuring program to cope with modern database layouts

I wonder how SQLite does this under the covers. They mapped different data types to storage classes. As I understand, things like e. g. varchar are an internal alias to "text". But then, SQLite doesn't bring with it a

Taking in account other DBMS, for example mysql, if you use unicode UTF8 fields, by default now it allocates 4 bytes "per character" (it allocates the maximum possible).

Regarding IBMi I think it has - historically - the correct approach, "more correct" than other platforms, because it has explicit CCSID attached to a piece of information.
A char buffer in RPG has-a CCSID. A prototype parameter has-a CCSID. A stream file on IFS has-a CCSID. A field in the db has-a CCSID. You don't guess like in other platforms and the conversion points are well defined if respected.
RPG has now support for unicode (i.e. substring in "natural" mode etc.).
To use richer services over unicode, like i.e. "remove all accents in a string", or translitarations, you call into ICU API (in QICU library on IBMi ... despite IBM should really take time to refresh it, the last mod dates I see on the SRVPGM on my system are 2009... cmon IBM.... refresh those C/C++ compilers....).