Maybe because I did not learn C before python, I often try to use a lazy direct (or mixup) way on code… So, for an experimental import script, I removed all Gramps IDs stuff and used a pseudo UID as handle for gramps object… Why ? Because there is an inherited check process (legacy) which complains for a conversion or something like that (old model for Gramps 1.0) during import. This has been removed on the script during cleanup, which was a little bit difficult to explain to AI during review…
So, even not a proper handle based on date but rather a pseudo UID for this script, the generated records into this custom Gramps XML will be imported into a new Family Tree.
True, maybe I should really use a right way for generating handle value, and except maybe a possible missing date value on the metadata record, what could be the risk to generate random handle record values before an import? About ID records, as we can import many dataset without ID, I did not really look at ID Formats, prefering a large range on handles values like maybe proper ID formats should do. Sure, an horrible and lazy mixup of design.
Note, does someone know how to easily corrupt a Gedcom file format?
Just set an ID twice…