Open pavelhoral opened 1 month ago
Played with this change a bit and there are quite a lot of corner cases that needs to be properly addressed (mainly in relation to phantom changes). For example right now there is no easy way to unset a property during synchronization process from within the mapping file. Also we probably need to make sure there is a way how to make the JDBC repo behave in a legacy compatible way to not break existing deployments.
Just to add what was the LEGACY behavior even before the recent JDBC repo refactor... NULL
s were alway null
s - https://github.com/WrenSecurity/wrenidm/blob/6.1.x/openidm-repo-jdbc/src/main/java/org/forgerock/openidm/repo/jdbc/impl/ExplicitResultSetMapper.java#L116 .
This changes the way how
NULL
values in mapped table columns are handled - allNULL
values are being treated asundefined
JSON values.This change makes sense as it aligns generic table behavior with the explicitly mapped behavior. However this also makes it impossible to actually save
null
values, but supportingundefined
is definitely a more reasonable default. This needs to be properly tested before merging (i.e. no phantom changes when working with Admin UI and during synchronization).