names and deduplication do not play together
since datahandle locations depend on the name of an algorithm and algorithms with different names can be deduplicated in case they have the same configuration, some algorithm locations cease to exist after the deduplication step, thus messing up the configuration if someone depends on these removed locations/producers.
I am looking for a smarter way to solve this than "for all deduplicatable algorithms with different names, update all downstream algorithm inputs"
One obvious way would be: Locations do not depend on names, but i kind of like the current way, which is "._" (hash maybe optional in the future)
Another way (not completely thought through):
At algorithm instantiation, it is also registered to a "global" container of configured algorithms, thus deduplicated at instantiation time.
In this scenario however, i see no way to change a name on an already registered algorithms without again having to update downstream input locations.
Name changing might wanted when some algorithm has the default name and an identical algorithm with a non default name needs to be registered, thus i would rather want to keep the non default name, since the user obviously cared more about its algorithm having a name than the user which didn't specify one.