Unable to pass custom AlgebraFactory to mapOperation

[rubensworks]

In sparqlalgebrajs, we could pass a custom AlgebraFactory to mapOperation, which was important in cases where we use a non-standard DataFactory (within this AlgebraFactory).
However, this doesn't seem possible anymore.

This probably is the case for other utility functions as well.

[jitsedesmet]

Since mapOperation does not use the algebraFactory itself, I do not see the problem? When you need an algebraFactory in the mapOperation callBacks, we expect you to use closures for this.

[rubensworks]

Right, I forgot you have this deep clone approach, making this not required.

However, I am still of the opinion that such a deep cloning approach is very unsafe.
It's only a matter of time before we will run into problems due to this, and I'm afraid these will be difficult to debug.
Imagine for example a named node that is backed by a very deep or a non-cloneable object (e.g. a WASM binding, ...).

So I would recommend making clones explicit, by going through an algebra factory.

[jitsedesmet]

Yes, I'd need to think about a way to selectively clone in case this is needed. I doubt a factory approach (using classes) will be the way to do this. Some kind of index-based Transformer that requires you explicitly mention what keys should be visited seems more useful from the point of view of Traqula's generalizability towards other tree structures (like the AST, and potential algebra's/ASTs from other languages)

[rubensworks]

Some kind of index-based Transformer that requires you explicitly mention what keys should be visited seems more useful from the point of view of Traqula's generalizability towards other tree structures (like the AST, and potential algebra's/ASTs from other languages)

The problem is that you may not always know what keys are required explicitly.

In Comunica, it's possible for users to pass custom data factories when they execute a query.
So at development time, Comunica developers can not predict which keys must and must not be cloned.
This is something that can only be determined once someone passes in a custom data factory at query time.
As such, I don't see any way around delegating this cloning behaviour to the data/algebra factory.