Monotonic assert?

[infogulch]

Is there something like assert that interacts more naturally with choicepoints? Where the assertion is only valid/exists in the same branch that made the assertion. (Perhaps there is a better way to state this.)

In scryer today:

?- asserta(fact), fail; fact.
   true.

I would want:

?- assert_infogulch(fact), fail; fact.
   false.

[triska]

This could be solved by a concept called transactions. See for example Extending the logical update view with transaction support by Jan Wielemaker:

https://www.researchgate.net/publication/235357818_Extending_the_logical_update_view_with_transaction_support#fullTextFileContent

A transaction would "locally" assert new clauses, and make sure that everything is retracted on backtracking.

This could be an interesting future extension of Scryer Prolog.

[infogulch]

I suspect an implementation that stores asserted clauses with some kind of immutable data structure might be efficient enough to use pervasively.

[pmoura]

Is there something like assert that interacts more naturally with choicepoints? Where the assertion is only valid/exists in the same branch that made the assertion. (Perhaps there is a better way to state this.)
...

Would linear assumptions help in your case? See the following simple example and its bibliographic reference:
https://github.com/LogtalkDotOrg/logtalk3/tree/master/examples/assumptions