You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Infrastructure for calling nodes as lambdas with unique arguments #883
Because a Graphene document is turned, at graph compilation time, into a static call graph where each input to a node (its construction argument) is hardcoded in the node's struct as the result of evaluating its input node with static inputs, we can't easily make runtime calls to other nodes with varying inputs. Maybe something equivalent to dynamic dispatch can be employed to solve this limitation, but this will require some discussions with @TrueDoctor about Graphene architecture. The approach taken involves generalizing () and Footprint call arguments with Contexts which allows attaching custom key-value pair data to the call argument when given to predecessor nodes. This can then support varying that data between each repeated call to a predecessor node, like giving a unique index: usize to each iteration's call of a predecessor node.
We do have a provision for this which are the call arguments which can be supplied at runtime but using this for complex inputs would be a bit annoying atm
Activity
[-]Add a UI for lambdas in the node graph[/-][+]Lambda nodes[/+]TrueDoctor commentedon Jan 4, 2025
We do have a provision for this which are the call arguments which can be supplied at runtime but using this for complex inputs would be a bit annoying atm
[-]Lambda nodes[/-][+]Tracking Issue: Lambda nodes[/+]Keavon commentedon Mar 29, 2025
Closed by #2232.
[-]Tracking Issue: Lambda nodes[/-][+]Infrastructure for calling nodes as lambdas with unique arguments[/+]