# Traverse link relation

### Description
Combines the functionalities of `Traverse relation` and `Cross dataset` blocks.
Like `Cross dataset`, it allows to map objects from a dataset onto another.
However, rather than comparing their identifiers to determine a mapping,
it relies on a third dataset (a "link dataset" `LS`) which contains explicit mappings in the form of additional triples.
The predicate to consider in the link dataset can be specified as a parameter. Mapping links can be traversed forward and backward.

### Input
- `A [OBJ]`: a list of objects
- `LS [OBJ]`: a dataset containing objects freom both `A` and `B` and mappings in the form of triples
- `B [OBJ]`: a list of objects


### Output
- `PAIR [OBJ,OBJ]`: a successful mapping where the first column comes from `A` and the second column comes from `B`
- `resA [OBJ]`: the first column of `PAIR`
- `resB [OBJ]`: the second column of `PAIR`

### Parameters
- `Predicate`: the predicate of a graph triple `(subject,predicate,object)`
- `Use sub-predicates`: when set to `true`, the values of all sub-predicates are also included. Sub-predicates can be defined in the data with the `rdfs:subPropertyOf` relation.
- `Direction`: the direction to use when traversing the relation with the given `Predicate` and `SOURCE`
  - `Forward`: `(A, Predicate, B = ?)`
  - `Backward`: `(A = ?, Predicate, B)`
  - `Either`: the result contains the union of `Forward` and `Backward` results

Output scores can be [aggregated](docs://score_aggregation) and/or [normalised](docs://score_normalisation).