Missing triples after a fragmentation

[woutslabbinck]

Missing type in the view

When executing npm run test with the substring fragmenter (INPUT_FRAGMENTATION_STRATEGY="substring") in the root.ttl file, the view <begin_of_IRI/root.ttl> is a tree:node as it contains the relations to the other nodes.
However, this node does not have its type declared in the file.

The expected triple that is missing is <begin_of_IRI/root.ttl> <http://www.w3.org/1999/02/22-rdf-syntax-ns#type> <https://w3id.org/tree#Node>.

Note: While this was tested using the substring behaviour, I assume it would be nice that the type is added when using any bucketizer in the LDES Action

Missing property path in relation

Furthermore, when known (which is the case in the substring fragmenter) could the tree:path be added to each relation. While it is not explicitly required according to the TREE specification, as when no property path, each triple from a member is evaluated, it gives more information and can improve performance when creating application on top of TREE/LDES.

[ddvlanck]

#55 contains the following fixes

• Every page contains a type, tree:Node → <https://ddvlanck.github.io/Republish-LDES/output/a.ttl> a <https://w3id.org/tree#Node>.
• For every relation a page has, tree:path is added when relevant:

Substring fragmentation

<https://ddvlanck.github.io/Republish-LDES/output/a.ttl> <https://w3id.org/tree#relation> _:df_1_52.
_:df_1_52 a <https://w3id.org/tree#SubstringRelation>;
    <https://w3id.org/tree#node> <https://ddvlanck.github.io/Republish-LDES/output/ar.ttl>;
    <https://w3id.org/tree#path> _:n3-0;
    <https://w3id.org/tree#value> "ar".
_:n3-0 <http://www.w3.org/1999/02/22-rdf-syntax-ns#first> <http://www.w3.org/2000/01/rdf-schema#label>;
    <http://www.w3.org/1999/02/22-rdf-syntax-ns#rest> <http://www.w3.org/1999/02/22-rdf-syntax-ns#nil>.

Subject pages → Does not have relations to other pages
Basic fragmentation → Does contain relations to other pages, but does not use the property path (so no tree:path)

<https://ddvlanck.github.io/Republish-LDES/output/0.ttl> <https://w3id.org/tree#relation> _:df_1_0.
_:df_1_0 a <https://w3id.org/tree#Relation>;
    <https://w3id.org/tree#node> <https://ddvlanck.github.io/Republish-LDES/output/1.ttl>.

[pietercolpaert]

For the basic fragmentation: can’t we make sure we add one based on a time property? Or is this too complicated?

[KasperZutterman]

If we want different relations for every fragmentation/bucketizer strategy, we should look into moving this into the bucketizers themselves maybe?

[pietercolpaert]

Can’t this then be fixed with this idea? TREEcg/bucketizers#3

[ddvlanck]

For a basic fragmentation it is not possible to add tree:path or a more specific relation, because we do not know the sorting of the original LDES. However, if we load the entire dataset into memory, we could apply a sorting based on the property path, which is then preferably set to a timestamp predicate, but I don't know if that is the most sustainable option.

The core relation information is defined in RelationParameters and allows different types of relations and properties tree:node, tree:value and tree:remainingItems. The property tree:path is added in the LDES Action itself