question about adding namespaces

[Luke-ebbis]

Dear developers, I have been exploring the usage of this tool in my data transformation pipeline. If find it to be quite useful in forming RDF documents. However, I am not sure on how to specify the namespace of my graphs I produce. For example:

clusterA(a, p1) .
clusterA(a, p2) .
clusterA(a, p3) .
clusterA(a, p4) .
clusterA(a, p0) .

clusterA(b, p100) .
clusterA(b, p101) .

clusterB(A, p0) .
clusterB(A, p1) .

clusterB(B, p2) .
clusterB(B, p100) .
% cluster(?name, ?proteins)


@prefix up: <http://purl.uniprot.org/uniprot/> .
@prefix base: <http://example.com/clustering/> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .


cluster1(IRI(?X), IRI(?y)) :- clusterB(?X, ?y).
cluster2(IRI(?X), IRI(?y)) :- clusterA(?X, ?y) .

% When a Cluster in set X is connected to a Cluster in set Y.
isConnected(?clusterNameX, ?p, ?clusterNameY) :- 
    cluster1(?clusterNameX, ?p) , % it must be in X
    cluster2(?clusterNameY, ?p) . % it must be in Y.

% When a cluster X has a connection to Y.
hasOutConnection(?X, ?Y) :-
    isConnected(?X, ?p, ?Y) . 

% When a cluster X has n connections to Y.
hasOutConnectionCount(?X, ?Y, #count(?p)) :-
    isConnected(?X, ?p, ?Y) .

% When a cluster X has n connections to Y.
hasTotalConnections(?X, #sum(?count, ?p)) :-
    hasOutConnectionCount(?X, ?p, ?count) .

hasWeightedConnection(?X, ?Y, ?w) :-
    hasOutConnectionCount(?X, ?Y, ?count) ,
    hasTotalConnections(?X, ?total),
    ?w = ?count / double(?total).    

hasConnection(?X, ?Y) :-
    hasOutConnection(?X, ?Y),
    hasOutConnection(?Y, ?X) .


% Exporting our analysis to RDF

cluster(?X, ?p, clusterA) :-
    cluster1(?X, ?p) .
cluster(?X, ?p, clusterB) :-
    cluster2(?X, ?p) .

up:Protein(?X, ?p, ?T) :- 
    cluster(?X, ?p, ?T) ,
    ?p = concat("-", str(?p)) .

graph(?X, base:hasMember, ?p) :- 
    cluster(?X, ?p, ?T).
graph(?p, base:memberOf, ?X) :- 
    cluster(?X, ?p, ?T).
graph(?X, base:cluster, ?T) :- 
    cluster(?X, ?p, ?T).
graph(?p, rdf:type, up:Protein) :- 
    up:Protein(?X, ?p, ?T). 
graph(?X, base:connectsTo, ?Y) :- 
    hasOutConnection(?X, ?Y) .
graph(?X, base:connectsTo, ?Y) :- 
    hasOutConnection(?Y, ?X) .
graph(?key, rdf:type, rdf:Statement),
graph(?key, rdf:predicate, base:hasOutConnection),
graph(?key, rdf:object, ?Y),
graph(?key, rdf:value, double(?w)),
graph(?key, rdf:subject, ?X),:-
    hasOutConnection(?X, ?Y) ,
    hasWeightedConnection(?X, ?Y, ?w),
    ?key = f"http://example.com/clustering/{?Y}-{?Y}", 
    ?key = uri(?key).

@export graph :- rdf{resource="rdf.ttl"}.
@export hasTotalConnections :-
    csv{resource="connectivity.csv"} .
@export hasWeightedConnection :-
    csv{resource="hasWeightedConnection.csv", 
    format=(any,any,double)} .

Then I get the error that complex terms are not supported. Is there a better way to do this? I am sure its possible but I cannot find it in the documentation.

[Laschoking]

Currently, fstrings are the correct way to combine prefixes and variables.
The error is caused because uri() is not an existing builtin function. You can use URIENCODE.
If you want to encode the fstring you could use:

    ?string = f"http://example.com/clustering/{?Y}-{?Y}", 
    ?key = URIENCODE(?string).

Fstrings also accept prefixes, so you can also write:

  @prefix base : <http://example.com/>.
  ...
   ?string = f"{base:clustering}/{?Y}-{?Y}", 

Unfortunately, prefixes require a non-empty local name so the following example is not valid:

  @prefix base : <http://example.com/clustering/>.
   ?string = f"{base:}{?Y}-{?Y}", 

[Luke-ebbis]

Thank you, that is clear! Two more questions,

1. How would I turn p1 to up:p1? I did not understand how this would work:
   an example such as this

graph(?p, rdf:type, up:Protein) :-
    cluster(?X, ?p, ?t) ,
    ?up = f"{<http://purl.uniprot.org/uniprot/>}{?p}", 
    ?p = IRI(?up).

does not work
2) Is it possible for me to add to the documentation of this tool? I would like to improve the documentation for other users with what I learned.

[Laschoking]

1. You can use prefixes to simplify constants and predicate names (see the docu here .)
   For facts you can simply write: clusterA(a, up:p1) .

your rule has a wrong design: you use the variable ?p twice and basically check the following (which is never evaluated to true):
?p = IRI(f"{<http://purl.uniprot.org/uniprot/>}{?p}").
You rather should use a new variable e.g, ?iri = IRI(?up) and use graph(?iri, rdf:type, up:Protein) in the head.

2. Our documentation resides at https://github.com/knowsys/nemo-doc.
   We would be happy if you would like to suggest changes, give feedback about the structure and content of the documentation or supply some of your own examples.

[Luke-ebbis]

Oke great, that works. I came up with this code to add the up namespace to my proteins.

clusterA(a, p1) .
clusterA(a, p2) .
clusterA(a, p3) .
clusterA(a, p4) .
clusterA(a, p0) .

clusterA(b, p100) .
clusterA(b, p101) .

clusterB(A, p0) .
clusterB(A, p1) .

clusterB(B, p2) .
clusterB(B, p100) .
% cluster(?name, ?proteins)


@prefix up: <http://purl.uniprot.org/uniprot/> .
@prefix base: <http://example.com/clustering/> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .


cluster1(IRI(?X), IRI(?y)) :- clusterB(?X, ?y).
cluster2(IRI(?X), IRI(?y)) :- clusterA(?X, ?y) .

% Exporting our analysis to RDF

cluster(?X, ?p, clusterA) :-
    cluster1(?X, ?p) .
cluster(?X, ?p, clusterB) :-
    cluster2(?X, ?p) .


graph(?X, base:hasMember, ?iri) :- 
    cluster(?X, ?p, ?T),
    protein(?iri) .

protein(?iri) :-
    cluster(_, ?p, _) ,
    ?up = f"{<http://purl.uniprot.org/uniprot/>}{?p}", 
    ?iri = IRI(?up).

graph(?iri, rdf:type, up:Protein) :-
    protein(?iri) .

@export graph :- rdf{resource="rdf.ttl"}.

[Luke-ebbis]

Is there also a way to have a default namespace so as to not use relative iri's?

[Laschoking]

I am not sure, if I understand your question correctly, but when importing rdf formats you can specify the base parameter. Imported relative IRIs will be made absolute based on this base value.

[mmarx]

There is also @base, which apparently is not (yet) documented, but can be used like so:

@base <https://example.org/> .
p(foo) .
p(<https://example.com/bar>) .
@output p.

This will yield https://example.org/foo and https://example.com/bar.

[Luke-ebbis]

Thanks @mmarx and @Laschoking . When I will find the time I will try to add this example in a generalised way to the documentation