SPARQL property path, 1st part is unique, path repeats after - sparql

Question

Assume we have the triples
a1 blah <http://foo.com/f>
a1 pt c1
c1 pt d1
How would I construct a SPARQL query to select a1 or c1 or d1? I need to be able to walk one blah predicate or multiple pt predicates.
So a1 can be part of a triple like : a1 blah <http://foo.com/f>
but I dont know how to specify the case where its not a1, but its an object connected to a1 through another path in the other direction. But I also need to retain the path using blah, since thats what a1 or f1 is connected to the object by
Here is how I got it to work, but its not pretty:
select * from
{
# other sparql stuff before it
{ ?s blah <http://foo.com/f>
?s pt+ <c1> }
UNION
{ <c1> blah <http://foo.com/f> }
}
I can change 'c1' to 'a1' or 'd1' and the above sparql still works

Answer 1

AKSW's comment was exactly right, you need to use a sequence, not an alternative. Based on your example, it sounds like you actually want subjects that are connected to a specific object by a chain of zero or one occurrences of "blah", and then zero or more occurrences of "pt".
If you want to do this for specific predicates p and q, that you can specify with IRI references, then it's just
?x p?/q* ?y
The * following the q means zero or more, and the ? following the p means zero or one.
However, if you don't know the specific IRIs in advance, e.g., you want to use variables, in something like:
?x (?p)?/(?q)* ?y
then you may be out of luck. SPARQL property paths don't support variables. You can get wildcards in property paths with something like s|!s, since every URI is either s or it isn't, but there's no provision for putting repeatable variables into property paths.