I would add that a possible solution could be some mechanism for
conditional binding, e.g., BIND ( IF ( ?surname != '', ?surname, UNBOUND
) AS ?mySurname ), but apparently there is nothing like that in SPARQL
(apart from FILTER inside OPTIONAL).
M
On 26/02/2017 20:35, Zak Mc Kracken wrote:
Hi Andy,
thanks for your answer. I knew about UNDEF, but the fact is I don't
have control over the data (as I said, I'm trying to fit this into
TARQL, which takes a csv and adds a VALUE block to a CONSTRUCT
template, to convert to RDF). I'll see what the authors of the tool
say about the option to recognise some special value as UNDEF,
although, even so, one would still have the problem of how to filter
out non-empty values that shouldn't produce a mapping (e.g., I'm
dealing with a file where '###' in a column of numbers means empty).
By the way, my current workaround is to produce triples with surname
set at '', and then to pass the resulting dataset to the Jena 'sparql'
command line, together with another CONSTRUCT that filters out those
values. It works, but it's a bit more cumbersome and not ideal for
large data sets.
Marco.
On 26/02/2017 16:54, Andy Seaborne wrote:
Zak,
Using ?surname in the OPTIONAL-BIND does do not what you want. You'll
want the VALUES block before the OPTIONAL but even then the problem
is that the OPTIONAL matches graph pattern (an empty one) so ?surname
is not set in the OPTIONAL.
Ideally, put UNDEF in the values.
Andy
PREFIX ex: <http://www.example.com/data#>
CONSTRUCT {
?id ex:has-name ?name; ex:has-surname ?mySurname
}
WHERE
{
VALUES (?id ?name ?surname)
{
( ex:1 "John" "Smith" )
( ex:2 "Zak" "Mc Kracken" )
( ex:3 "Ben" UNDEF )
( ex:4 "Alice" "Wonderland" )
( ex:5 "Mickey" UNDEF )
}
}
or use IF and make one branch an eval error:
PREFIX ex: <http://www.example.com/data#>
CONSTRUCT {
?id ex:has-name ?name; ex:has-surname ?mySurname
}
WHERE
{
VALUES (?id ?name ?surname)
{
( ex:1 "John" "Smith" )
( ex:2 "Zak" "Mc Kracken" )
( ex:3 "Ben" "" )
( ex:4 "Alice" "Wonderland" )
( ex:5 "Mickey" "" )
}
BIND ( IF ( ?surname!='' , ?surname, 1/0) AS ?mySurname ).
}
On 26/02/17 12:42, Zak Mc Kracken wrote:
Hi all, I was hoping to make this to work with Fuseki 2.3.0:
PREFIX ex: <http://www.example.com/data#>
CONSTRUCT {
?id ex:has-name ?name; ex:has-surname ?mySurname
}
WHERE
{
OPTIONAL { BIND ( ?surname AS ?mySurname ). FILTER ( ?surname != ''
) }
#BIND ( ?surname AS ?mySurname )
VALUES (?id ?name ?surname)
{
( ex:1 "John" "Smith" )
( ex:2 "Zak" "Mc Kracken" )
( ex:3 "Ben" "" )
( ex:4 "Alice" "Wonderland" )
( ex:5 "Mickey" "" )
}
}
i.e., I've a list of rows to build some RDF from, where certain field
values are empty or should be ignored. The result should produce
triples
with name+surname, when surname is valid, only name, when it is
invalid.
I always get back triples with name only, even when I use the
non-optional binding. So, how can I get this result, when using VALUES?
The longer story is that the problem have arisen from tests I was doing
with the very nice tool TARQL (https://github.com/tarql/tarql,
https://github.com/tarql/tarql/issues/55)
Thanks in advance,
Marco
