I take note of your keyword suggestion "LATERAL" vs "LATERAL JOIN".
On Tue, Dec 20, 2022 at 6:01 PM Andy Seaborne <a...@apache.org> wrote: > On 20/12/2022 17:24, Marco Neumann wrote: > > OK I will repeat this elsewhere again but as you are aware there was > > a good discussion on the SPARQL-12 list with plenty of suggestions. > > maybe something like Sideways Information Passing > > or correlate subquery join. > > Noting I didn't choose the name. > > The keyword is "LATERAL" not "LATERAL JOIN". > > (SIP is an implementation technique for joins - it is best if the > ordering of the left-right flow is "jump forwards") > > SQL has some several forms of correlated subquery. > > I see a suggestions - it's more like exploring the space - no consensus. > LATERAL is the one that examples come back to. > > The nearest seems to be FOR EACH but that isn't infix syntax which > causes more text. > > "flatMap" is not data-user friendly. > > ---- > > Without nested results tables, it won't solve the general GraphQL case. > As per discussion. > > And nested results tables would be a whole different design. > > ---- > > If variables must be named: > > 1: It is more work for the user in the common case. > > 2: Putting a projection in the subquery for the uncommon cases seems > better UX > > 3: What happens if a variable is not bound? > > 4: Not the same as EXISTS so no implementation sharing. > > Andy > > PS Historical note: > LATERAL is the semantics for SPARQL 1.0 LC1 > > -- --- Marco Neumann
