SPARQL 1.1: variable introduced by GRAPH is not affecting MINUS disjointness - #228

Since our graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This code adds metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228

Since our graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This code adds metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228

Since our graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This code adds metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228

Since our algebraic graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This is fixed by adding metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228

Since our algebraic graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This is fixed by adding metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228
joachimvh/SPARQLAlgebra.js#128

Since our algebraic graph translation is not really part of the spec,
there is a MINUS edge case we need to consider with GRAPH ?g.
If left and right of MINUS have disjoint variables,
the whole left solution sequence must be kept.
If GRAPH is defined outside of an operator (e.g. MINUS),
then the spec says that evaluation of the operators
must be done as union over the evaluation of that operator
within each graph separately,
and that the variable of ?g must only be bound
**after** that evaluation.
As such, MINUS will not be aware of this variable ?g,
and the disjoint case will apply.
This is fixed by adding metadata to the operation so that engines
can special-case this.

See w3c/rdf-tests#228
joachimvh/SPARQLAlgebra.js#128