> As far as I can understand RexCorrelVariable is like a reference to the
> "current row”
Yes, that’s correct.
Also, think of a join, say
select * from Depts as d, Emps as e where e.deptno = d.deptno
implemented as nested loops, reading from an index. Then the pseudo-code is
for (Dept d in Depts) {
for (Emp e in Emps.getByDeptno(d.deptno)) {
// output row (d, e)
}
}
You can think of “d” as several things: in SQL it is a table alias; in
RelNode/RexNode terms it is a correlating variable, and in the pseudo-code it
is a loop iteration variable.
“e” is all of those things too, but we don’t bother creating a correlating
variable because it is never accessed out of its scope; it is always the
current row read from a particular input RelNode.
Julian
> On Jan 12, 2018, at 7:49 AM, Enrico Olivelli <[email protected]> wrote:
>
> Hi,
> I am trying to understand the meaning of RexCorrelVariable
>
> With this query I get a Plan like this:
>
> Query: UPDATE tblspace1.tsql set n2=n2+n2,n3=n3+n2 WHERE k1=?
>
> -- Best Plan
> EnumerableTableModify(table=[[tblspace1, tsql]], operation=[UPDATE],
> updateColumnList=[[n2, n3]], sourceExpressionList=[[+($cor0.n2, $cor1.n2),
> +($cor2.n3, $cor3.n2)]], flattened=[true]): rowcount = 1.0, cumulative cost
> = {2.5 rows, 8.5 cpu, 0.0 io}, id = 157
> EnumerableProject(k1=[$0], n1=[$1], s1=[$2], t1=[$3], n2=[$4], n3=[$5],
> EXPR$0=[+($4, $4)], EXPR$1=[+($5, $4)]): rowcount = 1.0, cumulative cost =
> {1.5 rows, 8.5 cpu, 0.0 io}, id = 156
> EnumerableInterpreter: rowcount = 1.0, cumulative cost = {0.5 rows, 0.5
> cpu, 0.0 io}, id = 155
> BindableTableScan(table=[[tblspace1, tsql]], filters=[[=($0, ?0)]]):
> rowcount = 1.0, cumulative cost = {0.005 rows, 0.01 cpu, 0.0 io}, id = 148
>
> here we have 4 RexCorrelVariable:
>
> sourceExpressionList=[[+($cor0.n2, $cor1.n2), +($cor2.n3, $cor3.n2)]]
>
> As far as I can understand RexCorrelVariable is like a reference to the
> "current row"
>
> Is it correct my understanding ?
>
> Thank you
>
> Enrico
