feat(interactive): Support IS_NULL and IS_NOT_NULL in Cypher Queries
#3167
Conversation
…uery to anti join
| @Test | ||
| public void where_7_test() { | ||
| RelNode multiMatch = | ||
| Utils.eval("Match (a) Optional Match (a)-[]->(b) Where b is null Return a").build(); |
There was a problem hiding this comment.
how do we control whether b is null can or cannot be pushed down? how about
Match (a) Optional Match (a)-[]->(b) Where b.age is null Return acan b.age is null be pushed down?
There was a problem hiding this comment.
The official Calcite FilterIntoJoinRule determines whether filter conditions can be pushed down into a join. The condition b.age is null is actually equivalent to b is not null and b.age is null, and in such cases, it can be pushed down into the join.
However, currently, when the compiler interfaces with the Calcite type system, it assumes that types coming from storage cannot be null, so b.age is null is inferred to produce no results and generate the following logical plan:
// Match (a) Optional Match (a)-[]->(b) Where b.age is null Return a
GraphLogicalProject(a=[a], isAppend=[false])
LogicalValues(tuples=[[]])
Here LogicalValues denotes the logical plan before the Return a produces no results.
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## main #3167 +/- ##
=======================================
Coverage 42.05% 42.05%
=======================================
Files 101 101
Lines 10985 10985
=======================================
Hits 4620 4620
Misses 6365 6365 Continue to review full report in Codecov by Sentry.
|
What do these changes do?
as titled.
Related issue number
Fixes