How can I find a query's predicate lineage?

[rycowhi]

Hey there!

I've been playing around with EventListeners trying to automatically generate lineage for my queries. I'm a huge fan of how the sources of each output column are mapped on a per column basis in the ioMetadata making this operation very easy.

Something I've noticed however, is that only column projections and not predicates are present in the resulting metadata sources.

An example:

DDL

CREATE TABLE memory.default.threshold (threshold INT);
CREATE TABLE memory.default.input (input_value INT);
CREATE TABLE memory.default.output (output_value INT);

DML

INSERT INTO memory.default.output 
SELECT * FROM memory.default.input 
WHERE input_value >= (SELECT max(threshold) from memory.default.threshold);

In the resulting sources for the target column output_value, we only get column input_value from memory.default.input. Is there a better way to determine this at run time than an EventListener? I would like to see memory.default.threshold's threshold column as a source, preferably with a way to differentiate between predicate and projection lineage.