I have a data from ControlM (scheduler) this way
PREDECESSOR_JOB_ID,SUCCESSOR_JOB_ID XYZ,ABC ABC,LMN ABC,PQR XYZ,EFG EFG,STU STU,TUV STU,VWX . . .
I saw many solutions where they have started from first node and descended to last node. But that is likely to fan-out with this data. What I am looking for a fan-in solution, where I can start with any end node and come towards first node
In this case the answer that am looking for is
ABC-->XYZ LMN-->ABC-->XYZ PQR-->ABC-->XYZ EFG-->XYZ STU-->EFG-->XYZ TUV-->STU-->EFG-->XYZ VWX-->STU-->EFG-->XYZ
Consider below approach
with recursive iterations as ( select successor_job_id, predecessor_job_id, 1 pos from your_table union all select b.successor_job_id, a.predecessor_job_id, pos + 1 from your_table a join iterations b on b.predecessor_job_id = a.successor_job_id ) select successor_job_id || '-->' || string_agg(predecessor_job_id, '-->' order by pos) as jobs_path from iterations where not successor_job_id is null group by successor_job_id
if applied to sample data in your question – output is