Currently, master_create_distributed_table() works successfully if I pass an index name as the table name and the column on which the index is as the partition key. It also puts an entry into the partition metadata table. master_create_worker_shards() errors out saying that the relation is not a table when an index name is passed to it, which is how I feel master_create_distributed_table() should behave as well.
Again not something we’d expect users to do, but IMO its better if the UDF guards itself against invalid arguments, specially since it is directly exposed to the end users.
From @samay-sharma on November 20, 2014 15:52
Currently,
master_create_distributed_table()works successfully if I pass an index name as the table name and the column on which the index is as the partition key. It also puts an entry into the partition metadata table.master_create_worker_shards()errors out saying that the relation is not a table when an index name is passed to it, which is how I feelmaster_create_distributed_table()should behave as well.Again not something we’d expect users to do, but IMO its better if the UDF guards itself against invalid arguments, specially since it is directly exposed to the end users.