I have been having a chat in a different thread regarding the

tap-postgres

and overriding primary key. There are numerous answers that suggest using the metadata field to override this by setting

table-key-properties

,

view-key-properties

or

key-properties

. I’ve tried all of these with an empty array and it is still being inserted as a primary key into the snowflake table. I have tried: • deleting the snowflake database and starting from scratch. • reinstalling the tap and target • for a sanity check, creating a new tap that inherits from

tap-postgres

Looking through the logs for a job I can see that the

key_properties

in the stream doesn’t get overwritten, even if the catalog is showing the correct metadata. In other words, if I run

meltano invoke tap-postgres > output.json

, then

output.json

has

key_properties: ["id"]

in the SCHEMA json object (This is the actual primary key in postgres, which I want imported just as any other column) At this point, I’m not sure how else to debug this or where I should look…any ideas? I do see when I run

elt

that meltano sets the

key_properties

early in the job, via

2022-04-08T21:16:28.559100Z [debug    ] Setting '.streams[53].selected' to 'False'
2022-04-08T21:16:28.559184Z [debug    ] Setting '.streams[53].replication_method' to 'INCREMENTAL'
2022-04-08T21:16:28.559272Z [debug    ] Setting '.streams[53].replication_key' to 'updated_at'
2022-04-08T21:16:28.559357Z [debug    ] Setting '.streams[53].table_key_properties' to '[]'
2022-04-08T21:16:28.559436Z [debug    ] Setting '.streams[53].table_key_properties' to '[]'
2022-04-08T21:16:28.559514Z [debug    ] Setting '.streams[53].key_properties' to '[]'

but it doesn’t seem like it is actually doing anything.