Coverage Results ✅

Name    Stmts   Miss  Cover   Missing
-------------------------------------
TOTAL   18665      0   100%

236 files skipped due to complete coverage.

This feels like quite an extreme fix. I think there should be a better way of resolving this.

In the original issue the table name is there:

select
        pj.hash_key,
        pj.snapshot_date,
        trim(cast(pj.metadata_super."amount" as numeric))                                       as amount,
        ...
    from w_parse_json as pj

It's just qualified more than once (although that's Redshift). With the Athena example, I think it would be enough to only flag issues with the rule if there's more than one table in the query - at which point we should already expect that the first identifier should be the table name.

On interesting variant would be to just change the error message slightly, to indicate that if it's in Redshift or Athena - and it's a structured access type - we recommend having the table name as the first element.

Thoughts?

This feels like quite an extreme fix. I think there should be a better way of resolving this.

In the original issue the table name is there:

select
        pj.hash_key,
        pj.snapshot_date,
        trim(cast(pj.metadata_super."amount" as numeric))                                       as amount,
        ...
    from w_parse_json as pj

It's just qualified more than once (although that's Redshift). With the Athena example, I think it would be enough to only flag issues with the rule if there's more than one table in the query - at which point we should already expect that the first identifier should be the table name.

On interesting variant would be to just change the error message slightly, to indicate that if it's in Redshift or Athena - and it's a structured access type - we recommend having the table name as the first element.

Thoughts?

Alternatives which I was considering were to skip rule if query contains only one table reference (effectively assuming that columns are unqualified), or maybe skip rule if query contains one unaliased table reference.

This effectively solves both athena and redshift but was not consistent with repository main.

At the same time it's more fundamental rework - so let's get agreement that it's a valid direction before starting implementation.

@alanmcruickshank it ended up being a fairly modest uplift. I am happy with the essence of the fix, test cases / docs possibly can be refined.

I squeezed in for #6402 since it was somewhat close to touched code regions.