Kirill Müller

Still in c0af3bf87c45f1f918adee75004376eaf2041fb9. Taking a look now.

Proposing new column names: `reference_table_name`, `reference_table_oid`, `reference_column_indexes`, `reference_column_names` .

I only now noticed the second-to-last row in the example: `constraint_index = 4` . This row doesn't seem quite right. An alternative approach to adding more columns is to add...

Is it a requirement that `constraint_index` values are dense and start at 0 for each tables? Or could they be unique within the returned table?

With krlmlr/duckdb@14d5652e4e3c398008946f9edcddef28d073f1f1 and the following modified input file, I'm seeing: ``` CREATE TABLE tf_1 ( a integer, "b c" integer, "d e" integer, PRIMARY KEY (a), UNIQUE ("b c"), UNIQUE...

I filed #4309. I can continue working on this issue here if we decide that we want "wide form" (without new columns or with just one new column). I still...

Are we okay with the following modification that affects the `constraint_index` column, krlmlr/duckdb@009ac70820a12868c65bf953bf2d6990e50fa1e3? Moving forward, a `FOREIGN KEY` row would have a corresponding `REFERENCED KEY` row with the same `constraint_index`,...

Can you please confirm: we are not introducing any new column names, we are representing the constraints in long form, and all that remains to do is to link the...

Fine, I'll think about it a bit more. Are we ever going to support foreign key constraints across schemas? ```sql create schema a; create schema b; create table a.x (c1...

I'm making good progress, but struggling with default schemas. Ideally, if the assertion introduced in krlmlr/duckdb@e7c0e134e5411acaea5fbca82ae46e01046d6bd3 was always fulfilled, we could rely on schema and table name to be always...