Repository navigation
Seemingly pointless test #1807
Description
While implementing the support for INHERITS, I happened to find this test which contains in its name the word inherit and I suppose may reference the CREATE ROLE INHERIT, but it not present in the actual test.
If this test is actually pointless, it may be best to remove it. Do let me know how to proceed.
#[test]
fn parse_create_table_with_inherit() {
let sql = "\
CREATE TABLE bazaar.settings (\
settings_id UUID PRIMARY KEY DEFAULT uuid_generate_v4() NOT NULL, \
user_id UUID UNIQUE, \
value TEXT[], \
use_metric BOOLEAN DEFAULT true\
)";
pg().verified_stmt(sql);
}
Activity
LucaCappelletti94 commentedon Apr 27, 2025
Ping to request I am expected to do with this, happy to remove the test if it is indeed pointless.
alamb commentedon Sep 28, 2026
It would be good to look at the PR that added it to see if there was some additional history
LucaCappelletti94 commentedon Sep 28, 2026
When I wrote the issue a while back I had checked it, but now I did not recall anymore and so I started an opus investigation which matches my recollection of it. Here is its summary:
The test came from ivanceras in #20 (commit 78c054e, 2018). In that first version the SQL ended with
INHERITS (system.record), which the parser of the time silently ignored because it stopped after the closing parenthesis. A week later, the refactor in 5abd9e7 (same PR) turned the test into ato_string()round trip, and theINHERITSclause was dropped because the AST could not render it. The name outlived the clause it was about.INHERITSitself has been covered since #1806 byparse_create_table_with_inherits, and the column features the old SQL still exercised (TEXT[],BOOLEAN DEFAULT true,DEFAULT ... NOT NULL) are covered byparse_create_table_from_pg_dumpand the array DDL tests, so #2595 lost no coverage.