Hi Karina,
I am not very clear about why `build_reloptions` is removed in patch
`v2-0002-better-error-message-for-setting-parameters-for-p.patch`, if
you can help explain would be great.
From my observation, it seems the WITH option has different behavior
when creating partitioned table and index. For example,
pgbench -i --partitions=2 --partition-method=range -d postgres
postgres=# create index idx_bid on pgbench_accounts using btree(bid)
with (fillfactor = 90);
CREATE INDEX
postgres=# select relname, relkind, reloptions from pg_class where
relnamespace=2200 order by oid;
relname | relkind | reloptions
----------------------------+---------+------------------
idx_bid | I | {fillfactor=90}
pgbench_accounts_1_bid_idx | i | {fillfactor=90}
pgbench_accounts_2_bid_idx | i | {fillfactor=90}
I can see the `fillfactor` parameter has been added to the indexes,
however, if I try to alter `fillfactor`, I got an error like below.
postgres=# alter index idx_bid set (fillfactor=40);
ERROR: ALTER action SET cannot be performed on relation "idx_bid"
DETAIL: This operation is not supported for partitioned indexes.
This generic error message is reported by
`errdetail_relkind_not_supported`, and there is a similar error message
for partitioned tables. Anyone knows if this can be an option for
reporting this `fillfactor` parameter not supported error.
Best regards,
David
On 2022-10-14 8:16 a.m., Karina Litskevich wrote:
Hi, Simon!
The new error message looks better. But I believe it is better to use
"parameters" instead of "options" as it is called "storage parameters"
in the documentation. I also believe it is better to report error in
partitioned_table_reloptions() (thanks to Japin Li for mentioning this
function) as it also fixes the error message in the following situation:
test=# CREATE TABLE parted_col_comment (a int, b text) PARTITION BY
LIST (a);
CREATE TABLE
test=# ALTER TABLE parted_col_comment SET (fillfactor=100);
ERROR: unrecognized parameter "fillfactor"
I attached the new versions of patches.
I'm not sure about the errcode. May be it is better to report error with
ERRCODE_INVALID_OBJECT_DEFINITION for CREATE TABLE and with
ERRCODE_WRONG_OBJECT_TYPE for ALTER TABLE (as when you try "ALTER TABLE
partitioned INHERIT nonpartitioned;" an ERROR with
ERRCODE_WRONG_OBJECT_TYPE
is reported). Then either the desired code should be passed to
partitioned_table_reloptions() or similar checks and ereport calls
should be
placed in two different places. I'm not sure whether it is a good idea to
change the code in one of these ways just to change the error code.
Best regards,
Karina Litskevich
Postgres Professional: http://postgrespro.com/
