Query to check sequence overflow sql

[BLoHny]

relate : mfvanek/pg-index-health#363

[mfvanek]

Hi @BLoHny,
I've left some comments.

[BLoHny]

select
    nsp.nspname as schemaname, -- schema name
    case
        when increment_by > 0
        then
            100.0 * (max_value - coalesce(last_value, start_value)) / (max_value - min_value)
        else
            100.0 * (coalesce(last_value, start_value) - min_value) / (max_value - min_value)
    end :: numeric(5, 2) as remaining_percentage -- percentage of remaining values
    pg_get_serial_sequence(col.attrelid::regclass::text, col.attname) as sequence_name, -- check sequence growth direction (increment or decrement)
from pg_catalog.pg_sequences
join
    pg_catalog.pg_class t
    join pg_catalog.pg_namespace nsp on nsp.oid = t.relnamespace
    join pg_catalog.pg_attribute col on col.attrelid = t.oid
where
    not cycle -- exclude cyclic sequences
    and nsp.nspname = :schema_name_param::text
order by schemaname, sequencename;

Does this code satisfy request?

[mfvanek]

Does this code satisfy request?
It doesn't work. :(

It should be

create sequence if not exists demo.seq1; -- bigint by default
create sequence if not exists demo.seq2 as smallint;
create sequence if not exists demo.seq3 as smallint increment by -200;
select nextval('demo.seq3'); -- call several times

with all_sequences as (
    select
        s.seqrelid::regclass::text as sequence_name,
        s.seqtypid::regtype as data_type,
        s.seqstart as start_value,
        s.seqmin as min_value,
        s.seqmax as max_value,
        s.seqincrement as increment_by,
        case when has_sequence_privilege(c.oid, 'select,usage'::text)
        then pg_sequence_last_value(c.oid::regclass)
        else null::bigint end as last_value
    from
        pg_sequence s
        join pg_class c on c.oid = s.seqrelid
        left join pg_namespace n on n.oid = c.relnamespace
    where
        not pg_is_other_temp_schema(n.oid) -- not temporary
        and c.relkind = 'S'::char -- sequence object
        and not s.seqcycle -- skip cycle sequences
         -- and n.nspname = :schema_name_param::text
        and n.nspname = 'demo'::text
),
sequence_state as (
    select
        t.sequence_name,
        t.data_type,
        case when t.increment_by > 0 -- ascending or descending sequence
        then 100.0 * (t.max_value - coalesce(t.last_value, t.start_value)) / (t.max_value - t.min_value)
        else 100.0 * (coalesce(t.last_value, t.start_value) - t.min_value) / (t.max_value - t.min_value)
        end ::numeric(5, 2) as remaining_percentage -- percentage of remaining values
    from all_sequences as t
)
select *
from sequence_state as s
where
    --s.remaining_percentage <= :remaining_percentage_threshold::numeric(5, 2)
    s.remaining_percentage <= '99.99'::numeric(5, 2)
order by s.sequence_name;

Prints

demo.seq3	smallint	96.34

[mfvanek]

@BLoHny
You'd better to revert your last commit because the query is wrong now.
Please retest the query before requesting review again.

There was only one issue with query

  == [/github/workspace/sql/check_sequence_overflow.sql] FAIL
  L:  42 | P:   5 | RF03 | Qualified reference 's.remaining_percentage' found in
                         | single table select which is inconsistent with previous
                         | references. [references.consistent]

https://github.com/mfvanek/pg-index-health-sql/pull/42/checks?check_run_id=24604476669

I guess we can fix it with s.*

select s.*
from sequence_state as s