How count(1), count(*) and count(column) Differ in Execution

When a table is large, analysis shows count(1) can be slightly slower than count(*), but the execution plans are the same. After statistics are gathered, count(1) may be a bit faster for data under 10k rows, though the difference is minimal.

Both count(1) and count(*) count all rows, including rows where columns are NULL. count(column) counts only non‑NULL values of that column, so NULL rows are ignored.

count(1) counts every record, including NULLs.

count(column) counts occurrences of the column, ignoring NULLs.

Performance observations:

If the column is a primary key, count(column) is fastest.

If the column is not a primary key, count(1) is faster than count(column).

When a table has multiple columns and no primary key, count(1) outperforms count(*).

If a primary key exists, counting the primary key column gives the best performance.

For a single‑column table, count(*) is optimal.

Example:

create table counttest (
  name char(1),
  age  char(2)
);
insert into counttest values
  ('a','14'),('a','15'),('a','15'),('b',NULL),
  ('b','16'),('c','17'),('d',NULL),('e','');
select name, count(name), count(1), count(*), count(age), count(distinct age)
  from counttest group by name;

The result shows that count(name), count(1) and count(*) all return the same row count, while count(age) ignores the NULL ages, and count(distinct age) reports distinct non‑NULL ages.