Why does REGEXP_REPLACE concatenate with a previous result?

This looks like strange behaviour, and I get the same on the MySQL command line as via db<>fiddle:

create table t(foo varchar(255));

insert into t (foo) values
('the quick brown fox'),
('jumped over the lazy dog');

select * from t;

| foo |
| :----------------------- |
| the quick brown fox |
| jumped over the lazy dog |

select regexp_replace(foo, '[^a-z]+', '-') from t;

| regexp_replace(foo, '[^a-z]+', '-') |
| :------------------------------------------ |
| the-quick-brown-fox |
| the-quick-brown-foxjumped-over-the-lazy-dog |

db<>fiddle here

I'd expect to see:

select REGEXP_REPLACE(foo, '[^a-z]+', '-') from t;

| REGEXP_REPLACE(foo, '[^a-z]+', '-') |
| :---------------------------------- |
| the-quick-brown-fox |
| jumped-over-the-lazy-dog |

Am I doing something wrong or is this a MySQL bug?

It's a known bug, fixed in 8.0.12.


REGEXP_RELACE() results from one result set row could carry forward to
the next row, resulting in accumulation of previous results in the
current row.

StackExchange Database Administrators Q#212125, answer score: 5