"Vlad Marchenko" <vlad@infonet.com.ua> writes:
> test=# select count(*) from subscribers;
> count
> -------
> 96856
> (1 row)
> test=# select * from subscribers where email~'domain4';
> sid | email | domain | status | password | signup | name | sex | address |
> city | state | zip | country | phone
> -----+-------+--------+--------+----------+--------+------+-----+---------+-
> -----+-------+-----+---------+-------
> (0 rows)
> test=# INSERT INTO subscribers(email) SELECT (random()||'@domain4.com') from
> subscribers;
> ERROR: Cannot insert a duplicate key into unique index
> subscribers_email_key
> test=#
> I assume that this is because of pure transaction data isolation?
No, this just says that when you generated 96856 random() values, you got
at least one pair of duplicates.
regards, tom lane