Re: Adding same column to several tables.

On Thu, Jun 16, 2016 at 10:19:10AM +0200, Łukasz Brodziak wrote:
> Hello,
> I have a function to add column to all tables from the
> list:do$$declare relnam pg_class.relname%TYPE; kur cursor for
> select c.relname from pg_class as c inner join pg_attribute as
> a on a.attrelid = c.oid where a.attname = 'some_id' and c.relkind
> = 'r';begin open kur; LOOP Fetch kur into relnam; exit when
> not found; EXECUTE 'ALTER TABLE '|| relnam|| ' add column
> another_id integer default -1'; END LOOP; close kur;END;$$
> I have 22 table names returned by the query used by cursor. Whe I run
> the function it executes for 1.5 minutes. Is there a way of doing it
> faster?

Two problems:
1. you're adding column with default value, which means it has to
rewrite the table, to set value for this column is all rows to -1.
2. you're adding column to 22 tables, in single transaction. this means
it will be prone to deadlocks.

it would be better to write simple shell script to add these columns,
one per connection, in parallel.

Best regards,

depesz

--
The best thing about modern society is how easy it is to avoid contact with it.
http://depesz.com/