On Wed, Aug 13, 2008 at 1:49 PM, Yi Zhao <yi.zhao at alibaba-inc.com> wrote:
> I have a table which have unique like:
> unique(foo_id, bar_id);
>
> my slonik script is below:
> set add table (set id = 1, origin = 1, id = 2, full qualified name =
> 'search.thread', key = 'foo_id, bar_id', comment = 'nodesc');
>
> when I execute:
> <stdin>:16: PGRES_FATAL_ERROR select
> "_replication".determineIdxnameUnique('search.thread', 'foo_id,
> bar_id');  - ERROR:  Slony-I: table "search"."thread" has no unique
> index foo_id, bar_id
>
> how to add a table which have unique with two column into slony???

The unique constraint on your table will make use of an index.
You should specify the name of this index in the 'key' parameter, not
the name of the columns in the constraint definition.
Like:

postgres=# create table table1 (foo_id int, bar_id int);
CREATE TABLE
postgres=# alter table table1 add constraint unique_idx unique (foo_id,bar_id);
NOTICE:  ALTER TABLE / ADD UNIQUE will create implicit index
"unique_idx" for table "table1"
ALTER TABLE
postgres=# \d table1
    Table "public.table1"
 Column |  Type   | Modifiers
--------+---------+-----------
 foo_id | integer |
 bar_id | integer |
Indexes:
    "unique_idx" UNIQUE, btree (foo_id, bar_id)

Then your slonik command should look like:

set add table (set id = 1, origin = 1, id = 2, full qualified name =
'search.thread', key = 'unique_idx' comment = 'nodesc');

Note that you also need to have NOT NULL on all the columns composing
the index. As the manual says, you might as well define a primary key
consisting of these columns.

Thanks,

Charles Duffy