Re: 1.10 beta 1

Ashesh Vashi a écrit : 

Hi Guillaume,

Thanks for reviewing the patch.

Guillaume Lelarge wrote:   

Ashesh Vashi a écrit :     

Hi Dave,

Please find the patch for the Dependency bug       

I just checked this patch. I find disturbing that it finds a dependency
to a nextval function instead of a dependency to the sequence. Moreover,
the first table has two dependencies : on the sequence and on the
nextval function of this sequence.     

I think - you're right.
But, I could not find the direct dependency of the sequence on the table.   

 This patch is a first step, but I think we need to go further.     

Definitely, even I felt the same.
I need some guidance for it.

Could you please help on this?
   

Here are the definition of the two tables:

CREATE TABLE t1 (id serial NOT NULL);
CREATE TABLE t2 (LIKE t INCLUDING DEFAULTS);

Quite simple. t1_id_seq must be a dependency of t2 (ie you can't drop
t1_id_seq, or t1, without dropping t2 or at least change its sequence).

With this next query, we get every pg_attrdef dependencies:

ioguix=# select * from pg_depend where classid = 2604;classid | objid  | objsubid | refclassid | refobjid | refobjsubid | deptype
---------+--------+----------+------------+----------+-------------+---------   2604 | 122030 |        0 |       1259 |   122027 |           1 | a   2604 | 122030 |        0 |       1259 |   122025 |           0 | n   2604 | 122034 |        0 |       1259 |   122031 |           1 | a   2604 | 122034 |        0 |       1259 |   122025 |           0 | n
(4 lignes)

This one gives us the target column (122030 and 122034 are results of
the previous query, column objid) :

ioguix=# select relname||'.'||attname from pg_class cl, pg_attribute att, pg_attrdef atd where atd.oid in (122030, 122034)   and atd.adrelid=att.attrelid   and atd.adnum=att.attnum   and cl.oid=att.attrelid;?column?
----------t1.idt2.id
(2 lignes)

This one gives us the referenced object (122027, 122025, 122031 are
results of the attrdef query, column refobjid) :

ioguix=# select oid, relname from pg_class where oid in (122027, 122025,
122031); oid   |  relname
--------+-----------122025 | t1_id_seq122027 | t1122031 | t2
(3 lignes)

In the first query (the attrdef one), you'll notice t1_id_seq appears
two times, one for t1.id and one for t2.id.

I don't quite know how to build a query that will give us t1_id_seq as a
dependency to t2.

I don't actually have the time to work on this right now. But it must be
something with these queries.

Hope it helps. Ping me if you need more details.