Обсуждение: Insert in table with UNIQUE index
Hi All!
I have a table with unique index with 2 exactly the same rows.
How it can be possible?
CREATE TABLE _inforgchngr6716_test
(
_nodetref bytea NOT NULL,
_noderref bytea NOT NULL,
_messageno numeric(10,0)
)
WITH (
OIDS=FALSE
);
ALTER TABLE _inforgchngr6716_test
OWNER TO postgres;
-- Index: _inforg6716_bynodemsg_rn_test
-- DROP INDEX _inforg6716_bynodemsg_rn_test;
CREATE UNIQUE INDEX _inforg6716_bynodemsg_rn_test
ON _inforgchngr6716_test
USING btree
(_nodetref, _noderref, _messageno);
Artem Tomyuk wrote: > I have a table with unique index with 2 exactly the same rows. > How it can be possible? > > > CREATE TABLE _inforgchngr6716_test > ( > _nodetref bytea NOT NULL, > _noderref bytea NOT NULL, > _messageno numeric(10,0) > ) > WITH ( > OIDS=FALSE > ); > ALTER TABLE _inforgchngr6716_test > OWNER TO postgres; > > -- Index: _inforg6716_bynodemsg_rn_test > > -- DROP INDEX _inforg6716_bynodemsg_rn_test; > > CREATE UNIQUE INDEX _inforg6716_bynodemsg_rn_test > ON _inforgchngr6716_test > USING btree > (_nodetref, _noderref, _messageno); Maybe index corruption. Did you have any crashes? Do you get an error when you REINDEX INDEX _inforg6716_bynodemsg_rn_test; Yours, Laurenz Albe
<div>Hi Artem,</div><div> </div><div>You can see exact same record in your select results from _infogcngr6716_test
tabledespite uniq index exists, if you create a table inherited from the table.</div><div> </div><div>For
example:</div><div>Createinherited table</div><div><span style="font-family:terminal,monaco;font-size:x-small;">CREATE
TABLEpublic._inforgcngr6716_test_child</span><br /><span
style="font-family:terminal,monaco;font-size:x-small;">(</span><br/><span
style="font-family:terminal,monaco;font-size:x-small;"> LIKE public._inforgchngr6716_test INCLUDING DEFAULTS INCLUDING
CONSTRAINTSINCLUDING INDEXES</span><br /><span style="font-family:terminal,monaco;font-size:x-small;">) </span><br
/><spanstyle="font-family:terminal,monaco;font-size:x-small;">INHERITS (_inforgchngr6716_test)</span><br /><span
style="font-family:terminal,monaco;font-size:x-small;">WITH(</span><br /><span
style="font-family:terminal,monaco;font-size:x-small;"> OIDS = FALSE</span><br /><span
style="font-family:terminal,monaco;font-size:x-small;">);</span></div><div> </div><div>Insertsame records into these
twotable. (No error returns)</div><div><span style="font-family:terminal,monaco;font-size:x-small;">INSERT INTO
_infogcngr6716_test_child(</span><br/><span style="font-family:terminal,monaco;font-size:x-small;">
_nodetref,_noderref, _messageno)</span><br /><span style="font-family:terminal,monaco;font-size:x-small;"> VALUES
('asd','asd',10);</span><br /><br /><span style="font-family:terminal,monaco;font-size:x-small;">INSERT INTO
_infogcngr6716_test(</span><br/><span style="font-family:terminal,monaco;font-size:x-small;"> _nodetref,
_noderref,_messageno)</span><br /><span style="font-family:terminal,monaco;font-size:x-small;"> VALUES ('asd',
'asd',10);</span></div><div> </div><div>Make select query to parent table</div><div><span
style="font-family:terminal,monaco;font-size:x-small;">SELECT_nodetref, _noderref, _messageno</span><br /><span
style="font-family:terminal,monaco;font-size:x-small;"> FROM _inforgchngr6716_test;</span><br /><br
/></div><div>"asd";"asd";10<br/>"asd";"asd";10<br /><br /></div><div><div>Make select query to only parent
table</div><spanstyle="font-family:terminal,monaco;font-size:x-small;">SELECT _nodetref, _noderref,
_messageno</span><br/><span style="font-family:terminal,monaco;font-size:x-small;"> FROM ONLY
_inforgchngr6716_test;</span></div><div> </div><div>"asd";"asd";10</div><div> </div><div>İyiçalışmalar.</div><div>Samed
YILDIRIM</div><div> </div><div> </div><div> </div><div>27.01.2016,16:14, "Albe Laurenz"
<laurenz.albe@wien.gv.at>:</div><blockquotetype="cite"><p>Artem Tomyuk wrote:<blockquote> I have a table with
uniqueindex with 2 exactly the same rows.<br /> How it can be possible?<br /><br /><br /> CREATE TABLE
_inforgchngr6716_test<br/> (<br /> _nodetref bytea NOT NULL,<br /> _noderref bytea NOT NULL,<br /> _messageno
numeric(10,0)<br/> )<br /> WITH (<br /> OIDS=FALSE<br /> );<br /> ALTER TABLE _inforgchngr6716_test<br /> OWNER TO
postgres;<br/><br /> -- Index: _inforg6716_bynodemsg_rn_test<br /><br /> -- DROP INDEX
_inforg6716_bynodemsg_rn_test;<br/><br /> CREATE UNIQUE INDEX _inforg6716_bynodemsg_rn_test<br /> ON
_inforgchngr6716_test<br/> USING btree<br /> (_nodetref, _noderref, _messageno);</blockquote><p><br />Maybe index
corruption.<br/>Did you have any crashes?<br /><br />Do you get an error when you<br /> REINDEX INDEX
_inforg6716_bynodemsg_rn_test;<br/><br />Yours,<br />Laurenz Albe<br /><br /><span>-- <br />Sent via pgsql-admin
mailinglist (<a href="mailto:pgsql-admin@postgresql.org">pgsql-admin@postgresql.org</a>)<br />To make changes to your
subscription:<br/><a
href="http://www.postgresql.org/mailpref/pgsql-admin">http://www.postgresql.org/mailpref/pgsql-admin</a><br
/></span></blockquote>
Hi Artem,You can see exact same record in your select results from _infogcngr6716_test table despite uniq index exists, if you create a table inherited from the table.
So what? That is not the problem presented.
David J.
On Wed, Jan 27, 2016 at 5:30 AM, Artem Tomyuk <admin@leboutique.com> wrote:
I have a table with unique index with 2 exactly the same rows.How it can be possible?CREATE TABLE _inforgchngr6716_test(_nodetref bytea NOT NULL,_noderref bytea NOT NULL,_messageno numeric(10,0))CREATE UNIQUE INDEX _inforg6716_bynodemsg_rn_testON _inforgchngr6716_testUSING btree(_nodetref, _noderref, _messageno);
In the duplicated entries, are any of the "_messageno" values NULL? If so, that would explain it. NULL is not bound by the constraint, you have a few options: make it NOT NULL, COALESCE() it to a known value, or create a couple partial indexes to enforce the uniqueness.
El mié, 27-01-2016 a las 14:12 +0000, Albe Laurenz escribió: > Artem Tomyuk wrote: > > I have a table with unique index with 2 exactly the same rows. > > How it can be possible? > > > > > > CREATE TABLE _inforgchngr6716_test > > ( > > _nodetref bytea NOT NULL, > > _noderref bytea NOT NULL, > > _messageno numeric(10,0) > > ) > > WITH ( > > OIDS=FALSE > > ); > > ALTER TABLE _inforgchngr6716_test > > OWNER TO postgres; > > > > -- Index: _inforg6716_bynodemsg_rn_test > > > > -- DROP INDEX _inforg6716_bynodemsg_rn_test; > > > > CREATE UNIQUE INDEX _inforg6716_bynodemsg_rn_test > > ON _inforgchngr6716_test > > USING btree > > (_nodetref, _noderref, _messageno); > > Maybe index corruption. > Did you have any crashes? To discard a index corruption problem, try to select duplicates row setting enable_indexscan and enable_bitmapscan to false and check if you get 1 or 2 rows, if you get 2 -> I'll follow Albe advide and reindex > > Do you get an error when you > REINDEX INDEX _inforg6716_bynodemsg_rn_test; > > Yours, > Laurenz Albe >