Обсуждение: Upsert with a partial unique index constraint violation
Hi all,
I'm struggling to work out the correct way to use a partial unique index in an upsert context.
Here's the context: I'm creating a partial unique index on a table:
tmp=# CREATE TABLE foo (a INT NOT NULL, b int NOT NULL, c TEXT, d BOOLEAN DEFAULT false);
CREATE TABLE
tmp=# CREATE UNIQUE INDEX foo_unique_true ON foo (a, b) WHERE d = true;
CREATE INDEX
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'one', false);
INSERT 0 1
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'two', false);
INSERT 0 1
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'three', true);
INSERT 0 1
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true);
ERROR: duplicate key value violates unique constraint "foo_unique_true"
This is all working as expected.
What I want to do is alter my INSERT to be an upsert for the violation of the index constraint foo_unique_true. However, I cannot work out the syntax for doing this as the partial index is not a constraint, and my interpretation of index_expression and index_predicate in the grammar[1] don't seem to be working:
First, trying to upsert as if foo_unique_true was a constraint:
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true)
tmp-# ON CONFLICT ON CONSTRAINT "foo_unique_true"
tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d = true;
ERROR: constraint "foo_unique_true" for table "foo" does not exist
Second, trying with index_expression and index_predicate:
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true)
tmp-# ON CONFLICT (a, b) WHERE d = true
tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d = true;
ERROR: there is no unique or exclusion constraint matching the ON CONFLICT specification
If anyone knows what I'm doing wrong and how to get this to work, or knows that this is not possible to achieve, I'm all ears.
Cheers,
Tim
On Mon, Jul 11, 2016 at 12:06 AM, Tim Dawborn <tim.dawborn@gmail.com> wrote: > tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true) > tmp-# ON CONFLICT (a, b) WHERE d = true > tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d = > true; > ERROR: there is no unique or exclusion constraint matching the ON CONFLICT > specification > > If anyone knows what I'm doing wrong and how to get this to work, or knows > that this is not possible to achieve, I'm all ears. That should work. Are you sure you haven't spelled it "... WHERE d IS TRUE"? -- Peter Geoghegan
Peter Geoghegan <pg@bowt.ie> writes:
> On Mon, Jul 11, 2016 at 12:06 AM, Tim Dawborn <tim.dawborn@gmail.com> wrote:
>> tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true)
>> tmp-# ON CONFLICT (a, b) WHERE d = true
>> tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d =
>> true;
>> ERROR: there is no unique or exclusion constraint matching the ON CONFLICT
>> specification
>>
>> If anyone knows what I'm doing wrong and how to get this to work, or knows
>> that this is not possible to achieve, I'm all ears.
> That should work. Are you sure you haven't spelled it "... WHERE d IS TRUE"?
It does work for me, but I think it probably only started working after
this as-yet-unreleased patch:
Author: Tom Lane <tgl@sss.pgh.pa.us>
Branch: master [26e66184d] 2016-05-11 16:20:23 -0400
Branch: REL9_5_STABLE [58d802410] 2016-05-11 16:20:03 -0400
Fix assorted missing infrastructure for ON CONFLICT.
subquery_planner() failed to apply expression preprocessing to the
arbiterElems and arbiterWhere fields of an OnConflictExpr. No doubt the
theory was that this wasn't necessary because we don't actually try to
execute those expressions; but that's wrong, because it results in failure
to match to index expressions or index predicates that are changed at all
by preprocessing. Per bug #14132 from Reynold Smith.
The key point here being that "WHERE boolvar = true" will be simplified
to "WHERE boolvar" by preprocessing, and you don't get a match unless
that happened on both expressions. Tim could work around this in
unpatched releases by spelling the predicate as just "d".
regards, tom lane
Correct, there was no typo there. All of the psql examples I included were copy-pasted out of a clean psql 9.5 session on a clean psql 9.5 database (64 bit linux).
$ createdb tmp
$ psql --quiet tmp
tmp=# select version();
version
-------------------------------------------------------------------------------------------------
PostgreSQL 9.5.3 on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 4.8.2-19ubuntu1) 4.8.2, 64-bit
(1 row)
tmp=# CREATE TABLE foo (a INT NOT NULL, b INT NOT NULL, c TEXT, d BOOLEAN DEFAULT false);
tmp=# CREATE UNIQUE INDEX foo_unique_true ON foo (a, b) WHERE d = true;
tmp=# \d+ foo
Table "public.foo"
Column | Type | Modifiers | Storage | Stats target | Description
--------+---------+---------------+----------+--------------+-------------
a | integer | not null | plain | |
b | integer | not null | plain | |
c | text | | extended | |
d | boolean | default false | plain | |
Indexes:
"foo_unique_true" UNIQUE, btree (a, b) WHERE d = true
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'one', false);
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'two', false);
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'three', true);
tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true) ON CONFLICT (a, b) WHERE d = true DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 and foo.d = true;
ERROR: there is no unique or exclusion constraint matching the ON CONFLICT specification
On 12 July 2016 at 13:43, Peter Geoghegan <pg@bowt.ie> wrote:
On Mon, Jul 11, 2016 at 12:06 AM, Tim Dawborn <tim.dawborn@gmail.com> wrote:
> tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true)
> tmp-# ON CONFLICT (a, b) WHERE d = true
> tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d =
> true;
> ERROR: there is no unique or exclusion constraint matching the ON CONFLICT
> specification
>
> If anyone knows what I'm doing wrong and how to get this to work, or knows
> that this is not possible to achieve, I'm all ears.
That should work. Are you sure you haven't spelled it "... WHERE d IS TRUE"?
--
Peter Geoghegan
Awesome. Thanks, Tom. Glad to see this issue has been patched upstream.
I'll use the alternative syntax in the meantime.
Cheers,
Tim
On 13 July 2016 at 01:03, Tom Lane <tgl@sss.pgh.pa.us> wrote:
Peter Geoghegan <pg@bowt.ie> writes:
> On Mon, Jul 11, 2016 at 12:06 AM, Tim Dawborn <tim.dawborn@gmail.com> wrote:
>> tmp=# INSERT INTO foo (a, b, c, d) VALUES (1, 2, 'four', true)
>> tmp-# ON CONFLICT (a, b) WHERE d = true
>> tmp-# DO UPDATE SET c = 'four' WHERE foo.a = 1 AND foo.b = 2 AND foo.d =
>> true;
>> ERROR: there is no unique or exclusion constraint matching the ON CONFLICT
>> specification
>>
>> If anyone knows what I'm doing wrong and how to get this to work, or knows
>> that this is not possible to achieve, I'm all ears.
> That should work. Are you sure you haven't spelled it "... WHERE d IS TRUE"?
It does work for me, but I think it probably only started working after
this as-yet-unreleased patch:
Author: Tom Lane <tgl@sss.pgh.pa.us>
Branch: master [26e66184d] 2016-05-11 16:20:23 -0400
Branch: REL9_5_STABLE [58d802410] 2016-05-11 16:20:03 -0400
Fix assorted missing infrastructure for ON CONFLICT.
subquery_planner() failed to apply expression preprocessing to the
arbiterElems and arbiterWhere fields of an OnConflictExpr. No doubt the
theory was that this wasn't necessary because we don't actually try to
execute those expressions; but that's wrong, because it results in failure
to match to index expressions or index predicates that are changed at all
by preprocessing. Per bug #14132 from Reynold Smith.
The key point here being that "WHERE boolvar = true" will be simplified
to "WHERE boolvar" by preprocessing, and you don't get a match unless
that happened on both expressions. Tim could work around this in
unpatched releases by spelling the predicate as just "d".
regards, tom lane