IntegrityError 重复键值违反惟一约束-django/postgres

我正在跟进一个 我之前问的问题,在这个 我之前问的问题中,我试图寻求从一个愚蠢的/写得很差的 mysql 查询到 postgreql 的转换。我想我成功了。无论如何,我使用的数据是手动从 mysql 数据库转移到 postgres 数据库。我使用了一个看起来像这样的查询:

  UPDATE krypdos_coderound cru


set is_correct = case
when t.kv_values1 = t.kv_values2 then True
else False
end


from
  

(select cr.id,
array_agg(
case when kv1.code_round_id = cr.id
then kv1.option_id
else null end
) as kv_values1,


array_agg(
case when kv2.code_round_id = cr_m.id
then kv2.option_id
else null end
) as kv_values2


from krypdos_coderound cr
join krypdos_value kv1 on kv1.code_round_id = cr.id
join krypdos_coderound cr_m
on cr_m.object_id=cr.object_id
and cr_m.content_type_id =cr.content_type_id
join krypdos_value kv2 on kv2.code_round_id = cr_m.id


WHERE
cr.is_master= False
AND cr_m.is_master= True
AND cr.object_id=%s
AND cr.content_type_id=%s


GROUP BY cr.id
) t


where t.id = cru.id
""" % ( self.object_id, self.content_type.id)
)

我有理由相信这个方法很有效。然而,这导致了一个新的问题。在尝试提交时,我从 django 得到一个错误:

IntegrityError at (some url):
duplicate key value violates unique constraint "krypdos_value_pkey"

我在这里看了几个回复,我还没有找到解决我的问题的方法(虽然相关的问题已经作出了一些有趣的阅读)。我在日志中看到了这一点,这很有趣,因为我从未明确调用 insert-django must 处理它:

   STATEMENT:  INSERT INTO "krypdos_value" ("code_round_id", "variable_id", "option_id", "confidence", "freetext")
VALUES (1105935, 11, 55, NULL, E'')
RETURNING "krypdos_value"."id"

但是,尝试运行该命令会导致重复的键错误。

# Delete current coding
CodeRound.objects.filter(
object_id=o.id, content_type=object_type, is_master=True
).delete()
code_round = CodeRound(
object_id=o.id,
content_type=object_type,
coded_by=request.user, comments=request.POST.get('_comments',None),
is_master=True,
)
code_round.save()
for key in request.POST.keys():
if key[0] != '_' or key != 'csrfmiddlewaretoken':
options = request.POST.getlist(key)
for option in options:
Value(
code_round=code_round,
variable_id=key,
option_id=option,
confidence=request.POST.get('_confidence_'+key, None),
).save()  #This is where it dies
# Resave to set is_correct
code_round.save()
o.status = '3'
o.save()

我检查了序列之类的,它们似乎是有序的。在这一点上,我不知道该怎么办-我假设它的东西,姜戈的端,但我不确定。如有任何反馈将不胜感激!

69736 次浏览

If you have manually copied the databases, you may be running into the issue described here.

This happend to me - it turns out you need to resync your primary key fields in Postgres. The key is the SQL statement:

SELECT setval('tablename_id_seq', (SELECT MAX(id) FROM tablename)+1);

I encountered this error because I was passing extra arguments to the save method in the wrong way.

For anybody who encounters this, try forcing UPDATE with:

instance_name.save(..., force_update=True)

If you get an error that you cannot pass force_insert and force_update at the same time, you're probably passing some custom arguments the wrong way, like I did.

In addition to zapphods answer:

In my case the indexing was indeed incorrect, since I had deleted all migrations, and the database probably 10-15 times when developing as I wasn't in the stage of migrating anything.

I was getting an IntegrityError on finished_product_template_finishedproduct_pkey

Reindex the table and restart runserver:

I was using pgadmin3 and for whichever index was incorrect and throwing duplicate key errors I navigated to the constraints and reindexed.

enter image description here

And then reindexed.

enter image description here

It appears to be a known difference of behaviour between the MySQL and SQLite (they update the next available primary key even when inserting an object with an explicit id) backends, and other backends like Postgres, Oracle, ... (they do not).

There is a ticket describing the same issue. Even though it was closed as invalid, it provides a hint that there is a Django management command to update the next available key.

To display the SQL updating all next ids for the application MyApp:

python manage.py sqlsequencereset MyApp

In order to have the statement executed, you can provide it as the input for the dbshell management command. For bash, you could type:

python manage.py sqlsequencereset MyApp | python manage.py dbshell

The advantage of the management commands is that abstracts away the underlying DB backend, so it will work even if later migrating to a different backend.

I had an existing table in my "inventory" app and I wanted to add new records in Django admin and I got this error:

Duplicate key value violates unique constraint "inventory_part_pkey" DETAIL: Key (part_id)=(1) already exists.

As mentioned before, I run the code below to get the SQL command to reset the id-s:

python manage.py sqlsequencereset inventory

Piping the python manage.py sqlsequencereset inventory | python manage.py dbshell to the shell was not working

  • So I copied the generated raw SQL command
  • Then opened pgAdmin3 https://www.pgadmin.org for postgreSQL and opened my db
  • Clicked on the 6. icon (Execute arbitrary SQL queries)
  • Copied the statement what was generated

In my case the raw SQL command was:

BEGIN;
SELECT setval(pg_get_serial_sequence('"inventory_signup"','id'), coalesce(max("id"), 1), max("id") IS NOT null) FROM "inventory_signup";
SELECT setval(pg_get_serial_sequence('"inventory_supplier"','id'), coalesce(max("id"), 1), max("id") IS NOT null) FROM "inventory_supplier";
COMMIT;

Executed it with F5.

This fixed everything.

If you want to reset the PK on all of your tables, like me, you can use the PostgreSQL recommended way:

SELECT 'SELECT SETVAL(' ||
quote_literal(quote_ident(PGT.schemaname) || '.' || quote_ident(S.relname)) ||
', COALESCE(MAX(' ||quote_ident(C.attname)|| '), 1) ) FROM ' ||
quote_ident(PGT.schemaname)|| '.'||quote_ident(T.relname)|| ';'
FROM pg_class AS S,
pg_depend AS D,
pg_class AS T,
pg_attribute AS C,
pg_tables AS PGT
WHERE S.relkind = 'S'
AND S.oid = D.objid
AND D.refobjid = T.oid
AND D.refobjid = C.attrelid
AND D.refobjsubid = C.attnum
AND T.relname = PGT.tablename
ORDER BY S.relname;

After running this query, you will need to execute the results of the query. I typically copy and paste into Notepad. Then I find and replace "SELECT with SELECT and ;" with ;. I copy and paste into pgAdmin III and run the query. It resets all of the tables in the database. More "professional" instructions are provided at the link above.

The solution is that you need to resync your primary key fields as reported by "Hacking Life" who wrote an example SQL code but, as suggested by "Ad N" is better to run the Django command sqlsequencereset to get the exact SQL code that you can copy and past or run with another command.

As a further improvement to these answers I would suggest to you and other reader to dont' copy and paste the SQL code but, more safely, to execute the SQL query generated by sqlsequencereset from within your python code in this way (using the default database):

from django.core.management.color import no_style
from django.db import connection


from myapps.models import MyModel1, MyModel2




sequence_sql = connection.ops.sequence_reset_sql(no_style(), [MyModel1, MyModel2])
with connection.cursor() as cursor:
for sql in sequence_sql:
cursor.execute(sql)

I tested this code with Python3.6, Django 2.0 and PostgreSQL 10.

I was getting the same error as the OP.

I had created some Django models, created a Postgres table based on the models, and added some rows to the Postgres table via Django Admin. Then I fiddled with some of the columns in the models (changing around ForeignKeys, etc.) but had forgotten to migrate the changes.

Running the migration commands solved my problem, which makes sense given the SQL answers above.

To see what changes would be applied, without actually applying them:
python manage.py makemigrations --dry-run --verbosity 3

If you're happy with those changes, then run:
python manage.py makemigrations

Then run:
python manage.py migrate

I was getting a similar issue and nothing seemed to be working. If you need the data (ie cant exclude it when doing dump) make sure you have turned off (commented) any post_save receivers. I think the data would be imported but it would create the same model again because of these. Worked for me.

You just have to go to pgAdmin III and there execute your script with the name of the table:

SELECT setval('tablename_id_seq', (SELECT MAX(id) FROM tablename)+1);

Based on Paolo Melchiorre's answer, I wrote a chunk as a function to be called before any .save()

from django.db import connection
def setSqlCursor(db_table):
sql = """SELECT pg_catalog.setval(pg_get_serial_sequence('"""+db_table+"""', 'id'), MAX(id)) FROM """+db_table+""";"""
with connection.cursor() as cursor:
cursor.execute(sql)

This is the right statement. Mostly, It happens when we insert rows with id field.

SELECT setval('tablename_id_seq', (SELECT MAX(id) FROM tablename));

This question was asked about 9 years ago, and lots of people gave their own ways to solve it.

For me, I put unique=True in my email custom model field, but while creating superuser I didn't ask for the email to be mandatory.

Now after creating a superuser my email field is just saved as blank or Null. Now this is how I created and saved new user

obj = mymodel.objects.create_user(username='abc', password='abc')
obj.email = 'abc@abc.com'
obj.save()

It just threw the error saying duplicate-key-value-violates in the first line because the email was set to empty by default which was the same with the admin user. Django spotted a duplicate !!!

Solution

  • Option1: Make email mandatory while creating any user (for superuser as well)
  • Option2: Remove unique=True and run migrations
  • Option3: If you don't know where are the duplicates, you either drop the column or you can clear the database using python manage.py flush

It is highly recommended to know the reason why the error occurred in your case.