Your idea about ignoring conflicts in bulk_create is a great idea, i've
made a merge request that attempts to implements this. If anyone has any
time to review it it would be much appreciated.

I couldn't find a very elegant way of adding it though - Sqlite and mysql
need something added just after the INSERT, whereas postgres needs
something added after the values but before the RETURNING statement. Oracle
doesn't support this as far as I'm aware.

> Qui tacet consentire videtur.

Silence is rarely an agreement on a mailing list, certainly not regarding a
proposed new feature or potentially breaking change. Code speaks louder
than words though :)

On 3 Oct 2017 20:17, "Дилян Палаузов" <dpa-dja...@aegee.org> wrote:

Hello,

Qui tacet consentire videtur.

I agree there is still a concurrency problem with the approach I
suggested.  INSERT ON CONFLICT DO UPDATE might not work, if the
simultaneously inserted object has for some reason other default values,
that the current object, so that DO UPDATE will overwrite the data.  I am
not talking about other backends.

Another idea to start with:

1. Add on_conflict=IGNORE parameter to bulk_create() -
https://code.djangoproject.com/ticket/28668

The problem is that in order to insert a lot of data in a database fast, a
single INSERT must be sent, but it fails if any of the records were already
in the database, so if this could happen, one has to iterate over the input
and do a separate INSERT for each object, but this is slower than a single
INSERT.

However with something like INSERT ... ON CONFLICT DO NOTHING (which varies
in the different RDBMs) the database is instructed to absorb all the data
that is not already there.

2. Afterwards extend bulk_create(..., on_conflict=IGNORE) to detect for
Postgresql which objects were actually added, as described at
https://groups.google.com/forum/#!topic/django-developers/wdHIYdQHO_0 , and
return only those.

Regards
  Дилян


On 10/03/2017 06:00 PM, Aymeric Augustin wrote:

> Hello,
>
> Since I haven't seen positive feedback from a committer, I'm not convinced
> there's consensus about this change.
>
> Also this doesn't look like a particularly easy topic for a beginner. It
> raises the following questions:
>
> - INSERT ON CONFLICT UPDATE would likely be a better option on Postgres ≥
> 9.5, wouldn't it?
> - what about other databases with built-in backends? third-party backends?
> - is this implementation really appropriate at any isolation level?
> - what are the consequences for performance?
>
> Best regards,
>
> --
> Aymeric.
>
>
> 2017-10-03 17:23 GMT+02:00 Дилян Палаузов <dpa-dja...@aegee.org <mailto:
> dpa-dja...@aegee.org>>:
>
>
>     Hello Muhereza,
>
>     I assume you understand by now Django quite well and are willing to
>     give something "in return".
>
>     Currently QuerySet.get_or_create() consists of two SQL commands:
>     SELECT + optional INSERT.  They cause a concurrent problem, if
>     another get_or_create() is called between the SQL statements.
>
>     With the Postgresql backend it is possible to reduce the queries to
>     a single one.
>
>     Consider this table:
>
>     CREATE TABLE t (
>         id SERIAL PRIMARY KEY,
>         name VARCHAR(10) NOT NULL UNIQUE,
>         comment VARCHAR(10));
>
>     The following query can do what get_or_create() currently achieves:
>
>     WITH
>        maybe_found AS (SELECT * FROM t WHERE t.name
>     <http://t.name>='nameD'),
>
>        to_be_inserted AS (SELECT 'nameD' as "name", 'comment13' as
>     "comment"),
>        just_inserted AS (
>               INSERT INTO t (name, comment) SELECT * FROM to_be_inserted
>                              WHERE NOT EXISTS(SELECT * FROM maybe_found)
>               RETURNING *)
>     SELECT *, FALSE as "created" FROM maybe_found UNION ALL
>     SELECT *, TRUE AS "created" FROM just_inserted LIMIT 2;
>
>     where "to_be_inserted' contains the values for the new object
>     ('default' parameter of get_or_create) and 'nameB' in maybe_found is
>     the criterion passed to get().
>
>     The challenge is to integrate the WITH ... SELECT query in Django.
>  As guidance I can only suggest looking at the existing code.
>
>     Regards
>         Дилян
>
>
>     On 10/03/2017 10:24 AM, Muhereza Herman wrote:
>
>         Hello, anyone reading this please help me out.
>         am a new developer but i would like to contribute to_*django*_.
>         please guide me on how to do that. thank you.
>
>         --         You received this message because you are subscribed to
> the
>         Google Groups "Django developers (Contributions to Django
>         itself)" group.
>         To unsubscribe from this group and stop receiving emails from
>         it, send an email to
>         django-developers+unsubscr...@googlegroups.com
>         <mailto:django-developers%2bunsubscr...@googlegroups.com>
>         <mailto:django-developers+unsubscr...@googlegroups.com
>         <mailto:django-developers%2bunsubscr...@googlegroups.com>>.
>
>         To post to this group, send email to
>         django-developers@googlegroups.com
>         <mailto:django-developers@googlegroups.com>
>         <mailto:django-developers@googlegroups.com
>
>         <mailto:django-developers@googlegroups.com>>.
>         Visit this group at
>         https://groups.google.com/group/django-developers
>         <https://groups.google.com/group/django-developers>.
>
>         To view this discussion on the web visit
>         https://groups.google.com/d/msgid/django-developers/e6d798ac
> -4ede-45c4-9f20-ca62f0595131%40googlegroups.com
>         <https://groups.google.com/d/msgid/django-developers/e6d798a
> c-4ede-45c4-9f20-ca62f0595131%40googlegroups.com>
>         <https://groups.google.com/d/msgid/django-developers/e6d798a
> c-4ede-45c4-9f20-ca62f0595131%40googlegroups.com?utm_medium=
> email&utm_source=footer
>
>         <https://groups.google.com/d/msgid/django-developers/e6d798a
> c-4ede-45c4-9f20-ca62f0595131%40googlegroups.com?utm_medium=
> email&utm_source=footer>>.
>         For more options, visit https://groups.google.com/d/optout
>         <https://groups.google.com/d/optout>.
>
>
>
>     --     You received this message because you are subscribed to the
> Google
>     Groups "Django developers  (Contributions to Django itself)" group.
>     To unsubscribe from this group and stop receiving emails from it,
>     send an email to django-developers+unsubscr...@googlegroups.com
>     <mailto:django-developers%2bunsubscr...@googlegroups.com>.
>
>     To post to this group, send email to
>     django-developers@googlegroups.com
>     <mailto:django-developers@googlegroups.com>.
>     Visit this group at
>     https://groups.google.com/group/django-developers
>     <https://groups.google.com/group/django-developers>.
>
>     To view this discussion on the web visit
>     https://groups.google.com/d/msgid/django-developers/062e270c
> -85f6-23a5-64a3-b225ddcff3ab%40aegee.org
>     <https://groups.google.com/d/msgid/django-developers/062e270
> c-85f6-23a5-64a3-b225ddcff3ab%40aegee.org>.
>
>
>     For more options, visit https://groups.google.com/d/optout
>     <https://groups.google.com/d/optout>.
>
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "Django developers (Contributions to Django itself)" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to django-developers+unsubscr...@googlegroups.com <mailto:
> django-developers+unsubscr...@googlegroups.com>.
> To post to this group, send email to django-developers@googlegroups.com
> <mailto:django-developers@googlegroups.com>.
> Visit this group at https://groups.google.com/group/django-developers.
> To view this discussion on the web visit https://groups.google.com/d/ms
> gid/django-developers/CANE-7mVgudhd0wDAvX9_uLxFUJ89w2Lge6-
> 0rOh_U3hPCc4WnA%40mail.gmail.com <https://groups.google.com/d/m
> sgid/django-developers/CANE-7mVgudhd0wDAvX9_uLxFUJ89w2Lge6-
> 0rOh_U3hPCc4WnA%40mail.gmail.com?utm_medium=email&utm_source=footer>.
>
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups
"Django developers  (Contributions to Django itself)" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to django-developers+unsubscr...@googlegroups.com.
To post to this group, send email to django-developers@googlegroups.com.
Visit this group at https://groups.google.com/group/django-developers.
To view this discussion on the web visit https://groups.google.com/d/ms
gid/django-developers/9182fb5c-4410-b28d-9e30-8d9b295ea22d%40aegee.org.

For more options, visit https://groups.google.com/d/optout.

-- 
You received this message because you are subscribed to the Google Groups 
"Django developers  (Contributions to Django itself)" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to django-developers+unsubscr...@googlegroups.com.
To post to this group, send email to django-developers@googlegroups.com.
Visit this group at https://groups.google.com/group/django-developers.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/django-developers/CAFNZOJOpP39_-4PEUujz7FTdhERgjLGJmO8zC-BhMb1SFZUBQA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to