#2108 closed defect (fixed)
Empty models result in invalid SQL for Sqlite
Reported by: | Owned by: | Malcolm Tredinnick | |
---|---|---|---|
Component: | Database layer (models, ORM) | Version: | |
Severity: | normal | Keywords: | sqlite |
Cc: | Triage Stage: | Unreviewed | |
Has patch: | no | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
A model with no fields, e.g.
class EmptyModel(models.Model):
pass
will cause an OperationalError to be raised when an instance is saved:
m = EmptyModel()
m.save()
if the database is sqlite3.
In this case the SQL executed is:
INSERT INTO "testapp_emptymodel" () VALUES ()
It should instead be generating the following SQL:
INSERT INTO "testapp_emptymodel" (id) VALUES (NULL)
Change History (7)
comment:1 by , 18 years ago
comment:2 by , 18 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
comment:3 by , 18 years ago
Resolution: | fixed |
---|---|
Status: | closed → reopened |
Inserting NULL into an automatically-incremented field is a common idiom that works in sqlite, mysql, and sql server at least.
Regardless, your solution -- to not save anything -- is wrong. There are cases where a model might have no fields but an id. The one I've got at the moment is a shopping basket, which at the moment does not refer to any other models -- but other models refer to it. If the basket silently fails to be saved, then the other model objects will also fail, all without a clear reason.
Either Django should support tables with just an id field, or it should throw a ValidationError or some such. Silently failing is not good.
comment:4 by , 18 years ago
Owner: | changed from | to
---|---|
Status: | reopened → new |
I understand your use case now. Sounds reasonable, although a bit of an unusual case.
Inserting NULL in that field is, as a rule, not correct -- it is invalid SQL and it's sort of a shame that MySQL and SQL server permit it. PostgreSQL rejects it. The main problem is SQLite. We should be executing insert into mytable (id) values (DEFAULT)
but SQLite does not understand the standard DEFAULT value, so you must use NULL in that case for integer primary keys.
I have made a patch that now inserts the new elements each time and has a backend-specific hook to get the right value to insert (either DEFAULT or NULL, as appropriate). Will commit it in a moment and get somebody on the mailing list to test the SQL Server case.
comment:5 by , 18 years ago
Status: | new → assigned |
---|
comment:6 by , 18 years ago
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
This is indeed a bug, but the proposed solution is not correct. A primary key in a database is unique and not NULL, so trying to set id to be NULL is not a good idea, even if it might happen to work in SQLite. We should be not saving anything in this case.