Opened 9 years ago
Closed 9 years ago
#25064 closed Bug (fixed)
Join.as_sql() emits invalid SQL if get_joining_columns() returns an empty tuple
| Reported by: | Alex Hill | Owned by: | nobody |
|---|---|---|---|
| Component: | Database layer (models, ORM) | Version: | 1.8 |
| Severity: | Normal | Keywords: | |
| Cc: | Triage Stage: | Accepted | |
| Has patch: | yes | Needs documentation: | no |
| Needs tests: | no | Patch needs improvement: | yes |
| Easy pickings: | no | UI/UX: | no |
Description (last modified by )
Join.as_sql() gets its join conditions from two places: the get_joining_columns() and get_extra_restrictions().
The SQL for extra restrictions is unconditionally prefixed with AND, which means invalid SQL is emitted if no joining columns are specified. The fix to make the AND conditional is a one-liner and allows users to override the default join conditions instead of adding to them.
Change History (11)
comment:1 by , 9 years ago
| Description: | modified (diff) |
|---|---|
| Summary: | Allow get_joining_columns to return empty tuple → Join.as_sql() emits invalid SQL if get_joining_columns() returns an empty tuple |
comment:2 by , 9 years ago
comment:4 by , 9 years ago
| Has patch: | set |
|---|
Here's the patch and tests: https://github.com/django/django/pull/4978
The patch is against master, but I would love to get this fixed in 1.8 and would be happy to backport it.
comment:5 by , 9 years ago
| Triage Stage: | Unreviewed → Accepted |
|---|
I don't think this should be backpatched as this isn't an omission in a new feature (the Join type is completely internal, so its addition doesn't count as a new feature).
Otherwise I don't see any harm in doing the changes.
comment:6 by , 9 years ago
| Patch needs improvement: | set |
|---|
Patch to Join seems OK to me, but the tests need some improvement.
comment:7 by , 9 years ago
Thanks. Fixed the tests in line with your feedback, rebased and squashed.
comment:8 by , 9 years ago
| Patch needs improvement: | unset |
|---|
I've changed the example used in the tests which should hopefully clarify things a bit, and made the recommended style edits.
comment:9 by , 9 years ago
| Triage Stage: | Accepted → Ready for checkin |
|---|
comment:10 by , 9 years ago
| Patch needs improvement: | set |
|---|---|
| Triage Stage: | Ready for checkin → Accepted |
Marked RFC too soon as there is a test failure on Oracle.
Is a patch forthcoming? If not, details on how to reproduce the issue will be helpful.