Hello Django community,
We are reaching out after encountering a persistent and elusive issue that
manifests as an IntegrityError during the save operation of a Django model.
This problem has been sporadically occurring since last year and has
successfully stumped four seasoned Django developers on our team. The error
seems to involve the non-update of auto_now fields upon model save, leading
us to suspect a deeper issue within Django, though we are cautious about
drawing premature conclusions. Given the complex and intermittent nature of
this bug, we are seeking insights, advice, or any form of guidance from the
community.
*Issue Overview:*
Our model inherits date_created and date_modified fields from an abstract
base class designed to standardize these timestamps across our models. Here
is how the abstract base class is defined:
class AbstractDateTimeModel(models.Model):
"""An abstract base class for most models"""
date_created = models.DateTimeField(
help_text="The moment when the item was created.",
auto_now_add=True,
db_index=True,
)
date_modified = models.DateTimeField(
help_text="The last moment when the item was modified. A value in
year"
" 1750 indicates the value is unknown",
auto_now=True,
db_index=True,
)
class Meta:
abstract = True
*The Problem:*
Intermittently, the .save() method triggers an IntegrityError, apparently
because the auto_now field (date_modified) does not get created. A specific
instance of this error showed that while date_created is correctly
populated, date_modified remains null, which directly leads to the
IntegrityError upon attempting to insert the record:
[datetime.datetime(2023, 10, 2, 14, 33, 49, 99833,
tzinfo=datetime.timezone.utc), None, ...]
'INSERT INTO "search_docket" ("date_created", "date_modified",... ]
*What We've Tried:*
- Investigated the possibility of an issue with update_fields being
non-null and excluding date_modified during .save(), but confirmed through
Sentry logs that update_fields was indeed None in all instances of the
error.
- Attempted to reproduce the issue in a controlled environment without
success, leaving us without a clear direction for a solution.
*Request for Help:*
We're wondering if this could point to an undocumented edge case in
Django's auto_now and auto_now_add implementation or a specific database
behavior under certain conditions. Any advice on further debugging steps,
experiences with similar issues, or knowledge of potential Django nuances
that we might not be considering would be incredibly valuable.
We appreciate your time and any feedback or suggestions you can offer.
Thanks
Bill
django v5.0.2
python 3.12.2
db is postgres
--
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 view this discussion on the web visit
https://groups.google.com/d/msgid/django-developers/a788c065-3a96-472c-9b41-7c2aebca6967n%40googlegroups.com.
