You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I don't know yet Propel details.
The idea is just that some models can have a default ordering (I assume this functionality is present in Propel also).
And that some ORM may still add this ordering in the SQL query whilst it is not needed because only a single object will be prefetched per instance of the original result set.
On very big prefetches I noticed a gain of 10 - 15 %,
since the rest of the DB query is mainly an index scan.
#ClimateChangeBrake
Can someone please dump the queries of such prefetches and see if the bug is present in Propel ?
If so submitting a PR should not be too hard.
I will try to help.
But right now, I start by warning the frameworks/ORM I am aware of :)
Best regards,
Laurent Lyaudet
The text was updated successfully, but these errors were encountered:
Hello,
Recently I found a problem with Django and submitted a patch in a PR:
https://code.djangoproject.com/ticket/35309
django/django#17984
I also developped a work-around for old versions of Django:
https://github.com/LLyaudet/django-monkey-patches/blob/main/src/django_monkey_patches/django__orm__prefetch_without_useless_order_by.py
I don't know yet Propel details.
The idea is just that some models can have a default ordering (I assume this functionality is present in Propel also).
And that some ORM may still add this ordering in the SQL query whilst it is not needed because only a single object will be prefetched per instance of the original result set.
On very big prefetches I noticed a gain of 10 - 15 %,
since the rest of the DB query is mainly an index scan.
#ClimateChangeBrake
Can someone please dump the queries of such prefetches and see if the bug is present in Propel ?
If so submitting a PR should not be too hard.
I will try to help.
But right now, I start by warning the frameworks/ORM I am aware of :)
Best regards,
Laurent Lyaudet
The text was updated successfully, but these errors were encountered: