Adam Johnson

Yeah sparse results are difficult because there's no predicting when the density of results changes. What the current algorithm does is just react to this by decreasing the chunk size...

This is nice and looks good. A quick glance doesn't reveal anything problematic, but I'll give it a bit more testing during the week!

I've been meaning to review this but it's quite a big chunk of stuff to understand at once (and after this time, rebase). And I've never had the issue myself,...

N.B. https://github.com/simonpercivall/django-dbshell-plus performs the main use case I thought of (launching mycli for your db). However it woudl also be useful for sqlalchemy powered libraries like https://csvkit.readthedocs.org/en/0.9.1/scripts/sql2csv.html

The idea is to expand [`can_approx_count`](https://github.com/adamchainz/django-mysql/blob/2e4bb08488f737306f22b29ff217e2a33823d276/src/django_mysql/models/query.py#L589) to cover other situations, without hitting false positives. Looking again at it, I'm not sure how feasible it is, given the information we have...

Yes please send a PR! The tags are only loose and not up to date. Just beware I won't be able to review it for a couple weeks though. On...

Not gonna do this.

I've had a brief look. To be truly compatible we'd need to also add a check to the column, like `CHECK JSON_VALID(fieldname)`, so that inserting invalid JSON isn't possible -...

Triggers aren't quite the same as CHECK constraints, see https://mariadb.com/kb/en/mariadb/constraint/#check-constraint-expressions . Neither triggers nor CHECK constraints have any code representation in Django which is what makes them challenging to add.

@afausti I'm afraid not, as otherwise this ticket would have been closed :)