Aggregating an annotated field does not quote the alias column name

[Michael Manfre]

Given the below code snippet from aggregation.BaseAggregateTestCase.test_aggregate_annotation:

vals = Book.objects.annotate(num_authors=Count("authors__id")).aggregate(Avg("num_authors"))

The generated SQL does not quote the annotated field "num_authors" when used with the aggregate function. Generated SQL is of the form:

SELECT AVG(num_authors) FROM (... subquery ...)

[Michael Manfre]

I'm not aware of any databases that need to quote '*', but just in case there are I provided two patches. The second preloads the cache with '*' to prevent quoting it. Most commonly encountered by COUNT(*). Regardless of which patch is chosen (if any), 3rd party backends can override quote_cache to either have or omit this initial alias.

[Luke Plant]

Is there a bug caused by this behaviour? I can't see it at the moment.

[Anssi Kääriäinen]

Seems like if you use a SQL reserved keyword as the key in annotate, then you will get a broken query. No testing done here.

[Luke Plant]

Ah, I see, I was just thinking about characters that can't be part of keyword arguments anyway, so thought there wasn't an issue.

[Michael Manfre]

Updated pull request with test. ​https://github.com/django/django/pull/1674

[Tim Graham <timograham@…>]

In 9a041807fcde7ff7245696a743196d31ec7c7b5c:

Fixed #18333 - Quoted annotated column names

[Anssi Kääriäinen <akaariai@…>]

In 9027da65d3590a3bd319490d78c86ef09cd04f9e:

Added '*' to quote_cache

The commit for #18333 missed quote_cache default value for *.
Refs #18333.