Issue 40784: test_sqlite: CheckFuncDeterministic() fails with SQLite 3.32

With SQLite 3.32, test_sqlite fails with:

FAIL: CheckFuncDeterministic (sqlite3.test.userfunctions.FunctionTests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/builddir/build/BUILD/Python-3.9.0b1/Lib/sqlite3/test/userfunctions.py", line 290, in CheckFuncDeterministic
    self.assertEqual(mock.call_count, 1)
AssertionError: 2 != 1

This test defines a "deterministic" function and ensures that calling it twice in SQLite with only call the underyling Python function only once.

Copy of the test:

    @unittest.skipIf(sqlite.sqlite_version_info < (3, 8, 3), "deterministic parameter not supported")
    def CheckFuncDeterministic(self):
        mock = unittest.mock.Mock(return_value=None)
        self.con.create_function("deterministic", 0, mock, deterministic=True)
        self.con.execute("select deterministic() = deterministic()")
        self.assertEqual(mock.call_count, 1)

In pysqlite_connection_create_function() of Modules/_sqlite/connection.c, determistic=1 sets the following flag:

        flags |= SQLITE_DETERMINISTIC;

This flag is documented as:
"A deterministic function always gives the same answer when it has the same inputs."

* https://www.sqlite.org/deterministic.html
* https://www.sqlite.org/c3ref/c_deterministic.html

"SELECT 1 WHERE deterministic() = deterministic()" query also calls the mock twice.

Running "SELECT deterministic()" query twice also calls the mock twice.

It seems like SQLite 3.32 behaves differently.

Fedora issue: https://bugzilla.redhat.com/show_bug.cgi?id=1839826

Oh, I also tried with a function taking one argument and then call it twice by running "SELECT deterministic(1)" query twice: again, the mock is also called twice.

This seems like it's testing an implementation detail.

Indeed. The SQLite documentation talks about the limitations that non-deterministic functions have, not about deterministic functions being memoized: https://www.sqlite.org/deterministic.html

The flag would be better tested e.g. with a CHECK constraint: https://www.sqlite.org/lang_createtable.html#ckconst

Hi, folks. I took the liberty to create a PR for this; hope you don't mind. I've used partial indices to test deterministic behaviour.

https://www.sqlite.org/partialindex.html

New changeset c610d970f5373b143bf5f5900d4645e6a90fb460 by Erlend Egeberg Aasland in branch 'master':
bpo-40784: Fix sqlite3 deterministic test (GH-20448)
https://github.com/python/cpython/commit/c610d970f5373b143bf5f5900d4645e6a90fb460

Python 3.8 and Python 3.9 are also affected, I created backports.

Python 3.7 is not affected: deterministic parameter was added to Python 3.8.
https://docs.python.org/dev/library/sqlite3.html#sqlite3.Connection.create_function

New changeset 00a240bf7f95bbd220f1cfbf9eb58484a5f9681a by Miss Islington (bot) in branch '3.8':
bpo-40784: Fix sqlite3 deterministic test (GH-20448)
https://github.com/python/cpython/commit/00a240bf7f95bbd220f1cfbf9eb58484a5f9681a

New changeset 8fcc1474ef5d819c144309e048e71e6013544063 by Miss Islington (bot) in branch '3.9':
bpo-40784: Fix sqlite3 deterministic test (GH-20448)
https://github.com/python/cpython/commit/8fcc1474ef5d819c144309e048e71e6013544063

pull_requests: + pull_request19756