* Fix sqlite3 Connection initialization check
Add proper __init__ validation for sqlite3.Connection to ensure base class
__init__ is called before using connection methods. This fixes the
test_connection_constructor_call_check test case.
Changes:
- Modified Connection.py_new to detect subclassing
- For base Connection class, initialization happens immediately in py_new
- For subclassed Connection, db is initialized as None
- Added __init__ method that performs actual database initialization
- Updated _db_lock error message to match CPython: 'Base Connection.__init__ not called.'
This ensures CPython compatibility where attempting to use a Connection
subclass instance without calling the base __init__ raises ProgrammingError.
* use Initializer trait
Add proper __init__ validation for sqlite3.Cursor to ensure base class
__init__ is called before using cursor methods. This fixes the
test_cursor_constructor_call_check test case.
Changes:
- Modified Cursor to initialize with inner=None in py_new
- Added explicit __init__ method that sets up CursorInner
- Updated close() method to check for uninitialized state
- Changed error message to match CPython: 'Base Cursor.__init__ not called.'
This ensures CPython compatibility where attempting to use a Cursor
instance without calling the base __init__ raises ProgrammingError.
* Update ruff to 0.14.1
* Fix test regression in `test_compile`
* Unmark passing test
* Update `test_syntax` from 3.13.9
---------
Co-authored-by: Noa <coolreader18@gmail.com>
* Copy CPython 3.13 statistics module into RustPython
* Adjust CPython "magic constants" in KDE tests
## test_kde
I'm not too sure why but this one takes a few seconds to run the second
for loop which calculates the cumulative distribution and does a rough
calculation of the area under the curve.
## test_kde_random
I have a lower bound for RustPython to sort a random list of 1_000_000
numbers on my laptop of > 1 hour. By dropping n to 30_000 sort will not
take an egregious amount of time to run. It is then necessary to lower
the tolerance for the math.isclose check, or the computed values may
**randomly** fail due to the higher variance caused by the smaller
sample size.
* Reintroduce expected failure in test_statistics.TestNormalDict.test_slots
* Sync Rust `normal_dist_inv_cdf` with Python equivalent
See https://github.com/python/cpython/pull/95265.
To quote:
> Restores alignment with random.gauss(mu, sigma) and
random.normalvariate(mu, sigma) both. of which are equivalent to
sampling from NormalDist(mu, sigma).inv_cdf(random()). The two functions
in the random module happy accept sigma=0 and give a well-defined
result.
> This also lets the function gently handle a sigma getting smaller,
eventually becoming zero. As sigma decrease, NormalDist(mu,
sigma).inv_cdf(p) forms a tighter and tighter internal around mu and
becoming exactly mu in the limit. For example, NormalDist(100,
1E-300).inv_cdf(0.3) cleanly evaluates to 100.0but withsigma=1e-500``
the function previously would raised an unexpected error.
