PySeries.new_from_anyvalues loses microsecond precision

[MarcoGorelli]

Polars version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of Polars.

Issue description

In the below example, notice how the microseconds in the result match for the tz-naive case, but not for the tz-aware one

Reproducible example

In [62]: pl.Series([datetime(2514, 5, 30, 1, 53, 4, 986754)])
Out[62]:
shape: (1,)
Series: '' [datetime[μs]]
[
        2514-05-30 01:53:04.986754
]

In [63]: pl.Series([datetime(2514, 5, 30, 1, 53, 4, 986754, tzinfo=dt.timezone.utc)])
Out[63]:
shape: (1,)
Series: '' [datetime[μs, UTC]]
[
        2514-05-30 01:53:04.986756 UTC
]

Expected behavior

In the second case

In [63]: pl.Series([datetime(2514, 5, 30, 1, 53, 4, 986754, tzinfo=dt.timezone.utc)])
Out[63]:
shape: (1,)
Series: '' [datetime[μs, UTC]]
[
        2514-05-30 01:53:04.986754 UTC
]

The issue is that in the tz-aware case,

polars/py-polars/polars/internals/construction.py

Lines 329 to 332 in 8d5585d

	if dtype == Datetime and value.tzinfo is not None:
	py_series = PySeries.new_from_anyvalues(name, values)
	tz = str(value.tzinfo)
	return pli.wrap_s(py_series).dt.replace_time_zone(tz)._s

PySeries.new_from_anyvalues is used.

And in that one, in

polars/py-polars/src/series.rs

Lines 227 to 232 in 9cfc987

	pub fn new_from_anyvalues(name: &str, val: Vec<Wrap<AnyValue<'_>>>) -> PyResult<PySeries> {
	let avs = slice_extract_wrapped(&val);
	// from anyvalues is fallible
	let s = Series::from_any_values(name, avs).map_err(PyPolarsErr::from)?;
	Ok(s.into())
	}

avs will already have incorrectly converted to int:

[Datetime(17179869184986756, Microseconds, None)]

But that number is incorrect:

In [7]: dt.datetime(1970, 1, 1, tzinfo=dt.timezone.utc) + dt.timedelta(microseconds=17179869184986756)
Out[7]: datetime.datetime(2514, 5, 30, 1, 53, 4, 986756, tzinfo=datetime.timezone.utc)

It should be 17179869184986754:

In [8]: dt.datetime(1970, 1, 1, tzinfo=dt.timezone.utc) + dt.timedelta(microseconds=17179869184986754)
Out[8]: datetime.datetime(2514, 5, 30, 1, 53, 4, 986754, tzinfo=datetime.timezone.utc)

Installed versions

---Version info---
Polars: 0.16.6
Index type: UInt32
Platform: Linux-5.10.102.1-microsoft-standard-WSL2-x86_64-with-glibc2.35
Python: 3.11.1 (main, Dec  7 2022, 01:11:34) [GCC 11.3.0]
---Optional dependencies---
pyarrow: 11.0.0
pandas: 1.5.3
numpy: 1.24.1
fsspec: <not installed>
connectorx: <not installed>
xlsx2csv: <not installed>
deltalake: <not installed>
matplotlib: <not installed>

[aldanor]

Just to link it here, perhaps somewhat related or same code branches: #7212