Conversion to pyarrow crashes the interpreter with a core dump when using decimals

[cpcloud]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

In [9]: import polars as pl

In [10]: pl.__version__
Out[10]: '1.0.0'

In [11]: df = pl.DataFrame({"a": [1.0]})

In [12]: df
Out[12]:
shape: (1, 1)
┌─────┐
│ a   │
│ --- │
│ f64 │
╞═════╡
│ 1.0 │
└─────┘

In [13]: df = df.select(b=pl.col("a").cast(pl.Decimal(76, 38)))

In [14]: df
Out[14]:
shape: (1, 1)
┌─────────────────────────────────┐
│ b                               │
│ ---                             │
│ decimal[76,38]                  │
╞═════════════════════════════════╡
│ 0.9999999999999999774880982345… │
└─────────────────────────────────┘

In [15]: df.to_arrow()
/arrow/cpp/src/arrow/type.cc:1466:  Check failed: (precision) <= (kMaxPrecision)
zsh: abort (core dumped)  ipython

Log output

No response

Issue description

Polars dumps core when casting decimals.

Expected behavior

I would expect an error message, or a successful conversion to pyarrow, really anything but dumping core and crashing the Python interpreter.

Installed versions

--------Version info---------
Polars:               1.0.0
Index type:           UInt32
Platform:             Linux-6.6.36-x86_64-with-glibc2.39
Python:               3.10.14 (main, Mar 19 2024, 21:46:16) [GCC 13.3.0]

----Optional dependencies----
adbc_driver_manager:  <not installed>
cloudpickle:          3.0.0
connectorx:           <not installed>
deltalake:            0.18.1
fastexcel:            <not installed>
fsspec:               2024.6.1
gevent:               <not installed>
great_tables:         <not installed>
hvplot:               <not installed>
matplotlib:           3.9.0
nest_asyncio:         1.6.0
numpy:                2.0.0
openpyxl:             <not installed>
pandas:               2.2.2
pyarrow:              16.1.0
pydantic:             2.7.4
pyiceberg:            <not installed>
sqlalchemy:           2.0.31
torch:                <not installed>
xlsx2csv:             <not installed>
xlsxwriter:           <not installed>

[ritchie46]

It seems that we allow a higher precision than pyarrow/arrow. It seems that they abort the process once that check on pyarrow's side fails.

[ritchie46]

We should raise, pyarrows 16 byte decimal has a max precision of 38.

[cpcloud]

Yeah, it looks like they're using C asserts, which tend to call std::abort 😬

[ritchie46]

Yeah, Given that we cross the FFI boundary, an abort probably is the most sensible thing they can do if they cannot consume the pointers. Will ensure we raise. 👍