Binary type data disappears when pickled

[nrlugg]

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

When a pl.Binary type Series is pickled and unpickled, the data is lost (all becomes null) and the type is changed to pl.Int32.

This is a particular problem when using multiprocessing (or similar) since when I multiprocess over data with pl.Binary columns, that data is lost when sent to the other processes (via pickle).

This also happens for dill.

Reproducible example

import pickle
import numpy as np
import polars as pl

data = pl.Series(
    "binary_data",
    [np.random.bytes(l) for l in np.random.randint(4, 16, (8,))]
)

print(data)
"""
shape: (8,)
Series: 'binary_data' [binary]
[
	[binary data]
	[binary data]
	[binary data]
	[binary data]
	[binary data]
	[binary data]
	[binary data]
	[binary data]
]
"""

print(pickle.loads(pickle.dumps(data)))
"""
shape: (8,)
Series: 'binary_data' [i32]
[
	null
	null
	null
	null
	null
	null
	null
	null
]
"""

Expected behavior

There Series data should not change.

Installed versions

--Version info---
Polars: 0.15.18
Index type: UInt32
Platform: macOS-13.1-arm64-arm-64bit
Python: 3.10.9 (main, Jan 17 2023, 15:47:44) [Clang 14.0.0 (clang-1400.0.29.202)]
---Optional dependencies---
pyarrow: 7.0.0
pandas: 1.5.2
numpy: 1.23.5
fsspec: 2022.11.0
connectorx: <not installed>
xlsx2csv: <not installed>
deltalake: <not installed>
matplotlib: 3.6.2

[ritchie46]

This requires serde to implemented for dtype-binary. This is a good first issue. One can look at the serde for the other data types for inspiration.

[papparapa]

May I work on it?

[ritchie46]

May I work on it?

Please 🙂

[nrlugg]

Omg this is already in the latest release; thank you both!!