Constructing Series with dtype pl.List(pl.Float32) rounds inputs to integers

[SujeetO]

Research

• I have searched the above polars tags on Stack Overflow for similar questions.

• I have asked my usage related question on Stack Overflow.

Link to question on Stack Overflow

No response

Question about Polars

I am using '0.18.3' version of polars and seeing below problem:

x = pl.Series([[1,2],[2.6,3],[3,4,4.4]] )
print(x)

shape: (3,)
Series: '' [list[i64]]
[
[1, 2]
[2, 3]
[3, 4, 4]
]

it has converted the float number to int without any warning or error. it this expected?

[SujeetO]

If yes, how can I avoid that?

[ion-elgreco]

You can specify the dtype while creating it:

x = pl.Series([[1,2],[2.6,3],[3,4,4.4]], dtype=pl.List(pl.Float32)) 

[MarcoGorelli]

would your expected outcome be that dtype=pl.List(pl.Int64) raise an error, instead of rounding the 4.4 to 4? If so, then this sounds related to #9119 (it's different though, not a duplicate, just bringing it up for reference)

[deanm0000]

@ion-elgreco that doesn't work for me. I still get all ints.

[avimallu]

My output is a little different - it does get cast to float, but after first reading as an integer, it seems:

>>> import polars as pl
>>> pl.Series([[1,2],[2.6,3],[3,4,4.4]], dtype=pl.List(pl.Float32)) 
shape: (3,)
Series: '' [list[f32]]
[
        [1.0, 2.0]
        [2.0, 3.0]
        [3.0, 4.0, 4.0]
]
>>> pl.__version__
'0.18.1'

I would say it's a bug.

[ion-elgreco]

@ion-elgreco that doesn't work for me. I still get all ints.

My bad, I glanced over the printed result. Indeed, it's returning still integers, as @avimallu also shows.

[SujeetO]

how to convert label of this issue to bug from question?

[Wainberg]

Duplicate of #11009, which has now been accepted.

[stinodego]

Closing in favor of #11156