Unnecessary lines with starfile=0.5.10

[hanjinliu]

• starfile version: 0.5.10
• Python version: 3.11.9
• Operating System: Windows 11

Description

Since 0.5.10, writing a DataFrame to a star file results in additional lines for each row.
The test workflow of one of my repo that depends on starfile fails only for windows, so maybe this is a windows-specific problem.

What I Did

import starfile
import tempfile
import pandas as pd

df = pd.DataFrame({"a": [1,2,3]})
with tempfile.TemporaryDirectory() as d:
    path = Path(d) / "x.star"
    starfile.write(df, path)
    print(path.read_text())

# Created by the starfile Python package (version 0.5.10) at 22:02:00 on 07/01/2025


data_

loop_
_a #1
1

2

3

[alisterburt]

Really weird, thanks for the report - I'll take a look today 🙂

[alisterburt]

unable to reproduce on macOS so seems windows specific as you suggested

# Created by the starfile Python package (version 0.5.10) at 11:15:19 on 07/01/2025


data_

loop_
_a #1
1
2
3

[alisterburt]

is this pandas version specific? does up/downgrading pandas affect the behavior?

[hanjinliu]

Thank you for your reply, @alisterburt !
I was using the latest pandas=2.2.3, and pandas=2.1.4 resulted in the same problem. pandas<2.1 did not work at all because the method map is not defined in the older versions.
I think it is easier for me to fix this bug because I use Windows. Do you have any idea of which commits are probably related between 0.5.8 and 0.5.10?

[alisterburt]

thanks for taking a look yourself
#60 (812c6bb) was v0.5.8 so everything more recent than May 24th 2024 https://github.com/teamtomo/starfile/commits/main/

I suspect maybe @jahooker's updates to the writing internals are the issue but I'm not sure...

[jojoelfe]

Maybe try replacing '\n' in

starfile/src/starfile/writer.py

Line 179 in 62c7577

).split('\n'):

with os.linesep

[jojoelfe]

@alisterburt was there a reason why macos and windows are disabled here:

starfile/.github/workflows/ci.yml

Line 36 in 62c7577

platform: [ubuntu-latest, ] # ...macos-latest, windows-latest]

[alisterburt]

@jojoelfe I remember looking at the amount of free CI, how big the matrix was and thinking it might be too much... open to turning them back on :-)

[jahooker]

Hi! I think @hanjinliu's changes will fix things. I did assume that the line separator would always be '\n'. My bad!