import lost after formatting

[arthurxlee]

I have a file that looks like this

from xxxxxxxxxxxxxxxx import AAAAAAAAAA, BBBBBBBBBB
from xxxxxxxxxxxxxxxx import CCCCCCCCC, DDDDDDDDD  # xxxxxxxxxxxxxxxxxx

print(CCCCCCCCC)

The content from line 2 could be folded to line1, but I suspect the comment at the end of line 2 gets in the way and the entire line will be lost after formatting.

Using isort version 5.5.4

[timothycrosley]

@arthurxlee, can you provide the output of isort . --show-config?
isort should never lose imports (unless they are duplicated) running this code snippet with vanilla settings gives me:

from xxxxxxxxxxxxxxxx import (  # xxxxxxxxxxxxxxxxxx
    AAAAAAAAAA,
    BBBBBBBBBB,
    CCCCCCCCC,
    DDDDDDDDD,
)

print(CCCCCCCCC)

Which, I believe is correct. This leads me to believe it is happening only with certain settings, or that the actual code is different from the given example in some way.

In case it's helpful, this UI: https://pycqa.github.io/isort/docs/quick_start/0.-try/ can help to produce a minimal example.

Thanks!

~Timothy

[arthurxlee]

Attached is the output from show config.
isort_config.txt

[arthurxlee]

I would also suggest that the lines should not be folded since the inline comment would very probably not apply to the other lines, so them folding may not be desirable.

[timothycrosley]

Thank you for sharing your configuration, and I'm sorry you've been impacted by this issue! With the configuration, I was able to quickly confirm the issue was isolated to the wrap mode being used (which is a relatively newer community contributed mode). I've fixed this issue and added additional regression tests to ensure it doesn't happen going forward.

I would also suggest that the lines should not be folded since the inline comment would very probably not apply to the other lines, so them folding may not be desirable.

I agree with this in principle, though it is really very hard to tell comment intention.
For instance this slightly modified example where B should go to line 1 and C should go to line 2.

from xxxxxxxxxxxxxxxx import AAAAAAAAAA, CCCCCCCCC
from xxxxxxxxxxxxxxxx import BBBBBBBBB, DDDDDDDDD  # Comment for B

In theory there is some optimizations that could be made for just this wrap-mode, but if you really want to make sure comment intention is explicitly mapped, the only way to do that would be to use a mode that leads to one import per a line. Otherwise, for individual cases like this you can tell isort to skip the line with the comment:

from xxxxxxxxxxxxxxxx import AAAAAAAAAA, CCCCCCCCC
from xxxxxxxxxxxxxxxx import BBBBBBBBB, DDDDDDDDD  # Comment for D isort: skip

Thanks!

~Timothy