Multi Item Imports vs Duplicates "froms"

[cancan101]

What do people think about these two import styles:

from project.stuff import foo
from project.stuff import bar

vs.

from project.stuff import foo, bar
# or
from project.stuff import (
    foo,
    bar,
)

/CC @nathansilberman @taion @itajaja @nevsan

[itajaja]

second preferable

[taion]

One import statement per module please. See https://www.python.org/dev/peps/pep-0008/#imports.

[nevsan]

Yes. And we should remind people that you can use parentheses to break them into multiple lines if they exceed the line length (any opinions here? I don't think it's necessary to always use parentheses if it fits on one legal line length).

[nathansilberman]

I like the first one, but it seems I'm clearly in the minority. For option #2, if we cannot fit the list on one line, which is preferrable:

Option a:

from a.b.c import (
d, e, f,
)

or Option b:

from a.b..c import (
d,
e,
f,
)

My pref is option (a)

[nevsan]

I like option (a). YAPF does something more like option (a). I usually prefer the YAPF default decisions when I don't have strong opinions.

[taion]

I have something like this in the multiline section. The best way to go is to fit everything on one line if possible; otherwise, to put everything on its own line, per 4Catalyzer/flask-resty#136 (comment).

So

# When everything fits on one line:
(
    a, b, c,
)

# Assume these don't fit on one line:
(
    long_identifier_1,
    long_identifier_2,
    long_identifier_3,
)

# But almost never:
(
    a,
    b, c,
)

The latter avoids cruft in diffs from re-wrapping lines when e.g. alphabetizing imports.

There are very rare exceptions to the last; e.g. SQLAlchemy has Index(name, *columns), but:

# This:
Index(
    'ix_name',
    column_1, column_2, column_3,
)

# Is better than:
Index(
    'ix_name',
    column_1,
    column_2,
    column_3,
)