Comments interfering with fluent formatting

[max-sixty]

Copied over from #379 (comment)

Without a comment, formatted with black (-l 79)

    return (
        df_day.assign(date=date, fund_ticker=ticker)
        .pipe(lambda x: x.loc[x["Shares"].notnull()])
        .assign(Shares=df_day["Shares"].astype("float64"))
    )

Adding a comment, before formatting

    return (
        df_day.assign(date=date, fund_ticker=ticker)
        .pipe(lambda x: x.loc[x["Shares"].notnull()])
        # arrives as int sometimes
        .assign(Shares=df_day["Shares"].astype("float64"))
    )

With a comment, after formatting with black:

    return (
        df_day.assign(date=date, fund_ticker=ticker).pipe(
            lambda x: x.loc[x["Shares"].notnull()]
        )
        # arrives as int sometimes
        .assign(Shares=df_day["Shares"].astype("float64"))
    )

Without thinking about the more general problem, the second example here is ideal - the comment can be within a fluent expression and wouldn't change the fluent expression formatting.

Thanks again for such an awesome library and vision!

Operating system: MacOS
Python version: 3.7
Black version: 18.6b4
Does also happen on master:

[ambv]

I agree with you that comments should not break the flow of fluent interfaces. We'll address it next week.

[ambv]

The problem is that the standalone comment "splits" the fluent interface so Black sees two periods before (and fluent interfaces are implemented to only apply for >2 periods), then a comment, and one after.

[ambv]

See this Black playground.