Rule suggestion: recommend dict.fromkeys over comprehension

[tjkuson]

Replace

{elt: None for elt in foo}

with

dict.fromkeys(foo)

dict.fromkeys is more efficient and readable.

In [4]: import random

In [5]: choices = ("red", "green", "blue")

In [6]: dupes = [random.choice(choices) for _ in range(1000)]

In [7]: %timeit list({elt: None for elt in dupes})
18.8 µs ± 97.8 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

In [8]: %timeit list(dict.fromkeys(dupes))
12.6 µs ± 49.9 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)

Efficiency boost tested on both Python 3.9 and 3.12.

Note: would have to skip this rule for mutable values (see #4613).

[charliermarsh]

Looks great!

[mikeleppane]

I could pick up this one?

[charliermarsh]

Feel free unless @tjkuson planned to do it or had started on it -- I'll leave it to @tjkuson to confirm!

[tjkuson]

Feel free unless @tjkuson planned to do it or had started on it -- I'll leave it to @tjkuson to confirm!

Haven't started anything, have at it! @mikeleppane

[Skylion007]

This would be a great rule for refurb, C4, or perflint. You should consider pinging the author of those libraries as well.

[bswck]

This applies not only to None, but any constant.
In my opinion, {elt: 0 for elt in foo} should also be replaced with dict.fromkeys(foo, 0).

[bswck]

@tjkuson @charliermarsh would you agree? ↑

[tjkuson]

@tjkuson would you agree? ↑

Yes, this was my intention with the issue. Admittedly, I communicated that poorly by using only examples with None. (I mostly use dict.fromkeys in the context of ordered deduplication.)

[charliermarsh]

Closed by #9613. Thanks @mikeleppane!