Check if dictionary empty with == {}

Steven D'Aprano steve+comp.lang.python at pearwood.info
Thu Aug 20 05:16:52 CEST 2015


On Thursday 20 August 2015 08:57, Anton wrote:

> Probably a silly question.
> Let's say I have a dictionary mydict and I need to test if a dictionary is
> empty.
> 
> I would use
> 
> if not mydict:
>     """do something"""
> 
> But I just came across a line of code like:
> 
> if mydict == {}:
>     """do something"""
> 
> which seems odd to me, but maybe there is a valid use case, thus I decided
> to ask the community.

It's neither good code or bad code. It looks like something a beginner or 
casual Python user (e.g. a sys admin) might have written, but that isn't 
necessarily bad. It's no more wrong than writing `if x == 0` to test for an 
"empty" number (or "nothing", in the numeric sense).

Pros:

+ It's pretty short and explicit.

+ It allows for duck-typing: if some de facto mapping object wants to 
  support the dict API without inheriting from dict, it can.

+ Even a beginner can understand it: "does mydict equal an 
  empty dict?"


Cons:

- It looks a bit Python 1.5-ish. People used to more modern 
  idioms may have an (unjustified, in my opinion) "WTF" moment
  when looking at it.

- It *might* be a bit slow, since it takes time to create an
  empty dict; on the other hand, it also takes time to call
  isinstance(), so if you care about this, I want to see your
  profiling results and benchmarks.

Actually, it's not a bit slow, it's *significantly* faster than an 
isinstance check:

steve at runes:~$ python2.7 -m timeit -s "mydict = {1:2}" \
> "if mydict == {}: pass"
10000000 loops, best of 3: 0.0872 usec per loop
steve at runes:~$ python2.7 -m timeit -s "mydict = {1:2}" \
> "if isinstance(mydict, dict) and not mydict: pass"
1000000 loops, best of 3: 0.257 usec per loop

So maybe it's a micro-optimization?


TL;DR

There's nothing wrong with it. It is ever-so-subtly different from the 
various alternatives, so if you change it, don't be surprised if you break 
something.


-- 
Steven



More information about the Python-list mailing list