Re: What to do for faster `remove-duplicates'?

[Artur Malabarba]

>>> I attach the patch. I did a bunch of `benchmark-run' and it seems that
>>> 100 elements is the breaking point.
>>
>> Small question. How much slower is this patch compared to the current
>> version on a list of 99 elements? (Due to having to calculate the
>> length)
>
> For 99 unique candidates, the call to `length' takes 5% time compared to
> the call to `delete-dups':
>
> (/ 3.774e-06 6.3028e-05)
> 0.05987814939392017
>
> It becomes worse for a small amount of unique candidates, going to 30%
> for 10 candidates, but a lot of that is the standard cost of calling a
> function.
>
> I don't know if it's worth optimizing further.

Nah, I think it's fine, I was just curious.
Doing (nthcdr 100 list) instead of (> (length list) 100) might improve
that a little, but either way is fine.