uniq and array of enum members (That are all strings)

H. S. Teoh hsteoh at quickfur.ath.cx
Wed Jan 16 18:46:32 UTC 2019

On Wed, Jan 16, 2019 at 06:25:48PM +0000, bauss via Digitalmars-d-learn wrote:
> I'm aware of how to do it manually as I already stated I went with a
> similar approach.
> There should just be something standard for it and uniq should have an
> overload or something that allows for another behavior that doesn't
> rely on sorting.

Feel free to submit a PR to Phobos for this. The templatized function I
posted is probably already pretty close to Phobos standards, you just
need to add a few unittests and a nice ddoc header.

I wouldn't call it `uniq`, though.  It doesn't quite jive with the
"examine nearby elements only" philosophy of .uniq (which inherits from
the Unix `uniq` that does the same thing), and it's not @nogc unlike
.uniq, etc., and IME, trying to shoehorn something like this into an
existing symbol with already preconceived semantics will make the PR
review needlessly controversial.  I'd just submit it under a different

> Even if it's "slower".

Actually, filtering with an AA is not slower, since the AA amortizes the
overall cost to O(n), which is even better than the O(n log n) of the
pre-sorting approach.

Of course, this comes at the cost of not being @nogc, which might be a
turnoff for some folks.  It's probably possible to write a @nogc version
of this function, but it will be more complicated (much more complicated
if you're working with a range of unknown length). This is one of the
places where IMO the hassle of manually managing the memory of a lookup
table far outweighs the "cost" (perceived or real) of just embracing the


Elegant or ugly code as well as fine or rude sentences have something in common: they don't depend on the language. -- Luca De Vitis

More information about the Digitalmars-d-learn mailing list