[Baypiggies] hello and returning unique elements

[Hy Carrinski]

Erratum: the function I suggested is incomplete without this line
following the docstring.

if seq is None: return None


On Thu, Apr 7, 2011 at 6:42 PM, Hy Carrinski <hcarrinski at gmail.com> wrote:
> Dear BayPIGgies,
>
> Thank you for providing exceptional reading material over the past
> year for an avid conversation follower and occasional meeting
> attendee.
>
> I would like to join the conversation and contribute to the community
> beginning with this post. I will attend our meeting this month, and
> look forward to meeting at Noisebridge or elsewhere to watch PyCon
> 2011 videos together.
>
> While thinking about unique lists in the context of Vikram's recent
> question, I came across a related blog post at
> http://www.peterbe.com/plog/uniqifiers-benchmark
>
> The post emphasizes speed and maintaining original list order when
> removing duplicates from a list.
>
> Although the post is several years old, I thought of a solution to the
> question it poses that is different from the ones proposed by
> BayPIGgies either in the blog's comments or at a previous discussion
> (before reversed, sorted, and set were introduced in Python 2.4) at
> http://code.activestate.com/recipes/52560/
>
> I will be grateful for any comments on my ask forgiveness rather than
> permission implementation. It requires that all elements of the list
> be hashable and works for Python 2.x where x >= 4.
>
> def unique(seq):
>    """Return unique list of hashable objects while preserving order."""
>    seen = {}
>    for ix in reversed(xrange(len(seq))):
>        seen[seq[ix]] = ix
>    return sorted(seen,key=seen.__getitem__)
>
> For the case of hashable objects are there disadvantages to this
> implementation versus the ones referenced above? Is there a reason to
> prefer operator.itemgetter over __getitem__ in this limited context?
> Is there an advantage in clarity or performance to using
> collections.OrderedDict in Python 2.7 and above? For lists containing
> a great number of distinct objects the sort step will introduce a
> lower bound on speed for this implementation.
>
> I look forward to meeting more BayPIGgies in the coming months and years.
>
> Cheers,
> Hy
>