set() is a great way to deduplicate small lists, but it's important to note that it requires O(n) extra space (in-place sorting can avoid this overhead, but is more complex).
It is fragile design to write code that depends on 1) the data is so large that you don't have room for a set() BUT 2) it is small enough for an in-memory sort. (IOW, the almost-out-of-memory case invariably degrades over time to flat-out-of-memory).
Another thought: people seem to place too much concern about about the size of various data structures rather than thinking about the data itself. Python containers don't contain anything, they just hold references. (Usually, the weight of a bucket of water is mostly the water, not the bucket itself).
Finally, if your task is to dedup a lot of data, it doesn't make sense to read it all into memory in the first place (which you would need for a sorting approach). It is better dedup it using a set as you read in the data:
# Only the unique lines are ever kept in memory
with open('hugefile.txt') as f:
uniq_lines = set(f)
> rather than thinking about the data itself
…i.e, in this case, we need to realize that the set will require on the order of the size of how many unique items there are, and that in the worst case, that's O(n).
Of course, if you can guarantee that the size of unique set will fit into memory, then you're fine. But you might need a bit of knowledge about the data to do so.
But of course that will be unavoidable if you want fast deduplication. You could do fully lazy deduplication using a generator, but you'd have to avoid using sets.
A set by definition contains only unique items, that is all items in a set are different from each other.
What I meant is you could in theory process a generator and omit duplicates without any real memory usage (even with a file of millions of lines) by chaining generators together. This would be slower than a set but much more memory efficient.
ETA: See raymondh's post below.
* The with-statement only causes the file to be closed after use. It has nothing to do with iteration.
* Files themselves are self-iterable (anything that loops over a file object uses a line-at-a-time iterator). Even writing "for line in f: ..." causes you to loop a line at a time without the whole file being in memory all at once.
* Sets are just one of many objects that takes an iterable as an argument: min(f), max(f), list(f), tuple(f), etc.
I'm not sure I understand your third comment though. As I understand it, iterables have an __iter__ method. __iter__ methods return iterators. So an iterator for the iterable f traverses f and sends f's values to, in this case, set(). I believe we agree there. My initial concern was simply that the statement could be read "The iterator, set(), uses f.", and I don't see where the disagreement arises.
s = set()
for x in f: s.add(x)