Love your data

Intro to data structures and functional programming using Python.

Transcript

1. LOVE YOUR DATA
2. SMART DATA STRUCTURES AND DUMB CODE WORKS A LOT BETTER THAN THE OTHER WAY AROUND Eric S. Raymond
3. WHY SHOULD I LEARN DATA STRUCTURES? ▸ Well known trade-offs ▸ A lot of algorithms for any purposes ▸ Language independent ▸ They are fun!
4. COMMON DATA STRUCTURES ▸ Lists and Arrays ▸ Trees (Binary, Red-Black, B-Tree) and Heaps ▸ Maps or Dicts ▸ Graphs ▸ Queues, Stacks, Deques ▸ Sets
5. DO YOU KNOW WELL SIMPLE LIST? ▸ Collection size 10kk elements ▸ foo 3.08 s ± 169 ms per loop ▸ bar 3.51 s ± 56.7 ms per loop ▸ baz 5.15 s ± 364 ms per loop
6. BUT WHY?
7. TREES VS LISTS LIST TREE ACCESS O(n) O(log(n)) SEARCH O(n) O(log(n)) INSERTION O(1) O(log(n)) DELETION O(1) O(log(n))
9. TEXT IT’S NOT HARD AT ALL ▸ Only for spins on insert and delete ▸ Still O(log(n)) ▸ Sorting for free ▸ Average sorting algorithm 
 O(n log(n)) ▸ Worst sorting algorithm
 O(n^2)
10. WHAT ABOUT PROCESSING ALL THESE GRAPHS AND TREES? ▸ Complex is better than complicated. ▸ You shouldn’t do it every time. ▸ Most of algorithms are recursive. ▸ Complex code for iterating.
11. ONE MORE THING LIST IS A TREE TOO
13. NOT AGAIN. WHY DO WE NEED THIS? ▸ We already have first class functions ▸ Higher abstraction ▸ Code that obeys strict laws ▸ Clearer code
14. MAP, FILTER, REDUCE OVER DATA STRUCTURE Imperative Functional List For cycle with some mutable state map / filter / fold Graph Complex cycle with mutable state and stack or queue map / filter / fold Tree Complex cycle with mutable state and stack or queue map / filter / fold Set For cycle with some mutable state map / filter / fold
15. AND WHERE ARE STRICT LAWS? ▸ Monoid (mempty, mappend) ▸ Applicative (pure, liftA2, apply) ▸ Foldable (foldl, foldr) ▸ Functor (map)
16. GIMME SOME CODE
17. APPLICATIVE class Applicative(Functor): @classmethod def pure(cls): raise NotImplementedError def apply(self, other): return self.liftA2( other, lambda f, x: f(x)) def liftA2(self, other, f): return other \ .map(partial(partial, f)) \ .apply(self)
18. TREE class Tree: class Leaf: pass class Branch: pass def __iter__(self): stack = deque() stack.append(self) while stack: x = stack.pop() if isinstance(x.val, self.Leaf): yield x.val.val else: stack.append(x.val.right) stack.append(x.val.left)
19. BASE LIST class List(Monoid, Foldable, Applicative): def map(self, f): return self.foldr( lambda r, x: \ r.cons(f(x)), self.mempty()) @classmethod def pure(cls, x): return List(x) def liftA2(self, other, f): return List(*[f(s, o) for s in iter(self) for o in iter(other)])
20. SHOW ME REAL MAGIC! a = List(1, 2, 3) # => [1, 2, 3] b = a.map(partial(mul, 3)) # => [3, 6, 9] c = a.liftA2(b, add) # => [4, 7, 10, 5, 8, 11, 6, 9, 12] d = a.map(partial(partial, add)).apply(b) # => [4, 7, 10, 5, 8, 11, 6, 9, 12] c == d # True
21. THANK YOU! 
 QUESTIONS? @mikebohdan @mikhail_bohdan
22. LINKS ▸ https://gist.github.com/mikebohdan/ 47391c7109d426baa122c597e674586b ▸ https://speakerdeck.com/mikhail_bohdan/love-your-data ▸ https://www.youtube.com/playlist? list=PLbgaMIhjbmEnaH_LTkxLI7FMa2HsnawM_ ▸ http://www.cse.chalmers.se/research/group/logic/ TypesSS05/Extra/geuvers.pdf