% class-6 talking points

This week: comprehensions and dictionaries - the heart of Pythonic style!

Quiz:  (Try not to spend too much time on this)
almost everyone finished this fast 
 - but almost everyone made several important errors!
these aren't trick questions or obscure corners - they are fundamental
 misunderstanding these points results in common errors (usually harder to find)
4. y[1] is the second element, not the first
3. y[1] = 99 also updates x, x and y are aliases (sharing, need for copy module)
2. s.append(['e','f']) -> ['a','b','c','d',['e','f']] 
   append adds *one element*, here that element is itself a list,  nested
   likewise s.append('d') -> ['a','b','c','d'] not [..[d]],element is just 'd'
   append vs. extend, extend vs. +
1. s.append(d) == None, append is a mutator, changes s but returns None!
There is a HUGE disinction regarding operators, methods, functions
 *pure functions*, *functional* operators - do not change args, return new value
 *mutators* - change arg, do *not* return value (return None in Python)
Huge impact on programming style, alternate *functional programming* universe:
 sort vs. sorted (demo right in interpreter)
 also x if c else y vs if...:  + vs. append etc.

List comprehensions: comprehensions.py
Generate a new list from an existing collection
Review little tuple database from last week
Show for loop to build up list, compare to list comprehension
Show more elaborate pattern/expr (aside: string formating is not just for print)
map and filter patterns comprehensions vs. built-in functions
comprehensions are more versatile, Pythonic
functions support functional programming style, like Lisp, Haskell, F#, ...

Dictionaries:
Key/value pairs with *unique* keys.
for databases, lookup tables.  Much computing is looking things up.
Keys must be hashable (immutable) - numbers, strings, tuples but not lists
Chapter 13 word frequency example is very nice
Show Roman numerals: roman.py, nice combo of dict, comprehension, list
Mention Norvig spelling corrector.py - uses everything so far, including regexp

Quiz, break, presentations

sets - collections without order and repetition, demo set.py

arrays - all the elements are the same type, efficient numerical (and other) cals
arrays are not built in!  Must get numpy. SeaPIG this Thurs, Nov 18, 7pm, Capitol Hill

Reductions, product.py:
builtin: sum, max, min, all, any
reduce function: motivate operator module and lambda expressions
lambda is used heavily with map and filter - but not needed with comprehensions

That's it folks, full Pythonic style.  Next two weeks - object-oriented prog.