Today: lambda output, lambda/def, custom sorting with lambda, wordcount sorting, introduction to modules

Lambda - Advanced - A Small Superpower

Lambda is powerful feature, letting you express a lot of computation in very little space. As a result, it's weird looking at first, but when it clicks, you should feel like a Power Hacker when you wield it. Kind of a superpower.

These are well suited to little in-class exercises .. just one line long. Not easy, but they are short!

Syntax Reminder - Map Lambda

map() takes in a lambda of one parameter, and a list, and calls that lambda for every element in the list, like this:

>>> list(map(lambda n: 2 * n, [1, 2, 3, 4, 5]))
[2, 4, 6, 8, 10]

Recall: Lambda 1-2-3 Steps

1. The word "lambda"

2. The input to the lambda will be an element from the list. What type are these - int? string? Choose an appropriate name for the lambda parameter like n: or s:

3. Write an expression to produce the lambda output, no "return". Typically this all fits on one line.

Visualization - What Is Def? What is Lambda?

Previously, here is def

def double(n):
    return 2 * n

The def sets up the name of the function, and points it to that body of code.

What is Lambda?

The lambda creates the code, but without the need for the name.

In the interpreter, the code from the two prints out with brackets < .. > which Python uses when it needs to print something that is not printable.

>>> def double(n):
...   return 2 * n
... 
>>> double
<function double at 0x7fb944ab6ee0>
>>> 
>>> lambda n: 2 * n
<function <lambda> at 0x7fb944ad03a0>

(optional) Party Trick - Def vs. Lambda

This is just kind of a trick, but it shows how you can actually make your own def using lambda and an equal sign. A def has code and a name. Here we use = to make the name fn point to the lambda code. Then we can call it like any other function.

>>> lambda n: 10 * n
<function <lambda> at 0x1023d1ee0>
>>> 
>>> fn = lambda n: 10 * n     # assign to "fn"
>>> 
>>> fn
<function <lambda> at 0x1023d2020>
>>> 
>>> fn(4)                     # fn call works!
40
>>> fn(123)
1230
>>>

This is a peak behind the curtain, showing what def is doing under the hood. Python is in a way very simple. A variable is a name in the code that opints to a value, and this is true for any type of value, even code.

Use Lambda For Everything? No

Now we have lambda, do we just use it for everything? No. Most of a program is good old def, but lambda is a great time-saving technique for spots in the program which need a short phrase of code.

Lambda Example/Exercise min_x()

A case where lambda works great. Convert the elements in a list - the lambda converts 1 element, and map runs it over the whole list.

> min_x()

Given a non-empty list of len-2 (x, y) points, i.e. tuples. What is the leftmost x among the tuples? Return the smallest x value among all the tuples, e.g. [(4, 2), (1, 2) (2, 3)] returns the value 1. Solve with a map/lambda and the builtin min(). Recall: min([4, 1, 2]) returns 1

min_x([(4, 2), (1, 2), (2, 3)])  ->  1

Solution min_x()

Use lambda to extract just the x value from each (x, y). then feed that into the builtin min() function, and we're done!

      [(4, 2), (1, 2), (2, 3)]
         |       |       |
        [4       1       2]   ->  min()  ->  1

def min_x(points):
    return min(map(lambda point: point[0], points))

Think About Def Features

Def can do things that lambda cannot.

• def introduces a name for the code, so we can call it from multiple places
• Def has room for real code features:
• Multiple lines
• If statements
• Variables
• Loops
• Doctests
• Inline comments
• Lambda: best without any of that, just short code that fits on one line

Long Computation - Use def, not lambda

• What to do if computation does not fit in 1 line?
• Just write a def
• Then refer to the function by name to use it
• e.g. map() can use the def

map/def Example - map_parens()

> map_parens()

In lambda1, see the map_parens() problem.

['xx(hi)xx', 'abc(there)xyz', 'fish'] ->
  ['hi', 'there', 'fish']

map_parens() Solution

Solution Code. Write the "parens" helper function that works on one string.

'xx(hi)xx' -> 'hi'
'fish'     -> 'fish'


def parens(s):
    left = s.find('(')
    right = s.find(')', left)
    
    if left == -1 or right == -1:
        return s
    return s[left + 1:right]

Then use map(), using the name "parens" to refer to the helper function code.

def map_parens(strs):
    return map(parens, strs)

Custom Sort - Power Feature

• Python sorting has a lot of power in it
• Use lambda to guide the sorting
• This code feels powerful and dense
• More examples in section!

Python Custom Sort - Food Examples

# food tuple
# (name, tasty, healthy)
('donut', 10, 1)

• food tuple - (name, tasty 1-10, healthy 1-10)
• ('donut', 10, 1)
• Get data out of len-3 food tuple:
• food[0] = its name
• food[1] = how tasty it is 1-10
• food[2] = how healthy it is 1-10
• Food or garnish? - Radish

We'll try these food examples in the interpreter.

Default sorted()

By default sorted() works on list of tuples, compares [0] first, then [1], and so on

>>> foods = [('radish', 2, 8), ('donut', 10, 1), ('apple', 7, 9), ('broccoli', 6, 10)]
>>> 
>>> # By default, sorts food tuples by [0]
>>> sorted(foods)
[('apple', 7, 9), ('broccoli', 6, 10), ('donut', 10, 1), ('radish', 2, 8)]
>>> 

Sort By Tastiness

• Say I want to sort by tastiness
  e.g. the radish vs. donut dimension
• Control how sorted() looks a the data
• Like drawing a circle around tasty values - sort by these!
• How can we get the code to do this?

Project Out Sort-By Values

• How to code sort-by-tasty
• For each element in list
  "Project out" a sort-by value to be used in sorting comparisons
• Here, for each food, project out its tasty int
• aka "Proxy" strategy
  Each element, proxy value is used for sorting comparisons

Project Out With Lambda

• Q: how to project out these sort-by proxy values?
• A: lambda

Custom Sort Lambda - Plan

• 1. Call sorted() as usual
• 2. provide key=lambda to control sorting
• 3. The lambda input is one elem from the list
  The lambda output is the sort-by (proxy) value for that elem
• e.g. sort by tasty
  lambda food: food[1]
• e.g. sort by healthy
  lambda food: food[2]

Sort Tasty Increasing

>>> foods = [('radish', 2, 8), ('donut', 10, 1), ('apple', 7, 9), ('broccoli', 6, 10)]
>>> 
>>> sorted(foods, key=lambda food: food[1])
[('radish', 2, 8), ('broccoli', 6, 10), ('apple', 7, 9), ('donut', 10, 1)]

Most Tasty First (reverse=True)

>>> sorted(foods, key=lambda food: food[1], reverse=True)
[('donut', 10, 1), ('apple', 7, 9), ('broccoli', 6, 10), ('radish', 2, 8)]

Most Healthy First

>>> sorted(foods, key=lambda food: food[2], reverse=True)
[('broccoli', 6, 10), ('apple', 7, 9), ('radish', 2, 8), ('donut', 10, 1)]

Sort by tasty * healthy

Not limited to just projecting out existing values. We can project out a computed value. Here we compute tasty * healthy and sort on that. So apple is first, 7 * 9 = 63, broccoli is second with 6 * 10 = 60. Donut is last :(

>>> sorted(foods, key=lambda food: food[1] * food[2], reverse=True)
[('apple', 7, 9), ('broccoli', 6, 10), ('radish', 2, 8), ('donut', 10, 1)]
>>>

Sorted vs. Min Max

• What code give us the most tasty food?
  Or the least tasty?
• Sorting n things is kind of expensive
• Could sort, take the last item - overly expensive approach
• Use max(), max takes a key=lambda just like sorted()
• e.g. pull out most or least tasty food - change "sorted" to "max" or "min"

>>> foods = [('radish', 2, 8), ('donut', 10, 1), ('apple', 7, 9), ('broccoli', 6, 10)]
>>> max(foods)     # uses [0] by default - tragic!
('radish', 2, 8)
>>> 
>>> sorted(foods, key=lambda food: food[1])
[('radish', 2, 8), ('broccoli', 6, 10), ('apple', 7, 9), ('donut', 10, 1)]
>>> 
>>> min(foods, key=lambda food: food[1])  # least tasty
('radish', 2, 8)
>>> max(foods, key=lambda food: food[1])  # most tasty
('donut', 10, 1)

Performance point: computing one max/min element is much faster than sorting all n elements.

Movie Examples / Exercises

Given a list of movie tuples, (name, score, date-score), e.g.

[('alien', 8, 1), ('titanic', 6, 9), ('parasite', 10, 6), ('caddyshack', 4, 5)]

sort_score(movies)

> sort_score()

Given a list of movie tuples, (name, score, date-score), where score is a rating 1-10, and date 1-10 is a rating as a "date" movie. Return a list sorted in increasing order by score.

sort_date(movies)

> sort_date()

Given a list of movie tuples, (name, score, date-score), where score is a rating 1-10, and date-score 1-10 is a rating as a "date" movie. Return the list sorted in decreasing by date score.

sort21() Example/Exercise - Distance

> sort21()

sort21(nums): Given a list of numbers. Return the list of numbers sorted with the closest to 21 first and the farthest from 21 last. Note: abs(n) is the absolute value function. Use sorted/lambda.

[15, 19, 21, 30, 0]  -> [21, 19, 15, 30, 0]

Idea: subtract each number from 21, use that as the sort-by value. Try this, see what it does.

Want: sorted with closest to 21 first

      [15,   19,  21,  0,   30]

21-n:  6     2    0    21   -9

Solution idea:

The negative number is a problem. Use abs() function, Python's absolute value function. The absolute value of the different between two numbers is, in a sense, the "distance" between those two numbers.

Python String Sort Case-Sensitive - Fix With Lambda

By default, < places uppercase before lowercase, so this is what sorted() does. This is rarely what we want.

Fix: project out lowercase version of string as sort-by. The lambda takes in one elem from list - in this case 1 string

e.g. lambda s: s.lower()

>>> # The default sorting is not good with upper/lower case
>>> strs = ['coffee', 'Donut', 'Zebra', 'apple', 'Banana']
>>> sorted(strs)
['Banana', 'Donut', 'Zebra', 'apple', 'coffee']
>>> 
>>> sorted(strs, key=lambda s: s.lower())    # not case sensitive
['apple', 'Banana', 'coffee', 'Donut', 'Zebra']
>>> 
>>> sorted(strs, key=lambda s: s[len(s) - 1])  # by last char
['Zebra', 'Banana', 'coffee', 'apple', 'Donut']
>>> 

Put It All Together - WordCount + Sorted

Look at wordcount project, apply custom sorting to the output stage, a very realistic lambda application.

Wordcount - Top-Count - Lambda

Here is the WordCount project we had before. This time look at the print_counts() and print_top() functions.

> wordcount.zip

print_counts() - Alphabetic Output

Here is the output of the regular print_counts() function, which prints out in alphabetic order. Output looks like:

$ python3 wordcount.py poem.txt 
are 2
blue 2
red 2
roses 1
violets 1
$
$ python3 wordcount.py alice-book.txt
a 631
a-piece 1
abide 1
able 1
about 94
...
youth 6
zealand 1
zigzag 1
$

print_counts() Code

This is the standard dict-output sorted loop.

def print_counts(counts):
    """
    Given counts dict, print out each word and count
    one per line in alphabetical order, like this
    aardvark 1
    apple 13
    ...
    """
    for word in sorted(counts.keys()):
        print(word, counts[word])

    # Alternately use .items() to access all the key/value tuples
    # for key, value in sorted(counts.items()):
    #    print(key, value)

-top Output Feature

Now we'll think about a new -top feature.

The print_top(counts, n) function implements this — print the n most common words in decreasing order by count.

$ python3 wordcount-solution.py -top 10 alice-book.txt 
the 1639
and 866
to 725
a 631
she 541
it 530
of 511
said 462
i 410
alice 386

• Write code for print_top() function - work out the code in there
• 1. Print the counts.items() to see what we have
  Python technique - Python can print almost any intermediate data structure, so we're leveraging that
• Can use the file 'alice-start.txt' which is just the first page, so a smaller test case to look at
• Recall: dict.items() - random order of word/count pairs
  [('sister', 12), ('rabbit', 5), ...]
• 2. Need to sort the pairs: decreasing order by count
  Use sorted/lambda
  This code is incredibly short and powerful
• 3. Print only the first n items, not all of them
  How do you shorten a list to just the first n?

print_top() Demo/Exercise

def print_top(counts, n):
    """
    Given counts dict and int N, print the N most common words
    in decreasing order of count
    the 1045
    a 672
    ...
    """
    items = counts.items()
    # Could print the items in raw form, just to see what we have
    # print(items)
    pass
    # Your code - my solution is 3 lines long, but it's dense!
    # Sort the items with a lambda so the most common words are first.
    # Then print just the first N word,count pairs with a slice

print_top() Solution

Here's the lines - sort by count decreasing order. Then slice to take the top n.

    # 1. Sort largest count first
    items = sorted(items, key=lambda pair: pair[1], reverse=True)
    # 2. Slice to grab first N
    for word, count in items[:n]:
        print(word, count)