Python Patterns: max Instead of if

June 14, 2018 #python #software-development

When writing Python, I often have to look through a set of objects, determine a score for each one of them, and save both the best score and object associated with it. For example, looking for the highest scoring word that I can make in Scrabble with the letters I currently have.

One way to do this is to loop over all the objects and use a placeholder to remember the best one seen so far, like this:

# Set up placeholder variables
best_score = 0
best_word = None

# Try all possible words, saving the best one seen
for word in possible_words(my_letters):
  score = score_word(word)

  if score > best_score:
    best_score = score
    best_word = word

You have probably written this logic before in some of your own code. The code is not that complicated, but we can still improve its readability with a quick tweak.

Simplifying with max()

What does if score > best_score remind you of? The way we might implement the max() function! Using max() helps us simplify the code nicely:

# Set up placeholder variables
best_seen = (0, None)

# Try all possible words, saving the best one seen
for word in possible_words(my_letters):
  score = score_word(word)

  newest_seen = (score, word)
  best_seen = max(best_seen, newest_seen)

Storing all the data together in a single tuple means that assignment and comparison are now handled all at once. This makes it less likely that we will mix up one of the assignments, and makes it clearer what we’re doing.

There is one potential pitfall here: max() picks the tuple with the largest first element (the score in our case), which is what we want. But, if the first elements are the same in both tuples, max() continues through the remaining elements until the tie is broken. So if two words have the same score, max() will then compare the words next, which it does lexically.

To have max() only compare the first element, we can use the key parameter. The key parameter takes a function that is called on each object and returns another object to use in the comparison. We can use it to select just the first entry like so:

# Set up placeholder variables
best_seen = (0, None)

# Try all possible words, saving the best one seen
for word in possible_words(my_letters):
  score = score_word(word)

  newest_seen = (score, word)
  best_seen = max(best_seen, newest_seen, key=lambda x: x[0])

Even Simpler

In the above examples we wanted to save both the score and the word, but what if we only cared about the word that generated the highest score, not the score itself? Then there is an even simpler way!

By default max() uses the standard comparison operator, but we can change that to use our score_word() using the same key argument from above. Then we have:

words = possible_words(my_letters)
best_word = max(words, key=score_word)

Which gives us a very compact (and relatively fool proof) pattern, with all the looping and placeholders pushed into the implementation of max().