Lesson 10: Dictionaries (hash tables)

(c) 2019 Justin Bois. With the exception of pasted graphics, where the source is noted, this work is licensed under a Creative Commons Attribution License CC-BY 4.0. All code contained herein is licensed under an MIT license.

This document was prepared at Caltech with financial support from the Donna and Benjamin M. Rosen Bioengineering Center.

This lesson was generated from an Jupyter notebook. You can download the notebook here.

Mapping objects and dictionaries

A mapping object allows an arbitrary collection of objects to be indexed by an arbitrary collection of values. That's a mouthful. It is easier to understand instead by comparing to a sequence.

Let's take a sequence of two strings, say a tuple containing a first and last name.

name = ('jeffrey', 'lebowski')

We are restricted on how we reference the sequence. I.e., the first name is name[0], and the last name is name[1]. A mapping object could instead be indexed like name['first name'] and name['last name']. You can imagine this would be very useful! A classic example in biology might be looking up amino acids that are coded for by given codons. E.g., you might want


to give you 'Leucine'.

Python's only build-in mapping type is a dictionary. You might imagine that the Oxford English Dictionary might conveniently be stored as a dictionary (obviously). I.e., you would not want to store definitions that have to be referenced like


Rather, you would like to get definitions like this:


Importantly, note that Python dictionaries have no sense of order. Unlike lists and tuples, which are ordered collections, dictionaries have no order. The key is used to fetch the values from memory.

I pause to note that you will notice that the dictionaries seem to preserve the order in which they are created. This is due to a tweak to the implementation of dictionaries in Python 3.6, but as the documentation state, you should not count on this feature. If you really need ordered dictionaries, there are ordered dictionaries in the collections module of the standard library.

Dictionary syntax

The syntax for creating a dictionary, as usual, is best seen through example.

In [1]:
my_dict = {'a': 6, 'b': 7, 'c': 27.6}
{'a': 6, 'b': 7, 'c': 27.6}

A dictionary is created using curly braces ({}). Each entry has a key, followed by a colon, and then the value associated with the key. In the example above, the keys are all strings, which is the most common use case. Note that the items can be of any type; in the above example, they are ints and a float.

We could also create the dictionary using the built-in dict() function, which can take a tuple of 2-tuples, each one containing a key-value pair.

In [2]:
dict((('a', 6), ('b', 7), ('c', 27.6)))
{'a': 6, 'b': 7, 'c': 27.6}

Finally, we can make a dictionary with keyword arguments to the dict() function.

In [3]:
dict(a='yes', b='no', c='maybe')
{'a': 'yes', 'b': 'no', 'c': 'maybe'}

We do not need to have strings as the keys. In fact, any immutable object can be a key.

In [4]:
my_dict = {0: 'zero',
           1.7: [1, 2, 3],
           (5, 6, 'dummy string'): 3.14,
           'strings are immutable': 42}
{0: 'zero',
 1.7: [1, 2, 3],
 (5, 6, 'dummy string'): 3.14,
 'strings are immutable': 42}

However, mutable objects cannot be used as keys.

In [5]:
my_dict = {'immutable is ok': 1,
           ['mutable', 'not', 'ok']: 5}
TypeError                                 Traceback (most recent call last)
<ipython-input-5-ff51262de4d9> in <module>
      1 my_dict = {'immutable is ok': 1,
----> 2            ['mutable', 'not', 'ok']: 5}

TypeError: unhashable type: 'list'

Useful dictionaries in bioinformatics

It might be useful to quickly look up 3-letter amino acid codes. Dictionaries are particularly useful for this.

In [6]:
aa_dict = {'A': 'Ala',
           'R': 'Arg',
           'N': 'Asn',
           'D': 'Asp',
           'C': 'Cys',
           'Q': 'Gln',
           'E': 'Glu',
           'G': 'Gly',
           'H': 'His',
           'I': 'Ile',
           'L': 'Leu',
           'K': 'Lys',
           'M': 'Met',
           'F': 'Phe',
           'P': 'Pro',
           'S': 'Ser',
           'T': 'Thr',
           'W': 'Trp',
           'Y': 'Tyr',
           'V': 'Val'}

Another useful dictionary would contain the set of codons and the amino acids they code for. This is built in the code below using the zip() function we learned before. To see the logic on how this is constructed, see the codon table here.

In [7]:
# The set of DNA bases
bases = ['T', 'C', 'A', 'G']

# Build list of codons
codon_list = []
for first_base in bases:
    for second_base in bases:
        for third_base in bases:
            codon_list += [first_base + second_base + third_base]

# The amino acids that are coded for (* = STOP codon)

# Build dictionary from tuple of 2-tuples (technically an iterator, but it works)
codons = dict(zip(codon_list, amino_acids))

# Show that we did it
{'TTT': 'F', 'TTC': 'F', 'TTA': 'L', 'TTG': 'L', 'TCT': 'S', 'TCC': 'S', 'TCA': 'S', 'TCG': 'S', 'TAT': 'Y', 'TAC': 'Y', 'TAA': '*', 'TAG': '*', 'TGT': 'C', 'TGC': 'C', 'TGA': '*', 'TGG': 'W', 'CTT': 'L', 'CTC': 'L', 'CTA': 'L', 'CTG': 'L', 'CCT': 'P', 'CCC': 'P', 'CCA': 'P', 'CCG': 'P', 'CAT': 'H', 'CAC': 'H', 'CAA': 'Q', 'CAG': 'Q', 'CGT': 'R', 'CGC': 'R', 'CGA': 'R', 'CGG': 'R', 'ATT': 'I', 'ATC': 'I', 'ATA': 'I', 'ATG': 'M', 'ACT': 'T', 'ACC': 'T', 'ACA': 'T', 'ACG': 'T', 'AAT': 'N', 'AAC': 'N', 'AAA': 'K', 'AAG': 'K', 'AGT': 'S', 'AGC': 'S', 'AGA': 'R', 'AGG': 'R', 'GTT': 'V', 'GTC': 'V', 'GTA': 'V', 'GTG': 'V', 'GCT': 'A', 'GCC': 'A', 'GCA': 'A', 'GCG': 'A', 'GAT': 'D', 'GAC': 'D', 'GAA': 'E', 'GAG': 'E', 'GGT': 'G', 'GGC': 'G', 'GGA': 'G', 'GGG': 'G'}

These two dictionaries are particularly useful, so I put them in a little module which we will discuss in the next lesson on packages and modules.

A dictionary is an implementation of a hash table

It is useful to stop and think about how a dictionary works. Let's create a dictionary and look at where the values are stored in memory.

In [8]:
# Create dictionary
my_dict = {'a': 6, 'b': 7, 'c':12.6}

# Find where they are stored

So, each entry in the dictionary is stored at a different location in memory. The dictionary itself also has its own address. So, when I index a dictionary with a key, how does the dictionary know which address in memory to use to fetch the value I am interested in?

Dictionaries use a hash function to do this. A hash function converts its input to an integer. For example, we can use Python's built-in hash function to convert the keys to integers.

In [9]:
hash('a'), hash('b'), hash('c')
(-4447128484049440217, 8787448156945503081, 2487487073396208305)

Under the hood, Python then converts these integers to integers that could correspond to locations in memory. A collection of elements that can be indexed this way is called a hash table. This is a very common data structure in computing. Wikipedia has a pretty good discussion on them.

Given what you know about how dictionaries work, why do you think mutable objects are not acceptable as keys?

Dictionaries are mutable

The title says it all! Dictionaries are mutable. This means that they can be changed in place. For example, if we want to add an element to a dictionary, we use simple syntax.

In [10]:
# Remind ourselves what the dictionary is

# Add an entry
my_dict['d'] = 'Bootcamp is so much fun!'

# Look at dictionary again

# Change an entry
my_dict['a'] = 'I was not satisfied with entry a.'

# Look at it again
{'a': 6, 'b': 7, 'c': 12.6}
{'a': 6, 'b': 7, 'c': 12.6, 'd': 'Bootcamp is so much fun!'}
{'a': 'I was not satisfied with entry a.', 'b': 7, 'c': 12.6, 'd': 'Bootcamp is so much fun!'}

Membership operators with dictionaries

The in and not in operators work with dictionaries, but both only query keys and not values. We see this again by example.

In [11]:
# Make a fresh my_dict
my_dict = {'a': 1, 'b': 2, 'c': 3}

# in works with keys
'b' in my_dict, 'd' in my_dict, 'e' not in my_dict
(True, False, True)
In [12]:
# Try it with values
2 in my_dict

Yup! We get False. Why? Because 2 is not a key in my_dict. We can also iterate over the keys in a dictionary.

In [13]:
for key in my_dict:
    print(key, ':', my_dict[key])
a : 1
b : 2
c : 3

The best, and preferred, method, is to iterate over key,value pairs in a dictionary using the items() method of a dictionary.

In [14]:
for key, value in my_dict.items():
    print(key, value)
a 1
b 2
c 3

Note, however, that like lists, the items that come out of the my_dict.items() iterator are not items in the dictionary, but copies of them. If you make changes within the for loop, you will not change entries in the dictionary.

In [15]:
for key, value in my_dict.items():
    value = 'this string will not be in dictionary.'
{'a': 1, 'b': 2, 'c': 3}

You will, though, if you use the keys.

In [16]:
for key, _ in my_dict.items():
    my_dict[key] = 'this will be in the dictionary.'
{'a': 'this will be in the dictionary.',
 'b': 'this will be in the dictionary.',
 'c': 'this will be in the dictionary.'}

Built-in functions for dictionaries

The built-in len() function and del operation work on dictionaries.

  • len(d) gives the number of entries in dictionary d
  • del d[k] deletes entry with key k from dictionary d

This is the first time we've encountered the del keyword. This keyword is used to delete variables and their values from memory. The del keyword can also be to delete items from lists. Let's see things in practice.

In [17]:
# Create my_list and my_dict for reference
my_dict = dict(a=1, b=2, c=3, d=4)
my_list = [1, 2, 3, 4]

# Print them
print('my_dict:', my_dict)
print('my_list:', my_list)

# Get lengths
print('length of my_dict:', len(my_dict))
print('length of my_list:', len(my_list))

# Delete a key from my_dict
del my_dict['b']

# Delete entry from my_list
del my_list[1]

# Show post-deleted objects
print('post-deleted my_dict:', my_dict)
print('post-deleted my_list:', my_list)
my_dict: {'a': 1, 'b': 2, 'c': 3, 'd': 4}
my_list: [1, 2, 3, 4]
length of my_dict: 4
length of my_list: 4
post-deleted my_dict: {'a': 1, 'c': 3, 'd': 4}
post-deleted my_list: [1, 3, 4]

Note, though, that you cannot delete an item from a tuple, since it's immutable.

In [18]:
my_tuple = (1, 2, 3, 4)
del my_tuple[1]
TypeError                                 Traceback (most recent call last)
<ipython-input-18-db6fdb3f0eac> in <module>
      1 my_tuple = (1, 2, 3, 4)
----> 2 del my_tuple[1]

TypeError: 'tuple' object doesn't support item deletion

Dictionary methods

Dictionaries have several built-in methods in addition to the items() you have already seen. Following are a few of them, assuming the dictionary is d.

method effect
d.keys() return keys
d.pop(key) return value associated with key and delete key from d
d.values() return the values in d

Let's try these out.

In [19]:
my_dict = dict(a=1, b=2, c=3, d=4)

dict_keys(['a', 'b', 'c', 'd'])

Note that this is a dict_keys object. We cannot index it. If, say, we wanted to sort the keys and have them index-able, we would have to convert them to a list.

In [20]:
['a', 'b', 'c', 'd']

This is not a usual use case, though, and be warned that doing then when this is not explicitly what you want can lead to bugs. Now let's try popping an entry out of the dictionary.

In [21]:
In [22]:
{'a': 1, 'b': 2, 'd': 4}

...and, as we expect, key 'c' is now deleted, and its value was returned in the call to my_dict.pop('c'). Finally, we can look at the values.

In [23]:
dict_values([1, 2, 4])

We get a dict_values object, similar to the dict_keys object we got with the my_dict.keys() method.

You can get more information about build-in methods from the Python documentation.

List methods

As you may guess, the dictionary method pop() has an analog that works for lists. (Why don't the dictionary key() and values() methods work for lists?) We take this opportunity to introduce a few more useful list methods. Imagine the list is called s.

method effect
s.pop(i) return value at index i and delete it from the list
s.append(x) Put x at the end of the list
s.insert(i, x) Insert x at index i in the list
s.remove(x) Remove the first occurrence of x from the list
s.reverse() Reverse the order of items in the list

Using dictionaries as kwargs

A nifty feature of dictionaries is that they can be passed into functions as keyword arguments. We covered named keyword arguments in the Intro to functions lesson. In addition to the named keyword arguments, a function can take in arbitrary keyword arguments (not arbitrary non-keyword arguments). This is specified in the function definition by including a last argument with a double-asterisk, **. The kwargs with the double-asterisk get passed in as a dictionary.

In [24]:
def concatenate_sequences(a, b, **kwargs):
    """Concatenate (combine) 2 or more sequences."""
    seq = a + b

    for key in kwargs:
        seq += kwargs[key]
    return seq

Let's try it!

In [25]:
concatenate_sequences('TGACAC', 'CAGGGA', c='GGGGGGGGG', d='AAAATTTTT')

Now, imagine we have a dictionary that contains our values.

In [26]:
my_dict = {'a': 'TGACAC', 
           'b': 'CAGGGA', 
           'c': 'GGGGGGGGG', 
           'd': 'AAAATTTTT'}

We can now pass this directly into the function by preceding it with a double asterisk.

In [27]:

Beautiful! This example is kind of trivial, but you can imagine that it can come in handy, e.g. with large sets of sequence fragments that you read in from a file. We will use **kwargs later in the bootcamp.

Computing environment

In [28]:
%load_ext watermark
%watermark -v -p jupyterlab
CPython 3.7.3
IPython 7.1.1

jupyterlab 0.35.5