Structures Inside Structures

Nothing in the last chapter said a dictionary's values had to be simple. They can be lists. Those lists can hold dictionaries. Those dictionaries can hold more lists, and at no point does Python object or even notice.

club = {    "name": "Robotics",    "room": {"building": "C", "number": 214},    "members": [        {"name": "Amara", "badges": ["solder", "cad"]},        {"name": "Ben", "badges": ["python"]},    ],} print(club["room"]["number"])print(club["members"][0]["badges"][1])

Try itChange the last line so it prints Ben's only badge instead.

Three keys, and the three values are a string, a dictionary, and a list of dictionaries each holding a list. You did not choose this shape — it is what a club actually is, and any structure that described it faithfully would end up about this deep.

A long access path looks intimidating and is entirely mechanical. Read it left to right; each bracket takes the value you are holding and asks it for one thing.

1. 1

   club

   A dictionary of three keys.

2. 2

   club["members"]

   A list of two dictionaries.

3. 3

   club["members"][0]

   The first of those dictionaries.

4. 4

   club["members"][0]["badges"]

   A list of two strings.

5. 5

   club["members"][0]["badges"][1]

   The string "cad".

The bracket type follows from what you are holding, not from where you are: square brackets with a quoted name for a dictionary, square brackets with a number for a list. Write club["members"]["0"] and the list will tell you, fairly, that list indices must be integers.

Nested data usually wants nested loops, and the shape of the loops mirrors the shape of the data closely enough that you can almost read one off the other.

for member in club["members"]:    for badge in member["badges"]:        print(member["name"], "→", badge) # Amara → solder# Amara → cad# Ben → python

The outer loop names one member; the inner one names one badge belonging to that member. Naming the loop variables after what they actually hold is doing real work here — for m in club["members"] would be two characters shorter and considerably harder to follow three levels in.

A comprehension can flatten the same two levels into one list, which is the neatest way to answer "every badge anyone holds".

all_badges = [    badge    for member in club["members"]    for badge in member["badges"]]# ['solder', 'cad', 'python']

When a long path fails, the exception tells you precisely which step could not be taken — and that is more useful than it first appears, because the name it quotes is the one that was missing.

>>> club["members"][0]["email"]KeyError: 'email' >>> club["members"][5]IndexError: list index out of range >>> club["name"]["first"]TypeError: string indices must be integers

Three different failures at three different depths. The first says a key was missing, the second that a position did not exist, and the third that you tried to go down a level from something with no levels left. Read the exception name first: it tells you what kind of thing you were holding when the path ran out.

Reading depth is the common case; creating it has one wrinkle. You cannot write into a level that does not exist yet, so building nested data means creating each level before you fill it.

club = {"name": "Robotics"} club["members"] = []club["members"].append({"name": "Amara", "badges": []})club["members"][0]["badges"].append("solder") print(club)

Try itDelete the line club["members"] = [] and run it again.

There is no rule about how deep is too deep, but there is a reliable symptom: you have gone too far when you can no longer name what a level holds without going back and looking.

Two fixes, both cheap. Pull an intermediate level into a named variable, which turns one unreadable line into two obvious ones. Or, if the shape has become a thing your program uses beyond a passing response, give it a class — that was the argument for bundling data with behaviour, and it applies with more force at four levels than it did at one.

# Hard to read, and harder to debug when it raises:print(club["members"][0]["badges"][1].upper()) # The same thing, with the level that matters given a name:first_member = club["members"][0]print(first_member["badges"][1].upper())