Data structures: lists, tuples, dicts, sets

Data structures are how you hold data in memory. Four of them cover almost all data work.

Lists: ordered, changeable

cities = ["Berlin", "Toronto", "Seattle"]
cities.append("Vancouver")        # add to the end
cities[0]                         # "Berlin"   — first item, counting starts at 0
cities[-1]                        # "Vancouver" — last item
cities[1:3]                       # ["Toronto", "Seattle"] — slice, end not included
len(cities)                       # 4
"Berlin" in cities                # True

A list is a column of values, a batch of rows, a queue of files to process. You will loop over lists constantly.

Tuples: ordered, fixed

point = (49.28, -123.12)
lat, lon = point                  # unpacking

A tuple is a list you cannot change. Use it for things that belong together and never change: coordinates, a (year, month) pair, a database row.

Dictionaries: key to value

The most important structure for data work. A dict is a set of named values. It is exactly how JSON looks when it lands in Python.

offer = {"company": "Shopify", "role": "Data Analyst", "salary_usd": 95000}

offer["company"]                  # "Shopify"
offer["remote"] = True            # add a key
offer.get("bonus", 0)             # 0 — safe read with a default, no error if missing
offer.keys()                      # the keys
offer.items()                     # (key, value) pairs, for looping

An API response is a dict. A config file is a dict. An Airflow task’s parameters are a dict. A list of dicts is a table:

offers = [
    {"company": "Shopify", "salary_usd": 95000},
    {"company": "Wealthsimple", "salary_usd": 110000},
]
offers[1]["company"]              # "Wealthsimple"

Sets: unique values

countries = ["Canada", "Germany", "Canada", "Poland"]
set(countries)                    # {'Canada', 'Germany', 'Poland'}
len(set(countries))               # 3

Sets remove duplicates and test membership fast. set(a) & set(b) gives the common items, set(a) - set(b) the ones only in a. Handy for “which IDs are in the source but not in the target”.

Sets have no order. Never print a set when the order matters. Convert to a sorted list first: sorted(set(countries)).

Which one?

Need	Use
A sequence you will add to or loop over	list
A fixed group of values that belong together	tuple
Named fields, JSON, config	dict
Unique values, membership tests	set