Live Coding Session

This morning you learned a new code pattern - split-apply-combine - that answers most grouped questions in one line. This afternoon we will look at the questions it does not answer, and at the loop, which is the older and slower tool that handles them.

.groupby() gives you back one number per group, which covers a great deal of what you want. It does not cover printing a paragraph about each site, saving a separate file for each species, running a check that either passes or fails per group, or doing anything at all where the answer is not a number.

For any of those jobs, you take the groups apart and walk through them with a for loop. You met for on Tuesday afternoon over a list. Today the thing you loop over is a set of DataFrames.

We will also spend about ten minutes reading code you are not expected to write. List comprehensions are everywhere in real Python, including in code you may inherit from a colleague later on, and you should be able to look at one and say what it does.

How a live coding session works

We all start from an empty notebook and write everything together, live. There is no code on this page to copy, on purpose.

Cella or Kelly writes a line, explains it, and runs it. You type the same line and run it.
When something breaks, we fix it together. Things will break, and fixing them is the useful part.
If you fall behind, say so! Cella and Kelly would much rather slow down than lose you.
Type the code. Do not copy it from a neighbor’s screen. The typing is the point.

Set up your notebook

Create a new notebook named Session_5D_Loops_Over_Groups.ipynb.
Add a title cell:

# Day 5: Session 5D - When a Pattern Is Not Enough

Date: 09/04/2026

Add the following markdown cells now, in this order, leaving space under each. We will fill in the code as we go, and having the structure in place means your notes stay organized instead of becoming one long scroll.

## 1. What groupby is holding

## 2. Looping over the groups

## 3. One group is a DataFrame

## 4. Building something up as you go

## 5. Reading a list comprehension

## 6. Comprehensions you will meet in the wild

## 7. Which one should I reach for?

Under each heading, add an empty code cell.

Your seven headings are the scaffold for the session. Read the next section before you switch to your notebook, so you know what the afternoon covers.

What we will cover

Looping over groups (about 15 minutes)

what a groupby object is actually holding, and how to look inside it
the two-variable for loop: for name, group in grouped:
why one group is a full DataFrame, and what that lets you do
printing a formatted report, one line per group
collecting results into a list or a dictionary as you go

Reading comprehensions (about 10 minutes)

a for loop that builds a list, written the long way
the same loop as a list comprehension
a dictionary comprehension
a comprehension with an if in it
predicting the output of each before we run it

Choosing (about 5 minutes)

the three questions that tell you whether you need a loop at all
why .groupby(...).mean() is almost always the right answer, and when a loop is the better one

What you should be able to do afterwards

explain what .groupby() is holding before you call an aggregation on it
write a for name, group in grouped: loop and explain what each of the two variables holds
produce something for each group other than a summary number
read a list comprehension and predict what it produces
decide, for a given task, whether it wants the grouped pattern or a loop

You are not being asked to write comprehensions

Sections 5 and 6 are a reading exercise. Nothing this week, including the end-of-day practice and next week’s project, requires you to write a comprehension. We read a few today so that meeting one in somebody else’s notebook later is not the first time you have seen the syntax.