What a Blank Stare During a Pandas Question Really Means

Q: Remove the 'notes' column from df.

Candidate (after a pause):
'I want to drop a column, not a row. I think drop
defaults to rows, so I need to say columns.'

df = df.drop(columns=['notes'])

# equivalent older style:
# df = df.drop('notes', axis=1)

What I hear: the candidate reasoned about the axis
instead of guessing. Syntax recovered from logic.

Q: What does df.groupby('city')['sales'].mean() return?

Candidate: (blank)

Interviewer: Describe df for me first.

Candidate: 'Each row is an order with a city and a
sales amount. Grouping by city collapses all rows of
one city into one value. So I get one number per city,
and the city names become the index. It is a Series,
not a DataFrame, because I selected one column.'

Output shape:
city
Delhi     ...
Mumbai    ...
Name: sales, dtype: float64

import pandas as pd

df = pd.DataFrame({'name': ['Asha', 'Ravi', 'Meena', 'Kabir'],
                   'score': [72, 45, 88, 91]})

passed = df[df['score'] > 50]
print(passed.index.tolist())   # [0, 2, 3]

print(passed.iloc[1]['name'])  # Meena  (2nd position)
print(passed.loc[2, 'name'])   # Meena  (label 2)

# passed.loc[1] raises KeyError, label 1 was filtered out

passed = passed.reset_index(drop=True)
print(passed.index.tolist())   # [0, 1, 2]

import pandas as pd

marks = pd.DataFrame({'maths': [80, 60], 'science': [70, 90]},
                     index=['Asha', 'Ravi'])

print(marks.sum(axis=0))   # down the rows: one total per subject
# maths      140
# science    160

print(marks.sum(axis=1))   # across columns: one total per student
# Asha    150
# Ravi    150

marks['total'] = marks.sum(axis=1)
print(marks)

import pandas as pd

# Q: For each customer, how many orders and total spend?

# 1. Tiny input I can reason about
orders = pd.DataFrame({
    'customer': ['C1', 'C1', 'C2'],
    'amount':   [500, 300, 700]
})

# 2. Target: one row per customer, two columns
# customer  n_orders  total
# C1        2         800
# C2        1         700

# 3. Bridge: groupby with named aggregation
summary = (orders.groupby('customer', as_index=False)
                 .agg(n_orders=('amount', 'size'),
                      total=('amount', 'sum')))
print(summary)

Interviewer: Join customers to orders and keep
customers with no orders.

Candidate (thinking aloud):
'Customers has one row per customer. Orders has many
rows per customer. If I keep all customers, that is a
left join from customers.

Customers with two orders will appear twice. Customers
with no orders appear once with NaN in order columns.'

merged = customers.merge(orders, on='customer_id',
                         how='left')

'Then I can check: merged['order_id'].isna().sum()
gives the count of customers with no orders.'

FOR EACH CARD: predict shape and index, THEN run it

1. df.groupby('city')['sales'].sum()
2. df.groupby('city', as_index=False)['sales'].sum()
3. df.groupby(['city', 'month'])['sales'].sum()
4. df[df['sales'] > 1000].index
5. df.sum(axis=1)  (numeric columns only)
6. df.merge(other, on='id', how='left').shape
7. df.pivot_table(index='city', columns='month',
                  values='sales', aggfunc='sum')
8. df.sort_values('sales').iloc[0]

Write: Series or DataFrame? How many rows?
What is in the index?

import pandas as pd

df = pd.DataFrame({
    'order_id': [1, 2, 3, 4, 5, 6],
    'city':  ['Delhi', 'Delhi', 'Mumbai', 'Pune', 'Mumbai', 'Delhi'],
    'month': ['Jan', 'Feb', 'Jan', 'Jan', 'Feb', 'Feb'],
    'sales': [1200, 800, 1500, 400, 950, 1100]
})

other = pd.DataFrame({
    'id': [1, 2, 2, 7],
    'note': ['a', 'b', 'c', 'd']
})

# Use other with df.rename(columns={'order_id': 'id'})
# for card 6 and predict the row count first.

AFTER EACH CARD

My prediction:   Series / DataFrame,  ___ rows
Actual result:   Series / DataFrame,  ___ rows
Index predicted: ______________
Index actual:    ______________

If wrong, which picture was missing?
[ ] index labels
[ ] axis direction
[ ] groupby keys
[ ] merge duplicates

Write one sentence that would have fixed it.