Chapter 15, Basic

Review: the 20 questions

The original course ends with 20 multiple choice questions. Here each one has the answer and the **reason**, and a code cell that **proves** the 20 statements: if any were wrong, an `assert` would fail.

The questions and the answers

#	Question	Answer	Where
1	`s = pd.Series([10, 20, 30]); s[1]` prints...	20: label 1 (default index)	Ch. 4
2	`df[["salario"]]` returns...	A DataFrame with one column	Ch. 7
3	With default index, how many rows `df.loc[0:3]` does it return?	4 (the `loc` includes the end)	Ch. 7
4	What shows the missing values by column?	`df.isnull().sum()`	Ch. 9
5	What fills in instead of removing?	`fillna()`	Ch. 9
6	Best tool to replace values with a dictionary?	`map()`	Ch. 11
7	What is wrong with `df[df["a"] > 1 and df["b"] > 3]`?	Missing to use `&` with parentheses in each condition	Ch. 8
8	`astype(int)` breaks with missing values. The fix?	`pd.to_numeric(..., errors="coerce")` (or a nullable type)	Ch. 10
9	Why does a column created with `apply` get filled with `None`?	The function has no `return`	Ch. 11
10	What method matches any value from a list?	`isin`	Ch. 8
11	What method gives the average salary by department?	`groupby`	Ch. 13
12	How many rows `df.iloc[0:5]` does it return?	5 (the `iloc` excludes the end)	Ch. 7
13	With index `id_funcionario`, what `df.iloc[0]` does it return?	The first physical row, whatever the label	Ch. 7
14	What does the `~` in `df[~(df["c"] == "x")]`?	Negates the condition	Ch. 8
15	Why is the `dropna()` without arguments dangerous?	A bad column deletes good rows from others	Ch. 9
16	What does the `duplicated()` compare by default?	All columns	Ch. 10
17	Why is the `to_numeric(coerce)` is safer than the `astype(float)`?	Bad values become `NaN`, and the execution continues	Ch. 10
18	Which `lambda` is valid?	`lambda x: "Alto" if x > 70000 else "Normal"`	Ch. 11
19	`nlargest(5, c)` against `sort_values(c).head(5)`?	The same result, without sorting the entire column	Ch. 12
20	Why `resumo["departamento"]` gives `KeyError` after a `agg`?	The department turned the index	Ch. 13

The test

basico/cap15_revisao.pylinhas 10 a 78

import numpy as np
import pandas as pd

df = pd.read_csv("dados/funcionarios_15.csv")


def levanta(excecao, funcao):
    try:
        funcao()
    except excecao:
        return True
    return False


# 1 a 3
assert pd.Series([10, 20, 30])[1] == 20
assert isinstance(df[["salario"]], pd.DataFrame)
assert len(df.loc[0:3]) == 4

# 4 e 5
assert df.isnull().sum().equals(df.isna().sum()) and int(df.isnull().sum()["cidade"]) == 2
assert df["idade"].fillna(0).isna().sum() == 0 and len(df.dropna()) < len(df)

# 6
assert pd.Series(["a", "b"]).map({"a": "x", "b": "y"}).tolist() == ["x", "y"]

# 7: o `and` do Python não funciona em uma Series
assert levanta(ValueError, lambda: df[(df["experiencia"] > 3) and (df["nota_desempenho"] > 7)])

# 8 e 17: astype(int) quebra, to_numeric(coerce) converte o que dá
assert levanta(ValueError, lambda: df["idade"].astype(int))
texto = pd.Series(["10", "abc", "30"])
assert pd.to_numeric(texto, errors="coerce").tolist()[0] == 10 and pd.to_numeric(texto, errors="coerce").isna().sum() == 1


# 9: sem return, tudo vira None
def sem_return(x):
    if x > 3:
        "alto"


assert df["experiencia"].apply(sem_return).isna().all()

# 10 e 11
assert df["departamento"].isin(["Engenharia", "RH"]).sum() == 5
assert df.groupby("departamento")["experiencia"].mean().notna().all()

# 12 e 13
assert len(df.iloc[0:5]) == 5
por_id = df.set_index("id_funcionario")
assert por_id.iloc[0]["nome"] == "Ana Souza" and levanta(KeyError, lambda: por_id.loc[0])

# 14
assert (~(df["cidade"] == "Curitiba")).sum() == len(df) - 2

# 15 e 16
assert len(df.dropna()) == 9 and int(df.duplicated().sum()) == 0

# 18
assert (lambda x: "Alto" if x > 70000 else "Normal")(80000) == "Alto"

# 19
assert df.nlargest(5, "projetos_concluidos")["projetos_concluidos"].tolist() == \
    df.sort_values("projetos_concluidos", ascending=False).head(5)["projetos_concluidos"].tolist()

# 20
resumo = df.groupby("departamento").agg(media=("experiencia", "mean"))
assert levanta(KeyError, lambda: resumo["departamento"]) and "departamento" in resumo.reset_index().columns
print("as 20 respostas conferem")

Output

as 20 respostas conferem

If you got any question wrong, the best thing is to go back to the chapter of the last column of the table, run the example and change the code to see what happens. Memorizing the answer doesn't help: what remains is having seen Pandas fail.