Selection has three common tools:

[] — columns (and boolean row masks)
.loc — labels
.iloc — integer positions

Goal

Pick one column, several columns, a cell, and a slice without triggering SettingWithCopy confusion. Each block is a complete script.

Columns with `[]`

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
print(df["name"])
print()
print(df[["name", "score"]])
print()
print("one column as a frame:")
print(df[["name"]])

One pair of brackets returns a Series. Two brackets (a list) return a DataFrame, even for one column.

Rows and columns with `loc`

loc[row_labels, column_labels] — labels, inclusive slices.

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
print(df.loc[0])
print()
print(df.loc[0:2, ["name", "score"]])
print()
print(df.loc[df["team"] == "A", ["name", "score"]])

Positions with `iloc`

iloc is like NumPy: start inclusive, end exclusive.

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
print(df.iloc[0])
print()
print(df.iloc[0:2, 0:2])
print()
print(df.iloc[[0, 3], [0, 1]])

One cell: `at` / `iat`

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
print(df.at[1, "score"])
print(df.iat[1, 1])

Copy before you edit a slice

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
team_a = df.loc[df["team"] == "A"].copy()
team_a["bonus"] = 5
print(team_a)
print()
print("original unchanged:")
print(df)

Pitfall

df[df["team"] == "A"]["score"] = 100 may not write back to df. Select with loc in one step: df.loc[df["team"] == "A", "score"] = 100.

df = pd.DataFrame(
    {
        "name": ["Ada", "Alan", "Grace", "Linus"],
        "score": [98, 91, 95, 88],
        "team": ["A", "B", "A", "B"],
    }
)
df.loc[df["team"] == "A", "score"] = 100
print(df)

Columns with []

Rows and columns with loc

Positions with iloc