Selecting Data

# Single column → Series
df["col"]

# Multiple columns → DataFrame
df[["col1", "col2"]]

# Attribute access (only works when name is a valid Python identifier, no spaces/dots)
df.col

# All columns except one
df.drop(columns=["col"])

# Columns matching a pattern
df.filter(like="score")       # columns containing "score"
df.filter(regex=r"^val_")     # columns matching regex

df.iloc[0]          # first row → Series
df.iloc[-1]         # last row → Series
df.iloc[2:5]        # rows 2, 3, 4 → DataFrame
df.iloc[[0, 2, 4]]  # specific rows

# Row + column position
df.iloc[0, 1]           # scalar at row 0, col 1
df.iloc[0:3, 0:2]       # slice rows and columns
df.iloc[[0, 1], [2, 3]] # specific rows and cols
df.iloc[:, 2]           # all rows, column at position 2
df.iloc[:, -1]          # last column

df.loc["row_a"]           # row by label → Series
df.loc[["row_a", "row_c"]] # multiple rows
df.loc["row_a":"row_c"]   # inclusive label slice

# Label-based row + column
df.loc["row_a", "col1"]           # scalar
df.loc["row_a", ["col1", "col2"]] # row, specific cols
df.loc["row_a":"row_c", "col1":"col3"]  # both sliced
df.loc[:, "col1"]                 # all rows, one column

# With boolean mask
mask = df["age"] > 30
df.loc[mask, "name"]

df.at["row_label", "col_name"]   # label-based scalar
df.iat[2, 3]                     # position-based scalar

# Assignment
df.at["row_a", "score"] = 99
df.iat[0, 1] = 42

mask = df["age"] > 30
df[mask]
df.loc[mask]

# Multiple conditions — use & | ~ with parentheses
df[(df["age"] > 25) & (df["city"] == "NYC")]
df[(df["score"] < 50) | (df["score"] > 90)]
df[~(df["col"].isna())]

# .query() — readable string expressions
df.query("age > 30")
df.query("city == 'NYC' and age > 25")
df.query("score.between(50, 90)")   # method calls in query
threshold = 30
df.query("age > @threshold")        # reference local variable with @

df[df["city"].isin(["NYC", "LA", "Chicago"])]
df[~df["status"].isin(["inactive", "deleted"])]

df[df["score"].between(60, 90)]           # inclusive by default
df[df["score"].between(60, 90, inclusive="left")]  # "left" | "right" | "both" | "neither"

# Create a MultiIndex DataFrame
df = df.set_index(["year", "month"])

# .loc with tuples
df.loc[(2024, 1)]               # exact level values
df.loc[(2024, slice(None))]     # all months of 2024
df.loc[2024]                    # outer level only

# pd.IndexSlice for cleaner syntax
idx = pd.IndexSlice
df.loc[idx[2024, 1:6], :]       # 2024, months 1-6
df.loc[idx[:, 3], "value"]      # all years, month 3

# .xs — cross-section
df.xs(2024, level="year")
df.xs((2024, 6), level=["year", "month"])

df.select_dtypes(include="number")          # all numeric cols
df.select_dtypes(include=["int64", "float64"])
df.select_dtypes(exclude=["object", "category"])
df.select_dtypes(include="datetime")

# Keep values where condition is True; replace False with NaN (or other)
df.where(df > 0)
df.where(df > 0, other=-1)    # replace False with -1

# Inverse: keep where condition is False
df.mask(df < 0)
df.mask(df < 0, other=0)

df.head(10)           # first N rows
df.tail(10)           # last N rows
df.first("7D")        # first 7 days (DatetimeIndex)
df.last("30D")        # last 30 days
df.nth(2)             # 3rd row (works on GroupBy too)
df.sample(n=5)        # random 5 rows
df.sample(frac=0.1)   # random 10%
df.sample(n=5, random_state=42)