DataFrames — Core Operations

import pandas as pd

df = pd.DataFrame({
    "name": ["Rahul", "Priya", "Aman"],
    "age": [28, 25, 32],
    "salary": [60000, 75000, 90000]
})

df.shape    # (3, 3)
df.columns  # Index(["name", "age", "salary"])
df.dtypes   # types of each column

Each dict key becomes a column. shape gives (rows, columns). Always check shape and dtypes immediately after loading data — it tells you if the import worked correctly.

df["name"]           # single column → Series
df[["name","salary"]] # multiple columns → DataFrame

df.iloc[0]           # first row by position
df.iloc[0:3]         # first 3 rows
df.loc[0, "name"]    # row 0, column "name"
df.loc[df["age"] > 27, ["name","salary"]]  # filter + select

.iloc for position-based, .loc for label-based. The most important pattern: df.loc[condition, columns] — filter rows and select columns in one step.

# Add column:
df["bonus"] = df["salary"] * 0.10
df["grade"] = df["age"].apply(lambda x: "senior" if x > 30 else "junior")

# Rename:
df.rename(columns={"name": "full_name"}, inplace=True)

# Drop:
df.drop(columns=["bonus"], inplace=True)
df.drop(columns=["bonus"])  # returns new df, original unchanged

inplace=True modifies the DataFrame in memory. Without it, the operation returns a new DataFrame. Best practice: avoid inplace and reassign — it is clearer and avoids bugs.

df[df["salary"] > 70000]                     # single condition
df[(df["salary"] > 70000) & (df["age"] < 30)]  # AND
df[(df["salary"] > 90000) | (df["age"] < 26)]  # OR
df[df["name"].isin(["Rahul", "Aman"])]         # in list
df[df["name"].str.contains("ra", case=False)]  # text filter

Never use Python and/or in Pandas conditions — use & and |. Each condition in parentheses. This is one of the most common beginner mistakes that causes TypeErrors.

df.describe()     # count, mean, std, min, max, quartiles
df.info()         # column types, non-null counts, memory
df.head(5)        # first 5 rows
df.tail(5)        # last 5 rows
df.sample(3)      # 3 random rows
df.shape          # (rows, cols)
df.isnull().sum() # null count per column

This is your EDA starting sequence. Run all of these on any new dataset before touching a single column. info() reveals hidden issues — dtypes that should be numeric showing as object, for example.