rank Method¶

The rank() method assigns ranks to values within each group, useful for identifying top performers or creating rankings.

Basic Ranking¶

Rank values within groups.

1. Simple Rank¶

import pandas as pd

df = pd.DataFrame({
    'dept': ['A', 'A', 'A', 'B', 'B', 'B'],
    'employee': ['Alice', 'Bob', 'Carol', 'Dave', 'Eve', 'Frank'],
    'salary': [5000, 6000, 4000, 4500, 5500, 4500]
})

df['rank'] = df.groupby('dept')['salary'].rank(ascending=False)
print(df)

  dept employee  salary  rank
0    A    Alice    5000   2.0
1    A      Bob    6000   1.0
2    A    Carol    4000   3.0
3    B     Dave    4500   2.5
4    B      Eve    5500   1.0
5    B    Frank    4500   2.5

2. ascending Parameter¶

# ascending=False: highest value gets rank 1
# ascending=True: lowest value gets rank 1

3. Tie Handling¶

Default: tied values get average rank (2.5 for Dave and Frank).

LeetCode Example: Department Top Salaries¶

Find top 3 salaries per department.

1. Sample Data¶

employee = pd.DataFrame({
    'departmentId': [1, 1, 1, 2, 2, 2],
    'name': ['Alice', 'Bob', 'Carol', 'Dave', 'Eve', 'Frank'],
    'salary': [5000, 6000, 4000, 4500, 5500, 4500]
})

2. Dense Ranking¶

employee['rank'] = employee.groupby('departmentId')['salary'].rank(
    ascending=False,
    method='dense'
)
print(employee)

3. Filter Top 3¶

top_3 = employee[employee['rank'] <= 3]

method Parameter¶

Control how ties are handled.

1. method='average' (Default)¶

# Ties get average of ranks they would occupy
# [100, 100, 80] → [1.5, 1.5, 3.0]

2. method='min'¶

# Ties get lowest rank
# [100, 100, 80] → [1, 1, 3]

3. method='dense'¶

# Ties get same rank, next rank is consecutive
# [100, 100, 80] → [1, 1, 2]

Ranking Examples¶

Common ranking patterns.

1. Percentile Rank¶

df['percentile'] = df.groupby('dept')['salary'].rank(pct=True)

2. Row Number¶

df['row_num'] = df.groupby('dept').cumcount() + 1

3. First/Last in Group¶

df['is_top'] = df.groupby('dept')['salary'].rank(ascending=False) == 1

Comparison with SQL¶

Equivalent SQL window functions.

1. RANK()¶

df.groupby('dept')['salary'].rank(method='min', ascending=False)

2. DENSE_RANK()¶

df.groupby('dept')['salary'].rank(method='dense', ascending=False)

3. ROW_NUMBER()¶

df.groupby('dept').cumcount() + 1