Sorting and Ranking

Sort by Multiple Keys

Build (cat, score, name) tuples from each record and sort them with Python's tuple comparison, which compares left-to-right. With pandas, passing a list of column names to sort_values sorts hierarchically — primary key first, secondary key as a tiebreaker.

By hand

With pandas

df.sort_values(['cat', 'score']) sorts by cat first, using score to break ties within each category. The snapshot shows all three columns in their final sorted order.

naive.py 
# trace: ignore records
records = [
    {'name': 'al', 'cat': 'B', 'score': 80},
    {'name': 'bo', 'cat': 'A', 'score': 55},
    {'name': 'cy', 'cat': 'B', 'score': 70},
    {'name': 'di', 'cat': 'A', 'score': 90},
]
pairs = []
for r in records:
    pairs.append((r['cat'], r['score'], r['name']))
pairs.sort()
result = [p[2] for p in pairs]
print('RESULT:', result)
library.py 
import pandas as pd
from dalib.display import set_display
set_display()

records = [
    {'name': 'al', 'cat': 'B', 'score': 80},
    {'name': 'bo', 'cat': 'A', 'score': 55},
    {'name': 'cy', 'cat': 'B', 'score': 70},
    {'name': 'di', 'cat': 'A', 'score': 90},
]
df = pd.DataFrame(records)
sorted_df = df.sort_values(['cat', 'score'])
print('cats:', sorted_df['cat'].tolist())
print('scores:', sorted_df['score'].tolist())
print('names:', sorted_df['name'].tolist())
print('RESULT:', sorted_df['name'].tolist())

cats: ['A', 'A', 'B', 'B']
scores: [55, 90, 70, 80]
names: ['bo', 'di', 'cy', 'al']
RESULT: ['bo', 'di', 'cy', 'al']

Implementation notes

  • The list of column names is sorted hierarchically left-to-right: first column is primary, second is the tiebreaker, and so on. This mirrors how Python tuple comparison works — the naive half uses the same rule explicitly.
  • Control direction per column with a matching list: df.sort_values(['cat', 'score'], ascending=[True, False]) sorts categories ascending but scores descending within each category.
  • sort_values defaults to kind='quicksort', which is not guaranteed stable. Pass kind='stable' (or kind='mergesort') to preserve the original relative order of rows with equal keys, matching Python's list.sort() guarantee.
  • Cross-reference: sort-two-keys (python-data-basics) for the pure-Python tuple-sort version.