TopolearnSorted output is how people read results, and top-n queries appear in almost every report.
`df.sort_values(by)` sorts rows by one or more columns and returns a new DataFrame. Pass a list for multiple keys and a matching list for direction:
df.sort_values(["country", "sales"], ascending=[True, False])Missing values go last by default; na_position="first" changes that. The original index labels travel with their rows, so after sorting, df.loc[0] is still the same row it was before, and df.iloc[0] is the new first row. Add ignore_index=True to renumber.
For a single column, the default algorithm is quicksort, which isn't stable; pass kind="stable" if equal values must keep their existing order. Sorting by several columns is always stable.
`df.sort_index()` sorts by the row labels instead, which matters for time series and for fast label slicing. axis=1 sorts the columns by name.
For "top n" questions, `df.nlargest(5, "sales")` and nsmallest are clearer and faster than sorting everything and taking head.
`rank()` assigns positions to values: df["sales"].rank(ascending=False) gives 1 to the biggest. The method= argument decides how ties are treated: "average" (default), "min" (like sports rankings: 1, 2, 2, 4), "dense" (1, 2, 2, 3) or "first".
Sorting by a computed value doesn't need a temporary column: key= takes a function applied to the column before comparing, e.g. df.sort_values("name", key=lambda s: s.str.lower()) for case-insensitive order.
Example
import pandas as pd
df = pd.DataFrame({
"rep": ["ana", "Bo", "cy", "Di", "ed"],
"region": ["N", "S", "N", "S", "N"],
"sales": [120, 340, 340, 90, None],
})
print(df.sort_values("sales", ascending=False).rep.tolist())
# ['Bo', 'cy', 'ana', 'Di', 'ed'] NaN last
print(df.sort_values(["region", "sales"], ascending=[True, False]).rep.tolist())
# ['cy', 'ana', 'ed', 'Bo', 'Di']
print(df.nlargest(2, "sales").rep.tolist()) # ['Bo', 'cy']
print(df["sales"].rank(ascending=False, method="min").tolist())
# [3.0, 1.0, 1.0, 4.0, nan]
print(df.sort_values("rep", key=lambda s: s.str.lower()).rep.tolist())
# ['ana', 'Bo', 'cy', 'Di', 'ed']Remember
sort_values(by, ascending=...)accepts lists of keys and directions- Index labels move with rows; use
ignore_index=Trueto renumber nlargest/nsmallestanswer top-n questions directlyrank(method=...)controls how ties are numbered
Watch out for
- Calling
df.sort_values('x')without assigning the result and wondering whydfis unchanged.