Level 4 · The ten most expensive products

sort_values

sort_values orders the rows of a DataFrame by one or more columns:

products.sort_values("unit_price")                     # cheapest first
products.sort_values("unit_price", ascending=False)    # most expensive first

Like most pandas methods, it returns a new DataFrame and leaves products as it was.

Several columns, several directions

Pass a list of columns to sort by the first, then the second within ties, and so on. ascending can be a list too, one entry per column:

products.sort_values(["unit_price", "name"], ascending=[False, True])

This reads: highest price first; among products with the same price, A to Z by name. Ties are common: lots of products cost $189 or $159. Without a tie-breaker, their order depends on how the rows happened to be stored, so two runs on slightly different exports could give two different “top 10” lists. Add a tie-breaker whenever you sort for a top-N, and use a unique column (a name or an ID) as the last key so the order is fully determined.

Top N

Sort, then take the first rows:

top10 = products.sort_values(["unit_price", "name"], ascending=[False, True]).head(10)

pandas also has products.nlargest(10, "unit_price"), which is quick and efficient. Note that it keeps ties in their original order rather than letting you choose a tie-breaker, so for reports where the order must be explicit, sort_values is safer.

Reset the index for presentation

After sorting, the index labels stay attached to their original rows: the most expensive product might be labelled 0, the next 17, then 14. For a ranking, a clean 0 to 9 index is what people expect:

top10.reset_index(drop=True)

drop=True throws the old labels away. Without it, they become a new column called index.

Method chaining

Several steps written as a chain, wrapped in brackets so you can put one step per line, read top to bottom like a recipe:

result = (
    products
    .sort_values(["unit_price", "name"], ascending=[False, True])
    .head(10)
    [["name", "category", "unit_price"]]
    .reset_index(drop=True)
)

Selecting the columns before or after head gives the same result; just make sure you sort on columns that are still there.

Sorting text and missing values

Text sorts alphabetically, with uppercase before lowercase ("Zebra" before "apple"), which is one more reason to clean casing first. Missing values go to the end by default, whatever the direction; na_position="first" changes that.

Common mistakes

• Using ascending=False alone with two columns, which sorts the names Z to A as well.
• Taking .head(10) before sorting, which gives the first ten rows of the table, not the top ten.
• Forgetting drop=True in reset_index and getting an extra index column.

If you know SQL

This is ORDER BY unit_price DESC, name ASC LIMIT 10.