Level 3 · Combine conditions with & and |

Combining masks

Real filters have several conditions. pandas combines boolean masks with three operators:

operator	meaning	Python keyword it replaces
&	both are true	and
|	at least one is true	or
~	not	not

completed = orders["status"] == "completed"
paid_ship = orders["shipping_fee"] > 0

orders[completed & paid_ship]
orders[completed | paid_ship]
orders[~completed]               # everything that isn't completed

Why not and and or? Those Python keywords need a single True or False, and a Series of 12,000 values isn’t one. Using them raises ValueError: The truth value of a Series is ambiguous. When you see that error, you’ve used and/or where you needed &/|.

Parentheses are required

& and | bind more tightly than == and >. Without brackets, orders["status"] == "completed" & orders["shipping_fee"] > 0 is read as orders["status"] == ("completed" & orders["shipping_fee"]) > 0 and fails. Wrap every comparison:

(orders["status"] == "completed") & (orders["shipping_fee"] > 0)

Naming each mask, as above, avoids the problem entirely and makes long filters readable.

isin: one of several values

To match any value from a list, isin beats a chain of |:

orders["channel"].isin(["web", "app"])
# same as
(orders["channel"] == "web") | (orders["channel"] == "app")

isin returns a mask like any comparison, so it combines with & the same way. Its opposite is ~orders["channel"].isin([...]), the pandas version of “not in”.

Putting it together with loc

mask = (
    (orders["status"] == "completed")
    & orders["channel"].isin(["web", "app"])
    & (orders["shipping_fee"] > 0)
)
orders.loc[mask, ["order_id", "customer_id", "channel", "shipping_fee"]]

Wrapping the expression in brackets lets you break it across lines, one condition per line, which is how most teams format filters.

Check your filter

Before you hand over a filtered table, check the counts make sense:

mask.sum()                                  # rows kept
orders.loc[mask, "channel"].value_counts()  # only web and app?
orders.loc[mask, "shipping_fee"].min()      # above 0?

These three lines take seconds and catch inverted conditions, wrong values and missing brackets.

query as an alternative

pandas also has orders.query("status == 'completed' and shipping_fee > 0"), where and/or are allowed because the string is parsed by pandas. It’s tidy for quick exploration. Boolean masks are more common in production code because they work with any Python variable or function without string building.

Common mistakes

• Using and/or instead of &/|.
• Missing parentheses around comparisons.
• Using | when the brief needs &: “completed orders from web or app” is completed & (web | app), not completed | web | app.

If you know SQL

isin is IN (...), & is AND, | is OR and ~ is NOT.