Filtering rows with a boolean mask

The line df[df["cups"] > 100] looks like it's doing something clever. It isn't — it's doing two boring things back to back, and the only reason it reads strangely is that both happen on one line.

First, df["cups"] > 100 compares an entire column at once and hands back a Series of True/False, one per row. That's the boolean mask. Second, putting a boolean Series inside df[...] means "keep the rows where this is True". Split it across two lines and it stops looking like magic:

busy = df["cups"] > 100   # a Series of True/False
df[busy]                  # the rows where it's True

The index does not renumber

This is the part that catches people. Filtering keeps the original row labels. Drop rows 1, 2 and 4 and what's left is labelled 0, 3, 5 — not 0, 1, 2. The rows kept their names when they moved.

So result[1] raises KeyError: there's no row labelled 1 any more. If you want "the second row, whatever it's called", ask by position with result.iloc[1]. If you genuinely want fresh labels, say so explicitly with result.reset_index(drop=True). Both are fine; picking one by accident is not.

Combining conditions

Two traps here, and they bite in the same line of code. Use &, | and ~ — not and, or, not. The Python keywords want a single true-or-false answer and a mask has hundreds, so pandas raises rather than guess. And wrap each condition in parentheses, because & binds tighter than > in Python:

df[(df["city"] == "Delhi") & (df["cups"] > 100)]   # correct
df[df["city"] == "Delhi" & df["cups"] > 100]      # TypeError

For a long list of allowed values, don't chain a dozen |s — df[df["city"].isin(["Delhi", "Pune"])] says the same thing and reads better.