pandas essentials

How do you stop 07030 becoming 7030 when reading a CSV?

Read the column as text: pd.read_csv(f, dtype={'zip': str}).

What does dtype object on a date column mean?

It is stored as text, not a date. Convert with pd.to_datetime.

Why use & instead of and when filtering?

pandas needs element-wise operators. Using and raises "truth value of a Series is ambiguous".

Why parenthesise each filter condition?

& binds tighter than comparison operators, so (a > 1) & (b < 5) is required.

loc versus iloc?

loc is label-based and slice-inclusive; iloc is position-based and slice-exclusive.

What does mean() do with missing values?

Skips them, dividing by the non-null count rather than the row count.

What does df.dropna() with no arguments remove?

Every row containing at least one NaN — often most of a wide table.

agg versus transform after groupby?

agg collapses to one row per group; transform returns a value for every original row.

Why call reset_index() after groupby?

The grouping key sits in the index otherwise, which complicates plotting, merging and export.

How do you catch a merge that duplicated rows?

Compare len() before and after, or pass validate="many_to_one" to raise on a bad relationship.

What does melt do?

Unpivots wide data into long form — turns month columns into month rows.

Why does SettingWithCopyWarning appear?

You assigned into a filtered slice. Add .copy() after filtering, or assign via df.loc.

Why avoid looping over DataFrame rows?

Vectorised column operations are far faster and clearer.

Why pass index=False to to_csv?

Otherwise the index is written as an unnamed extra column.

Why call savefig before show?

show clears the figure, so saving afterwards produces a blank file.