Fill nulls

Fill nulls — `fillNulls`

Replace missing values using a strategy.

Use cases

Backfill a constant like "Unknown" for a missing category.
Impute a numeric column with its mean/median so rows aren't lost.
Carry the last known value forward (ffill) in a time series.

What it does

Replaces nulls in the target columns without removing rows — unlike Drop nulls, every row survives.

Before

order_id	region	amount
1001	North	120.5
1002	null	89
1003	South	null
1004	null	42.25

4 rows · 3 cols

Fill nulls (strategy=constant, value=Unknown, columns=[region])

After

order_id	region	amount
1001	North	120.5
1002	Unknown	89
1003	South	null
1004	Unknown	42.25

4 rows · 3 cols

Configuration

Config key	Type	Required	Description
`strategy`	string	No	`constant` (default), `mean`, `median`, `mode`, `min`, `max`, `zero`, `ffill`, `bfill`
`value`	any	Conditional	Required when `strategy` is `constant`
`columns`	string[]	No	Limit to these columns (otherwise all)

The mean, median, min, and max strategies are computed per column from the non-null values; mode uses the most frequent value; ffill/bfill propagate the previous/next value.

Generated Python code

# strategy: "constant", value: "Unknown", columns: ["region"]
df_2 = df_1.fillna({'region': 'Unknown'})

Tips & common mistakes

A value is only needed for constant. Computed strategies (mean, median, …) ignore it.
mean/median need numeric columns. Restrict columns so the strategy only touches columns it applies to.