Run any polars aggregation expression against the array’s elements

Source code

Description

This looks similar to $arr$eval(), but the key difference is that $arr$agg() automatically explodes the array if the expression inside returns a scalar (while $arr$eval() always returns an array).

Usage

<Expr>$arr$agg(expr)

Arguments

expr	Expression to run. Note that you can select an element with pl$element(), pl$first(), and more. See Examples.

Value

A polars expression

Examples

library("polars")

df <- pl$DataFrame(a = list(c(1, NA), c(42, 13), c(NA, NA)))$
  cast(pl$Array(pl$Int64, 2))

# The column "null_count" has dtype u32 because `$null_count()` returns a
# scalar for each sub-array. Using `$arr$eval()` instead would return a
# column with dtype arr(u32).
df$with_columns(
  null_count = pl$col("a")$arr$agg(pl$element()$null_count())
)

#> shape: (3, 2)
#> ┌───────────────┬────────────┐
#> │ a             ┆ null_count │
#> │ ---           ┆ ---        │
#> │ array[i64, 2] ┆ u32        │
#> ╞═══════════════╪════════════╡
#> │ [1, null]     ┆ 1          │
#> │ [42, 13]      ┆ 0          │
#> │ [null, null]  ┆ 2          │
#> └───────────────┴────────────┘

# The column "no_nulls" has dtype arr(u32) because the expression doesn't
# guarantee to return a scalar.
df$with_columns(
  no_nulls = pl$col("a")$arr$agg(pl$element()$drop_nulls())
)

#> shape: (3, 2)
#> ┌───────────────┬───────────┐
#> │ a             ┆ no_nulls  │
#> │ ---           ┆ ---       │
#> │ array[i64, 2] ┆ list[i64] │
#> ╞═══════════════╪═══════════╡
#> │ [1, null]     ┆ [1]       │
#> │ [42, 13]      ┆ [42, 13]  │
#> │ [null, null]  ┆ []        │
#> └───────────────┴───────────┘