Summarise a data frame's fields for a model

spec_fields() returns one row per column of data with its name, inferred encoding type (quantitative / nominal / ordinal / temporal), R class, distinct-value count, and a few example values. It is the grounding step for LLM plot generation (vplot_ask(), the MCP list_fields tool): handing a model the real vocabulary up front is what prevents hallucinated column names, and vplot_from_spec() then validates against exactly these.

Usage

spec_fields(data, n_examples = 3L)

Arguments

data

A data frame.

n_examples

How many example values to show per field (default 3).

Value

A data frame with columns name, type, class, n_unique, examples.

See also

vplot_from_spec(), spec_schema()

Examples

spec_fields(mtcars)
#>    name         type   class n_unique            examples
#> 1   mpg quantitative numeric       25    21.0, 22.8, 21.4
#> 2   cyl quantitative numeric        3             6, 4, 8
#> 3  disp quantitative numeric       27       160, 108, 258
#> 4    hp quantitative numeric       22        110, 93, 175
#> 5  drat quantitative numeric       22    3.90, 3.85, 3.08
#> 6    wt quantitative numeric       29 2.620, 2.875, 2.320
#> 7  qsec quantitative numeric       30 16.46, 17.02, 18.61
#> 8    vs quantitative numeric        2                0, 1
#> 9    am quantitative numeric        2                1, 0
#> 10 gear quantitative numeric        3             4, 3, 5
#> 11 carb quantitative numeric        6             4, 1, 2