Infer column types after cleaning, then optionally layer factor detection

POLIS returns every field as character. This parses character columns to their natural base type (numeric, integer, date, datetime, logical) via readr::type_convert(), protecting identifier-like names and leading-zero codes from coercion, and optionally proposes low-cardinality character columns as factors. The cleaners (clean_afp() etc.) call it with apply = FALSE (base types only) when polis_config(parse_types = TRUE).

Usage

auto_parse_types(
  data,
  max_levels = 50,
  max_unique_ratio = 0.2,
  protect_patterns = c("id$", "uid$", "code$", "ref$", "key$"),
  keep_leading_zero_chars = TRUE,
  apply = TRUE,
  return = c("data", "both", "plan")
)

Arguments

data: A data frame or tibble.
max_levels: Maximum distinct values for a factor candidate. Default 50.
max_unique_ratio: Maximum unique/non-NA ratio for a factor. Default 0.2.
protect_patterns: Regexes for names kept as character. Default c("id$", "uid$", "code$", "ref$", "key$").
keep_leading_zero_chars: Keep a character column when any value is a leading-zero digit string (e.g. "00123"). Default TRUE.
apply: If TRUE (default) apply factor conversions on top of the parsed base types; if FALSE parse base types only.
return: One of "data", "both", "plan". Default "data".

Value

Depending on return: the parsed tibble ("data"), the type plan ("plan"), or list(plan, data) ("both").

Examples

df <- tibble::tibble(id = c("001", "002"), age = c("1", "2"))
auto_parse_types(df, apply = FALSE)
#> # A tibble: 2 × 2
#>   id      age
#>   <chr> <int>
#> 1 001       1
#> 2 002       2