tidyselect and across()

Selectors

Selectors pick columns by name, pattern, or type. They work in select(), relocate(), rename_with(), arrange(), distinct(), group_by(), count(), across(), pick(), and the tidyr functions.

where() accepts 'numeric', 'integer', 'double', 'character', 'logical', 'factor', 'date', 'datetime', and 'list', or a Phibble\Type. 'numeric' selects integer and double columns.

across()

across(cols, fns, names:) applies functions to several columns inside mutate() or summarize(). Grouping columns are never selected. fns is one function or an associative array of named functions. Output names default to {col} for one function and {col}_{fn} for several.

$schools = DataFrame::fromCsv('schools.csv') |> group_by('school');

// Phplyr helpers can be passed by name; each gets the column name.
echo $schools |>
    summarize(across(where('numeric'), ['avg' => mean(...), 'sd' => sd(...)])) |>
    head(3);

// Other closures get the column's values (summarize) or each value (mutate).
echo "\n", $schools |>
    summarize(across(['study_hours', 'score'], fn (array $v) => round(\max($v) - \min($v), 1), names: 'range_{col}')) |>
    head(3);
echo "\n", $schools |> ungroup() |> mutate(across(ends_with('score'), fn ($v) => round($v))) |> head(3);

$schools = phplyr(DataFrame::fromCsv('schools.csv'))->groupBy('school')->get();

// Phplyr helpers can be passed by name; each gets the column name.
echo phplyr($schools)
    ->summarize(across(where('numeric'), ['avg' => mean(...), 'sd' => sd(...)]))
    ->head(3);

// Other closures get the column's values (summarize) or each value (mutate).
echo "\n", phplyr($schools)
    ->summarize(across(['study_hours', 'score'], fn (array $v) => round(\max($v) - \min($v), 1), names: 'range_{col}'))
    ->head(3);
echo "\n", phplyr($schools)->ungroup()->mutate(across(ends_with('score'), fn ($v) => round($v)))->head(3);

# A tibble: 3 × 7
  school study_hours_avg study_hours_sd prior_score_avg prior_score_sd score_avg
  <chr>            <dbl>          <dbl>           <dbl>          <dbl>     <dbl>
1 schoo…            6.60           3.26            63.0           12.3      72.9
2 schoo…            6.20           3.57            59.3           13.3      72.9
3 schoo…            6.52           3.34            56.1           11.3      69.0
# ℹ 1 more variable: score_sd <dbl>
# A tibble: 3 × 3
  school    range_study_hours range_score
  <chr>                 <dbl>       <dbl>
1 school-01              10.3        30.9
2 school-02              10.8        34.4
3 school-03              10.3        35.7
# A tibble: 3 × 4
  school    study_hours prior_score score
  <chr>           <dbl>       <dbl> <dbl>
1 school-01        11.6          77    78
2 school-01         2.7          52    67
3 school-01        10.0          79    88

What each function receives:

• A phplyr helper passed as a first-class callable, such as mean(...), is called with the column name.
• In summarize(), other closures receive the column’s values as a list.
• In mutate(), other closures receive one value at a time, or the whole column if the first parameter is typed array.
• A closure may return a helper’s result, as in fn ($v) => mean($v, naRm: true), which is then evaluated on the group.

Inside any of these functions, cur_column() returns the name of the column being processed.

if_any(cols, fn) and if_all(cols, fn) test a predicate across columns, in filter() or as a logical column in mutate(). They use R’s NA logic, so a predicate may return null for NA. With no columns selected, if_any() is false and if_all() is true, as in dplyr 1.1. pick(...) returns the selected columns of the current group as a table, for use inside closures.