Allow expressions in data_summary() that return > 1 column summaries

[strengejacke]

Revision

[mattansb]

I think summary dfs should have one row (to be shape consistent), but allow multi-value expressions - just expanding them to columns.

[strengejacke]

Errors:

library(datawizard)

set.seed(123)
d <- data.frame(
  x = rnorm(100, 1, 1),
  y = rnorm(100, 2, 2),
  groups = rep(1:4, each = 25)
)

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  quant_y = quantile(y, c(0.1, 0.9)),
  suffix = c("a", "b", "c")
)
#> Error:
#> ! Argument `suffix` must have the same length as the result of the
#>   regarding summary expression. `suffix` has 3 elements (`a`, `b` and `c`)
#>   for the expression `quantile(x, c(0.25, 0.75))`, which returned 2
#>   values.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  quant_y = quantile(y, c(0.1, 0.9)),
  suffix = list(c("a", "b"), c("c", "d"), c("e", "f"))
)
#> Error:
#> ! If `suffix` is a list of character vectors, it should have the same
#>   length as the number of expressions. `suffix` has 3 elements, but
#>   there are 2 expressions.

data_summary(
  mtcars,
  n = unique(mpg),
  j = c(min(am), max(am)),
  by = c("am", "gear")
)
#> Error:
#> ! Each expression must return the same number of values for each group.
#>   Some of the expressions seem to return varying numbers of values.

^{Created on 2026-03-11 with reprex v2.1.1}

[strengejacke]

@etiennebacher WDYT about the current implementation? It's no longer an additional "off-label-use" functionality like reframe(), but instead enhances the summary functionality to allow expressions that return arbitrary number of summary results, making the function more flexible. Furthermore, the suffix argument is also optional - if not provided, columns with identical names are simply renumbered (see above example).

[mattansb]

If the summary is a named vector can those be used as suffixes instead of the stuffix argument?

[strengejacke]

Which behaviour would you suggest? To make it less complex, we could do:

When the summary expression returns more than one value (and only then)

• suffix has to be a named list, to name multiple columns
• For all non-matching named elements, names of the returned summary are used
• When names are not present, automatic numbering is done.

I would then not allow suffix to be just a character vector that applies to all expressions, because all these options would be quite complex and difficult to understand and document.

[etiennebacher]

Looks good to me but I'd like to wait for @mattansb's opinion before merging.

[mattansb]

Sounds good to me, thanks!

[strengejacke]

Here's the (hopefully) final implementation:

library(datawizard)

set.seed(123)
d <- data.frame(
  x = rnorm(100, 1, 1),
  y = rnorm(100, 2, 2),
  w = rnorm(100, 3, 0.5),
  z = rnorm(100, 4, 3),
  groups = rep(1:4, each = 25)
)

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75))
)
#> quant_x25% | quant_x75% | mean_x | quant_y25% | quant_y50% | quant_y75%
#> -----------------------------------------------------------------------
#>       0.51 |       1.69 |   1.09 |       0.40 |       1.55 |       2.94

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  fivenum = fivenum(y)
)
#> quant_x25% | quant_x75% | mean_x | fivenum_1 | fivenum_2 | fivenum_3
#> --------------------------------------------------------------------
#>       0.51 |       1.69 |   1.09 |     -2.11 |      0.37 |      1.55
#> 
#> quant_x25% | fivenum_4 | fivenum_5
#> ----------------------------------
#>       0.51 |      2.97 |      8.48

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(quant_y = c("_Q1", "_Q2", "_Q3"))
)
#> quant_x25% | quant_x75% | mean_x | quant_y_Q1 | quant_y_Q2 | quant_y_Q3
#> -----------------------------------------------------------------------
#>       0.51 |       1.69 |   1.09 |       0.40 |       1.55 |       2.94

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(quant_x = c("Q1", "Q3"), quant_y = c("_Q1", "_Q2", "_Q3"))
)
#> quant_xQ1 | quant_xQ3 | mean_x | quant_y_Q1 | quant_y_Q2 | quant_y_Q3
#> ---------------------------------------------------------------------
#>      0.51 |      1.69 |   1.09 |       0.40 |       1.55 |       2.94

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.5)),
  quant_w = quantile(w, c(0.25, 0.5)),
  quant_y = quantile(y, c(0.25, 0.5)),
  quant_z = quantile(z, c(0.25, 0.5)),
  suffix = c("_Q1", "_Q2")
)
#> quant_x_Q1 | quant_x_Q2 | quant_w_Q1 | quant_w_Q2 | quant_y_Q1 | quant_y_Q2
#> ---------------------------------------------------------------------------
#>       0.51 |       1.06 |       2.73 |       3.02 |       0.40 |       1.55
#> 
#> quant_x_Q1 | quant_z_Q1 | quant_z_Q2
#> ------------------------------------
#>       0.51 |       1.81 |       3.99

# errors
data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(quant_xy = c("_Q1", "_Q2", "_Q3"))
)
#> Error:
#> ! Names of `suffix` must match the names of the expressions. Suffix
#>   `quant_xy` has no corresponding expression.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(c("Q1", "Q3"), "mean", c("_Q1", "_Q2", "_Q3"))
)
#> Error:
#> ! All elements of `suffix` must have names.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = c("_Q1", "_Q2", "_Q3")
)
#> Error:
#> ! Argument `suffix` must have the same length as the result of the
#>   corresponding summary expression. `suffix` has 3 elements (`_Q1`, `_Q2`
#>   and `_Q3`) for the expression `quantile(x, c(0.25, 0.75))`, which
#>   returned 2 values.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(quant_x = c("_Q1", "_Q2", "_Q3"))
)
#> Error:
#> ! Argument `suffix` must have the same length as the result of the
#>   corresponding summary expression. `suffix` has 3 elements (`_Q1`, `_Q2`
#>   and `_Q3`) for the expression `quantile(x, c(0.25, 0.75))`, which
#>   returned 2 values.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.75)),
  mean_x = mean(x),
  quant_y = quantile(y, c(0.25, 0.5, 0.75)),
  suffix = list(quant_x = c("Q1", "Q3"), quant_y = c("_Q1", "_Q2", "_Q2"))
)
#> Error:
#> ! All suffixes for a single expression must be unique. Suffix for element
#>   `quant_y` has duplicate values.

data_summary(
  d,
  quant_x = quantile(x, c(0.25, 0.5)),
  quant_w = quantile(w, c(0.25, 0.5)),
  quant_y = quantile(y, c(0.25, 0.5)),
  quant_z = quantile(z, c(0.25, 0.5)),
  suffix = c("_Q1", "_Q2", "_Q3")
)
#> Error:
#> ! Argument `suffix` must have the same length as the result of the
#>   corresponding summary expression. `suffix` has 3 elements (`_Q1`, `_Q2`
#>   and `_Q3`) for the expression `quantile(x, c(0.25, 0.5))`, which
#>   returned 2 values.

^{Created on 2026-03-12 with reprex v2.1.1}

[mattansb]

looks great!

[etiennebacher]

Looks good, just a suggestion to reorganize a bit the param description.

Thanks!