Draws a one-line sketch of a column's distribution, as block characters
(or ASCII where the console cannot render them). The point is density of
information: a mean and a standard deviation cannot tell you a column is
bimodal, skewed or has a gap, and this can, in the width of a table cell.
profcols() shows one for every numeric column.
Details
Non-finite values (NA, NaN, Inf) are dropped. The
range of the rest is cut into bins equal-width bins (as
graphics::hist() with those breaks) and each bin's count is drawn
as one of eight Unicode block heights, scaled to the largest bin; an
empty bin is a space, and a bin with any values is at least the lowest
block. When the session is not UTF-8, or the environment variable
RMBL_ASCII_ONLY is set, five ASCII levels (. : - = #) are
used instead. A column with no spread gives one full block in the middle;
no finite values give spaces.
Errors: x that is not numeric; bins that is not one
positive whole number.
See also
profcols(), which includes one
per numeric column.
Examples
set.seed(1)
# A symmetric distribution peaks in the middle.
inlhist(stats::rnorm(1000))
#> [1] "▁▂▄██▇▄▂▁▁"
# A skewed one leans left.
inlhist(stats::rexp(1000))
#> [1] "█▄▂▁▁▁▁▁▁▁"
# Bimodality is visible here and in no single summary number.
inlhist(c(stats::rnorm(500, -3), stats::rnorm(500, 3)))
#> [1] "▁▅█▄▁▁▃█▆▁"
# A constant column has no spread to show.
inlhist(rep(5, 10))
#> [1] " █ "
# more bins for more detail
inlhist(stats::rexp(1000), bins = 30)
#> [1] "█▇▅▄▃▃▂▂▁▁▁▁▁▁▁▁▁▁▁▁▁ ▁ ▁ ▁ ▁"
# text is not a distribution
try(inlhist(c("a", "b")))
#> Error : inlhist: `x` must be numeric, got character