skimr_table_header_width. The default is to use
the console width, i.e. the value of the width option.haven_labelled columns is now
supported. These columns are summarized using skimmers for the
underlying data, typically either numeric or character.skim_list (most commonly generated by the
partition() function) also inherits from a
listknitr.We’ve made to_long() generic, supporting a more
intuitive interface.
skim_df, it reshapes the output into the V1
long style.skim_tee()Thanks @sethlatimer for suggesting this feature.
tibble.skimr::summarize().Address failed build in CRAN due to lack of UTF-8 support in some platforms.
V2 is a complete rewrite of skimr, incorporating all of
the great feedback the developers have received over the last year. A
big thank you goes to @GShotwell, @akraemer007, @puterleat, @tonyfischetti, @Nowosad, @rgayler, @jrosen48, @randomgambit, @elben10, @koliii, @AndreaPi, @rubenarslan, @GegznaV, @svraka, @dpprdan and to our ROpenSci reviewers @jenniferthompson
and @jimhester for
all of the great support and feedback over the last year. We couldn’t
have done this without you.
For most users using skimr will not change in terms of
visual outputs. However for users who use skimr outputs as
part of a larger workflow the differences are substantial.
skim_dfWe’ve changed the way data is represented within skimr
to closer match expectations. It is now wide by default. This makes
piping statistics much simpler
skim(iris) %>%
dplyr::filter(numeric.sd > 1)This means that the old reshaping functions
skim_to_wide() and skim_to_list() are
deprecated. The latter is replaced with a reshaping function called
partition() that breaks a skim_df into a list
by data type. Similarly, yank() gets a specific data type
from the skim_df. to_long() gets you data that
is closest to the format in the old API.
As the above example suggests, columns of summary statistics are
prefixed by skim_type. That is, statistics from numeric
columns all begin numeric., those for factors all begin
factor., and so on.
We’ve deprecated support for pander() and our
kable() method. Instead, we now support knitr
through the knit_print() API. This is much more seamless
than before. Having a skim_df as the final object in a code
chunk should produce nice results in the majority of RMarkdown
formats.
We’ve deprecated the previous approach customization. We no longer
use skim_format() and skim_with() no longer
depends on a global state. Instead skim_with() is now a
function factory. Customization creates a new skimming function.
my_skim <- skim_with(numeric = sfl(mad = mad))The fundamental tool for customization is the sfl
object, a skimmer function list. It is used within
skim_with() and also within our new API for adding default
functions for new data types, the generic
get_skimmers().
Most of the options set in skim_format are now either in
function arguments or print arguments. The former can be updated using
skim_with, the latter in a call to print(). In
RMarkdown documents, you can change the number of displayed digits by
adding the skimr_digits option to your code chunk.
summary(), and it is now
incorporated into print() methods.focus() is like dplyr::select(), but it
keeps around the columns skim_type and
skim_variable.dplyr
verbs to make sure that they place nice with skimr
objects.skimr has never really focused on performance, it
should do a better job on big data sets with lots of different
columns.skim_without_charts() as a fallback for
when unicode support is not possible.skimr removes the tibble metadata when
generating output. On some platforms, this can lead to all output
getting removed. To disable that behavior, set either
strip_metadata = FALSE when calling print or use
options(skimr_strip_metadata = FALSE).This is likely to be the last release of skimr version 1. Version 2 has major changes to the API. Users should review and prepare for those changes now.
This is likely to be the last release of skimr version 1. Version 2 has major changes to the API. Users should review and prepare for those changes now.
skim_with(.list = mylist) or
skim_with(!!!mylist)rlang:
skim_with(iqr = ~IQR(.x, na.rm = TRUE)).skim_with() to add and remove skimmers at
the same time, i.e. skim_with(iqr = IQR, hist = NULL) works
as expected.spark_line() and spark_bar() are no longer
exportedmin(x) and
max(x) to quantile(x, probs = 0) and
quantile(x, probs = 1). These changes lead to more
predictable behaviors when a column is all NA values.NAs threw
an errordplyr::do()skim_v() is no longer exported. Vectors are now
directly supported via skim.default().kable() and pander() for
skim_df objects.skim_df objects.skim.default().dplyr::group_by())