Note that huxtable attempts to follow semantic versioning (https://semver.org). Therefore, major version increments reflect backwards-incompatible API changes, not necessarily big changes.
Huxtable 5.0.0 brings numerous changes. For a more user-friendly introduction, see https://hughjonesd.github.io/whats-new-in-huxtable-5.0.0.html.
There are changes to LaTeX output.
\tabcolsep is now set to 0 within huxtable tables, while left and right padding should now take effect even when wrap is FALSE.width is set. If width is set, it is “tabularx”.wrap only matters if width is set. Otherwise, cell wrapping is off.\centerbox macro from the LaTeX “adjustbox” package is used to centre tables. This should improve centring when tables are too wide. You may need to update the LaTeX “adjustbox” package to a recent version. check_latex_dependencies() can inform you about this.As previously signalled, add_colnames has now become TRUE by default in huxtable() and as_huxtable(). Set options(huxtable.add_colnames = FALSE) to go back to the old behaviour.
Newlines in cell contents are now respected (in LaTeX, so long as wrap = TRUE and width has been set).
Huxtable borders have been reworked, fixing some longstanding bugs and adding new features.
Borders are now automatically collapsed. For example:
will set the border in between the columns of jams to 0.4, overwriting the previous value. This is more in line with what you would expect. For example, the following code now does what you probably want:
jams %>%
set_rowspan(2, 1, 3) %>%
set_bottom_border(4, everywhere, 1)
## Type Price
## Strawberry 1.90
## 2.10
## 1.80
## ---------------------------instead of the old behaviour:
jams %>%
set_rowspan(2, 1, 3) %>%
set_bottom_border(4, everywhere, 1)
## Type Price
## Strawberry 1.90
## 2.10
## 1.80
## -----------set_left_border(), set_all_borders() and friends all use a default value of 0.4. So to set a default border, write e.g.
A new brdr() class encapsulates border thickness, style and colour. You can set all properties at once by writing, e.g.:
left_border(ht) and friends return a brdr object. To access the border thickness, write brdr_thickness(left_border(ht)).
Various deprecated items have been removed:
set_*. Instead, use map_*.byrow argument to set_*. Instead, use map_* and by_cols().error_style and pad_decimal arguments in huxreg. Use error_format and align(hx) <- ".".where(), is_a_number() and pad_decimal() functions. Use map_* functions, ! is.na(as.numeric(x)), and align(ht) <- ".".Default padding has been increased to 6 points.
By default, width() is now unset.
By default, wrap() is now TRUE.
every() has been renamed to stripe(), to avoid a clash with purrr::every(). everywhere, evens and odds are still the same.
The little-used ability to set copy_cell_props to a character vector in rbind.huxtable and cbind.huxtable has been removed. You can still set it to FALSE.
add_rows() and add_columns() now always call rbind.huxtable() or cbind.huxtable() and return a huxtable.
Huxtable no longer supports dplyr versions less than 0.7.0 (released mid-2017).
set_cell_properties() has been renamed style_cells(). It is retained as a soft-deprecated alias.
Various themes have been tweaked:
theme_basic() now has bold headers and no header column by default.theme_plain() defaults to position = "centre".theme_striped() uses grey stripes, a white border, and subtler headers.theme_article() has thinner borders.You can now use markdown within table cells.
set_markdown(ht, rows, cols) to turn this on.set_markdown_contents() to set cell contents that will be interpreted as markdown.Huxtable now has the concept of header row and columns.
set_header_rows(ht, row_numbers, TRUE). For columns, use header_cols() or set_header_cols().style_headers(), style_header_cols(), and style_header_rows() to set multiple properties on headers.header_row/col = TRUE set the first row/col to a header, and style all header rows/cols.set_bold() and set_italic() now use a default value of TRUE. So you can write e.g.
Console output in R now shows table position and caption position.
By default, huxtable now sets labels from the current knitr chunk label, if there is one. This is consistent with kable(). In bookdown, you can then do e.g.
Some iris species are shown in \@ref(tab:mytable):
```{r mytable}
as_hux(iris)
```Set options(huxtable.autolabel = FALSE) to turn off this behaviour.
The one-argument form of [ now works for huxtables just as it does for data frames. For example, ht[2:3] selects columns 2 and 3.
New functions fmt_percent() and fmt_pretty() for passing into number_format():
split_across() and split_down() split a huxtable into a list of sub-tables. Headers can be automatically included.
restack_across() and restack_down() split a huxtable, then join it back up. This is useful for making a table fit on a page.
merge_across() and merge_down() merge an area of cells horizontally across rows, or vertically down columns.
New functions set_lr_borders()/_border_colors()/_border_styles()/_padding()
set left and right borders and padding simultaneously. New functions set_tb_borders() etc. set top and bottom properties simultaneously. There are map_ equivalents of all of these.
set_outer_padding() sets padding around a range of cells, similarly to set_outer_borders().
A new table-level property, caption_width(), allows you to set the width of the caption. The default, NA, sets the width equal to the table width.
There are two new themes: theme_compact() and theme_bright().
For huxreg(), a new function tidy_replace() allows you to replace the output of tidy(x) entirely.
huxtable now only sets options(huxtable.knit_print_df = TRUE) if it is attached, not if it is loaded.
huxtable supports dplyr::relocate(), new in dplyr 1.0.0.
Improvements to as_flextable().
Improvements to quick_pptx() (thanks @davidgohel).
Bugfixes for options(huxtable.use_fontspec = TRUE).
Bugfix: add_rownames = "string" now works as promised.
Bugfix: non-ASCII characters are now supported in RTF.
New versions of the gtsummary package will have an as_huxtable() method.
Package texreg on CRAN includes a huxtablereg() function for creating a table of regression outputs.
by_quantiles(), by_equal_groups() and by_colorspace() have gained a colwise argument, which calculates quantiles or colors separately for each column.as_flextable() (thanks @sjewo).merge_repeated_rows() function: merge repeated rows into a single cell.fill and colspan/rowspan arguments for insert_row()/insert_column():
insert_row(ht, "blah", "", "", "", "", ...) can be written insert_row(ht, "blah", fill = "").colspan/rowspan set colspan/rowspan of the first cell in the inserted row/column.to_screen().options('huxtable.latex_use_fontspec') to TRUE to use portable font names in TeX documents, with the LaTeX “fontspec” package.hux_logo().row_height correctly.quick_latex() function.texreg package now includes a huxtablereg function, analogous to huxreg, which outputs a huxtable from a list of regressions. This will be available from the next version of texreg.flextable package and recent versions of Pandoc. (Powerpoint printing requires Pandoc >= 2.4.0.)position() allow text wrapping around tables.set_outer_border_colors() and set_outer_border_styles() functions, like set_outer_borders().broom package, instead using the generics package. If you use huxreg(), you will still need e.g. broom or broom.mixed to provide tidy() and glance() methods for specific models.tidy.tidy_override() and glance.tidy_override() should work even if underlying object has no tidy() or glance() method.echo = TRUE in Rmd pdf_document format.caption() and height() weren’t playing nicely.mutate(..., copy_cell_props = FALSE) was adding a column named copy_cell_props.check_latex_dependencies and install_latex_dependencies gave misleading errors.stars is NULL in huxreg, don’t print a note by default.tinytex when available, allowing autoinstallation of latex packages.map_xxx functions to set properties variably by cell values.by_rows, by_values, by_ranges, by_quantiles etc.tidy_override function to override p values etc. in huxreg.set_contents function to change huxtable contents within dplyr pipes.install_latex_dependencies() or tlmgr if it is not already on your system.huxtable() and friends, add_rownames = "Colname" now sets the name for the new column.set_xxx functions is deprecated, as is the where function. Use map_xxx instead.byrow is soft-deprecated. Use by_cols() instead.wrap=TRUE caused squeezed text in RTF.TeX code was getting escaped by pandoc. To avoid this, if possible, huxtable now adds fenced code blocks round latex tables (see https://pandoc.org/MANUAL.html#extension-raw_attribute). You must add
md_extensions: +raw_attribute
to your YAML header for this to work, and you will need a recent (> 2.0.0) version of Pandoc.
print_rtf, to_rtf and quick_rtf functions.border_style properties to set “solid”, “double”, “dotted” or “dashed” borders. (At present, LaTeX only allows “solid” or “double”.)merge_cells function, an alternative interface to colspan and rowspan.quick_pptx function to print data frames and huxtables into Powerpoint.install_latex_dependencies and check_latex_dependencies utility functions.add_rows and add_columns now accept data frames as arguments.theme_mondrian theme :-Dprint_md now handles bold and italic cells.quick_pdf has new width and height options to change paper size.print_screen and print_md would hang with a wide huxtable.tribble_hux function wrapping tibble::tribble() for readable data input.add_rows and add_columns functions to insert one or more rows into the middle of a huxtable.is_a_number is deprecated.hux_logo() ;-)format method which returns the result of to_html, to_latex etc. as appropriate.to_html printing e.g. “left-border: NA;” in cell CSS.set_all_* not working when huxtable is not attached.as_Workbook failing with non-numeric width.hux_logo was using multiple fonts, fails with Excel output.as_flextable borders not working in cells with colspan > 1.theme_plain theme.add_colnames is going to become TRUE. At present it remains FALSE. Set options("huxtable.add_colnames") to TRUE or FALSE to set the default and avoid warnings in future.quick_* functions now automatically open documents if used interactively. Use open = FALSE to avoid.pad_decimal is deprecated in favour of align(ht) <- ".".huxreg continues with a warning if statistics are unavailable for some models.knit_print.data.frame methods. This means that bare data frames will be pretty-printed via huxtable if the package is loaded.
options("huxtable.knit_print_df") to FALSE if you don’t want this.theme_plain theme. Set options(“huxtable.knit_print_df_theme”) to a different one-argument function if you want to use a different theme.autoformat argument lets huxtable() and as_huxtable() automatically choose alignment and number format based on column type. Set options("huxtable.autoformat") to FALSE to turn off this feature by default.number_format has changed from “%5.3g” to “%.3g”, which no longer space-pads numbers.as_flextable now does not print column names in the header. This matches the standard huxtable behaviour whereby headers are “just another row/column”. To get the old behaviour, use colnames_to_header = TRUE.add_colnames.huxreg was never using built-in confidence intervals.number_format, huxreg, as_hux.table, as_flextable.openxlsx package.quick_xlsx function.set_* column specifications: e.g. set_bold(ht, 1:3, matches(“ab”), TRUE)after argument to insert_column.quick_* functions: when the file argument is not explicitly specified, confirm overwrites manually, or fail if called non-interactively.tidy_args argument to huxreg allows per-model customization of the call to tidy.quick_xxx functions without an explicit file argument throw an error if called non-interactively, and prompt before overwriting files if called interactively.number_format to exponents in scientific notation.quick_pdf error when moving output across filesystems.quick_html, quick_pdf and quick_docx functions to print table-like objects to a new document.to_screen only shows colnames if there are any non-zero-length column names.number_format now applies to any number-like substrings in cells. This means you can include e.g. significance stars in a cell and still use number_format to format the content.number_format is NA, numbers are unchanged.number_format has changed from “%5.2f” to “%5.3g”, which plays nicer with integers but may surprise you by using scientific format for large numbers.outer_borders argument for huxreg. This changes default behaviour slightly.border argument for add_footnote to choose width of footnote’s top border.set_outer_borders now accepts the same row/column arguments as other set_ functions.align, position and caption_pos.huxreg.set_outer_borders function to set borders round a rectangle of cells.to_screen and to_md now respect wrap and col_widths properties.to_screen and to_md gain a min_width argument; to_md gains a logical header argument; to_screen gains a compact argument replacing blank = NULL.crayon package is installed. New huxtable.color_screen option.ReporteRs to officer and flextable. No more RJava horror.error_format argument to huxreg for flexible control over uncertainty estimates.collapse, borders, blank and colname_color from to_screen/print_screen.as_FlexTable is deprecated and calls as_flextable with a warning. header_rows and footer_rows arguments are ignored. If you need this feature, tell me.huxreg:
ci_level is NULL by default. Set it to a number to calculate confidence intervals.error_style is deprecated with a warning in favour of error_format.{stars} not %stars% to display significance levels in the note argument.borders becomes a number specifying border width. Set to 0 for no borders.insert_row and insert_column.latex_float property allows you to change positioning in LaTeX.caption_pos.set_default_properties function to set default properties for new huxtables.huxreg.huxreg function to convert a list of models to a huxtable.subset, and filling in values by row.$<-, [<- and [[<- now work better.set_cell_properties to set multiple properties on cells.evens, odds, everywhere, every(n, from), final(n), where(cond): convenience functions to select rows, columns and cells.ReporteRs.filter and select.guess_knitr_output_format.huxreg.theme_minimal has been renamed theme_basic to avoid a name clash with ggplot2.NEWS.md file to track changes to the package.