Description

This is a function for outputting a more advanced tables using HTML. The core philosophy is to bring column and row groups into the table and allow for a dense representation of complex tables. The HTML-output is designed for maximum compatibility with copy-paste functionality into word-processors. For adding styles, see addHtmlTableStyle() and themes setHtmlTableTheme(). Note: If you are using tidyverse and dplyr you may want to check out tidyHtmlTable() that automates many of the arguments that htmlTable requires.

Usage

Arguments

Value

string Returns a string of class htmlTable

Multiple rows of column spanners cgroup

If you want to have a column spanner in multiple levels you can set the cgroup and n.cgroup arguments to a matrix or list.

If the different levels have different number of elements and you have provided a matrix you need to set the ones that lack elements to NA. For instance cgroup = rbind(c("first", "second", NA), c("a", "b", "c")). And the corresponding n.cgroup would be n.cgroup = rbind(c(1, 2, NA), c(2, 1, 2)). for a table consisting of 5 columns. The "first" spans the first two columns, the "second" spans the last three columns, "a" spans the first two, "b" the middle column, and "c" the last two columns.

It is recommended to use list as you will not have to bother with the NA.

If you want leave a cgroup empty then simply provide "" as the cgroup.

The rgroup argument

The rgroup allows you to smoothly group rows. Each row within a group receives an indention of two blank spaces and are grouped with their corresponding rgroup element. The sum(n.rgroup) should always be equal or less than the matrix rows. If less then it will pad the remaining rows with either an empty rgroup, i.e. an "" or if the rgroup is one longer than the n.rgroup the last n.rgroup element will be calculated through nrow(x) - sum(n.rgroup) in order to make the table generating smoother.

The add attribute to rgroup

You can now have an additional element at the rgroup level by specifying the attr(rgroup, 'add'). The value can either be a vector, a list, or a matrix. See vignette("general", package = "htmlTable") for examples.

• A vector of either equal number of rgroups to the number of rgroups that aren't empty, i.e. rgroup[rgroup != ""]. Or a named vector where the name must correspond to either an rgroup or to an rgroup number.

• A list that has exactly the same requirements as the vector. In addition to the previous we can also have a list with column numbers within as names within the list.

• A matrix with the dimension nrow(x) x ncol(x) or nrow(x) x 1 where the latter is equivalent to a named vector. If you have rownames these will resolve similarly to the names to the list/vector arguments. The same thing applies to colnames.

Important knitr-note

This function will only work with knitr outputting HTML, i.e. markdown mode. As the function returns raw HTML-code the compatibility with non-HTML formatting is limited, even with pandoc.

Thanks to the the knitr::knit_print() and the knitr::asis_output() the results='asis' is no longer needed except within for-loops. If you have a knitr-chunk with a for loop and use print() to produce raw HTML you must set the chunk option results='asis'. Note: the print-function relies on the base::interactive() function for determining if the output should be sent to a browser or to the terminal. In vignettes and other directly knitted documents you may need to either set useViewer = FALSE alternatively set options(htmlTable.cat = TRUE).

RStudio's notebook

RStudio has an interactive notebook that allows output directly into the document. In order for the output to be properly formatted it needs to have the class of html. The htmlTable tries to identify if the environment is a notebook document (uses the rstudioapi and identifies if its a file with and Rmd file ending or if there is an element with html_notebook). If you don't want this behavior you can remove it using the options(htmlTable.skip_notebook = TRUE).

Table counter

If you set the option table_counter you will get a Table 1,2,3 etc before each table, just set options(table_counter=TRUE). If you set it to a number then that number will correspond to the start of the table_counter. The table_counter option will also contain the number of the last table, this can be useful when referencing it in text. By setting the option options(table_counter_str = "<b>Table %s:</b> ") you can manipulate the counter table text that is added prior to the actual caption. Note, you should use the sprintf() %s instead of %d as the software converts all numbers to characters for compatibility reasons. If you set options(table_counter_roman = TRUE) then the table counter will use Roman numerals instead of Arabic.

Empty data frames

An empty data frame will result in a warning and output an empty table, provided that rgroup and n.rgroup are not specified. All other row layout options will be ignored.

Options

There are multiple options that can be set, here is a set of the perhaps most used

• table_counter - logical - activates a counter for each table

• table_counter_roman - logical - if true the counter is in Roman numbers, i.e. I, II, III, IV...

• table_counter_str - string - the string used for generating the table counter text

• useViewer - logical - if viewer should be used fro printing the table

• htmlTable.cat - logical - if the output should be directly sent to cat()

• htmlTable.skip_notebook - logical - skips the logic for detecting notebook

• htmlTable.pretty_indentation - logical - there was some issues in previous Pandoc versions where HTML indentation caused everything to be interpreted as code. This seems to be fixed and if you want to look at the raw HTML code it is nice to have this set to TRUE so that the tags and elements are properly indented.

• htmlTableCompat - string - see parameter description

Other

Copy-pasting: As you copy-paste results into Word you need to keep the original formatting. Either right click and choose that paste option or click on the icon appearing after a paste. Currently the following compatibilities have been tested with MS Word 2016:

• Internet Explorer (v. 11.20.10586.0) Works perfectly when copy-pasting into Word

• RStudio (v. 0.99.448) Works perfectly when copy-pasting into Word. Note: can have issues with multi-line cgroups - see bug

• Chrome (v. 47.0.2526.106) Works perfectly when copy-pasting into Word. Note: can have issues with multi-line cgroups - see bug

• Firefox (v. 43.0.3) Works poorly - looses font-styling, lines and general feel

• Edge (v. 25.10586.0.0) Works poorly - looses lines and general feel

Direct word processor opening: Opening directly in Libre Office or Word is no longer recommended. You get much prettier results using the cut-and-paste option.

Google docs: Copy-paste directly into a Google docs document is handled rather well. This seems to work especially well when the paste comes directly from a Chrome browser.

Note that when using complex cgroup alignments with multiple levels not every browser is able to handle this. For instance the RStudio webkit browser seems to have issues with this and a bug has been filed.

As the table uses HTML for rendering you need to be aware of that headers, row names, and cell values should try respect this for optimal display. Browsers try to compensate and frequently the tables still turn out fine but it is not advised. Most importantly you should try to use < instead of < and > instead of >. You can find a complete list of HTML characters here.

Lastly, I want to mention that function was inspired by the Hmisc::latex() that can be an excellent alternative if you wish to switch to PDF-output. For the sibling function tidyHtmlTable() you can directly switch between the two using the table_fn argument.

See Also

addHtmlTableStyle(), setHtmlTableTheme(), tidyHtmlTable(). txtMergeLines(), Hmisc::latex()

Other table functions: tblNoLast(), tblNoNext()

Examples