docs/awesome_table_in_pdf.Rmd

---
title: "Create Awesome LaTeX Table with knitr::kable and kableExtra"
author: "Hao Zhu"
date: "`r Sys.Date()`"
output: 
  pdf_document:
    toc: true
    toc_depth: 2
header-includes:
  - \usepackage{booktabs}
  - \usepackage{longtable}
  - \usepackage{array}
  - \usepackage{multirow}
  - \usepackage[table]{xcolor}
  - \usepackage{wrapfig}
  - \usepackage{float}
  - \usepackage{colortbl}
  - \usepackage{pdflscape}
  - \usepackage{tabu}
  - \usepackage{threeparttable}
vignette: >
  %\VignetteIndexEntry{Create Awesome PDF Table with knitr::kable and kableExtra}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

> Please see the package [documentation site](http://haozhu233.github.io/kableExtra) for how to use this package in HTML and more.  

# Overview
The goal of `kableExtra` is to help you build common complex tables and manipulate table styles. It imports the pipe `%>%` symbol from `magrittr` and verbalize all the functions, so basically you can add "layers" to a kable output in a way that is similar with `ggplot2` and `plotly`. 

To learn how to generate complex tables in LaTeX, please visit [http://haozhu233.github.io/kableExtra/awesome_table_in_html.html](http://haozhu233.github.io/kableExtra/awesome_table_in_html.html).

# Installation
```r
install.packages("kableExtra")

# For dev version
# install.packages("devtools")
devtools::install_github("haozhu233/kableExtra")
```

# Getting Started
Here we are using the first few columns and rows from dataset `mtcars`
```{r}
library(knitr)
library(kableExtra)
dt <- mtcars[1:5, 1:6]
```

When you are using `kable()`, if you don't specify `format`, by default it will generate a markdown table and let pandoc handle the conversion from markdown to HTML/PDF. This is the most favorable approach to render most simple tables as it is format independent. If you switch from HTML to pdf, you basically don't need to change anything in your code. However, markdown doesn't support complex table. For example, if you want to have a double-row header table, markdown just cannot provide you the functionality you need. As a result, when you have such a need, you should **define `format` in `kable()`** as either "html" or "latex". *You can also define a global option at the beginning using `options(knitr.table.format = "latex")` so you don't repeat the step everytime.* **In this tutorial, I'll still put `format="latex"` in the function in case users just want to quickly replicate the results.** 

```{r}
options(knitr.table.format = "latex") 
## If you don't define format here, you'll need put `format = "latex"` 
## in every kable function.
```

## LaTeX packages used in this package
If you are using a recent version of rmarkdown, you are recommended to load this package entirely via `library(kableExtra)` or `require(kableExtra)` because this package will load all necessary LaTeX packages, such as `booktabs` or `multirow`, for you automatically. Note that, if you are calling functions from `kableExtra` via `kableExtra::kable_styling()` or if you put `library(kableExtra)` in a separate R file that is **sourced** by the rmarkdown document, these packages won't be loaded. Furthermore, you can suppress this auto-loading behavior by setting a global option `kableExtra.latex.load_packages` to be `FALSE` before you load `kableExtra`.

```{r, eval = FALSE}
# Not evaluated. Ilustration purpose
options(kableExtra.latex.load_package = FALSE)
library(kableExtra)
```

If you are using R Sweave, beamer, R package vignette template, tufte or some customized rmarkdown templates, you can put the following meta data into the `yaml` section. If you are familar with LaTeX and you know what you are doing, feel free to remove unnecessary packages from the list. 

```
header-includes:
  - \usepackage{booktabs}
  - \usepackage{longtable}
  - \usepackage{array}
  - \usepackage{multirow}
  - \usepackage[table]{xcolor}
  - \usepackage{wrapfig}
  - \usepackage{float}
  - \usepackage{colortbl}
  - \usepackage{pdflscape}
  - \usepackage{tabu}
  - \usepackage{threeparttable}
```

## Plain LaTeX
Plain LaTeX table looks relatively ugly in 2017.
```{r}
kable(dt)
```

## LaTeX table with booktabs
Similar with Bootstrap in HTML, in LaTeX, you can also use a trick to make your table look prettier as well. The different part is that, this time you don't need to pipe kable outputs to another function. Instead, you should call `booktabs = T` directly in `kable()`
```{r}
kable(dt, format = "latex", booktabs = T)
```

# Table Styles
`kable_styling` in LaTeX uses the same syntax and structure as `kable_styling` in HTML. However, instead of `bootstrap_options`, you should specify `latex_options` instead. 

## LaTeX options
Similar with `bootstap_options`, `latex_options` is also a charter vector with a bunch of options including `striped`, `hold_position` and `scale_down`.

### Striped
Even though in the LaTeX world, people usually call it `alternative row colors` but here I'm using its bootstrap name for consistency. Note that to make it happen, LaTeX package `xcolor` is required to be loaded. In an environment like rmarkdown::pdf_document (rmarkdown 1.4.0 +), `kable_styling` will load it automatically if `striped` is enabled. However, in other cases, you probably need to import that package by yourself. 
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(latex_options = "striped")
```

### Hold position
If you provide a table caption in `kable()`, it will put your LaTeX tabular in a `table` environment, unless you are using `longtable`. A `table` environment will automatically find the best place (it thinks) to put your table. However, in many cases, you do want your table to appear in a position you want it to be. In this case, you can use this `hold_position` options here. 
```{r}
kable(dt, format = "latex", caption = "Demo table", booktabs = T) %>%
  kable_styling(latex_options = c("striped", "hold_position"))
```

If you find `hold_position` is not powerful enough to literally PIN your table in the exact position, you may want to use `HOLD_position`, which is a more powerful version of this feature. For those who are familar with LaTeX, `hold_position` uses `[!h]` and `HOLD_position` uses `[H]` and the `float` package.

### Scale down
When you have a wide table that will normally go out of the page and you want to scale down the table to fit the page, you can use the `scale_down` option here. Note that, if your table is too small, it will also scale up your table. It was named in this way only because scaling up isn't very useful in most cases. 
```{r}
kable(cbind(dt, dt, dt), format = "latex", booktabs = T) %>%
  kable_styling(latex_options = c("striped", "scale_down"))
```
```{r}
kable(cbind(dt), format = "latex", booktabs = T) %>%
  kable_styling(latex_options = c("striped", "scale_down"))
```

### Repeat header in longtable
In `kableExtra` 0.3.0 or above, a new option `repeat_header` was introduced into `kable_styling`. It will add header rows to longtables spanning multiple pages. For table captions on following pages, it will append *"continued"* to the caption to differentiate. If you need texts other than *"(continued)"* (for example, other languages), you can specify it using `kable_styling(..., repeat_header_text = "xxx")`. If you want to complete replace the table caption instead of appending, you can specify it in the option `repeat_header_method`.
```{r}
long_dt <- rbind(mtcars, mtcars) 

kable(long_dt, format = "latex", longtable = T, booktabs = T, caption = "Longtable") %>%
  add_header_above(c(" ", "Group 1" = 5, "Group 2" = 6)) %>%
  kable_styling(latex_options = c("repeat_header"))
```


## Full width?
If you have a small table and you want it to spread wide on the page, you can try the `full_width` option. Unlike `scale_down`, it won't change your font size. You can use `column_spec`, which will be explained later, together with `full_width` to achieve the best result.
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(full_width = T) %>%
  column_spec(1, width = "8cm")
```

## Position
Table Position only matters when the table doesn't have `full_width`. You can choose to align the table to `center` or `left` side of the page. The default value of position is `center`. 

Note that even though you can select to `right` align your table but the table will actually be centered. Somehow it is very difficult to right align a table in LaTeX (since it's not very useful in the real world?). If you know how to do it, please send out an issue or PR and let me know. 
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(position = "center")
```

Becides these three common options, you can also wrap text around the table using the `float-left` or `float-right` options. Note that, like `striped`, this feature will load another non-default LaTeX package `wrapfig` which requires rmarkdown 1.4.0 +. If you rmarkdown version < 1.4.0, you need to load the package through a customed LaTeX template file. 
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(position = "float_right")
```
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Cras sit amet mauris in ex ultricies elementum vel rutrum dolor. Phasellus tempor convallis dui, in hendrerit mauris placerat scelerisque. Maecenas a accumsan enim, a maximus velit. Pellentesque in risus eget est faucibus convallis nec at nulla. Phasellus nec lacinia justo. Morbi fermentum, orci id varius accumsan, nibh neque porttitor ipsum, consectetur luctus risus arcu ac ex. Aenean a luctus augue. Suspendisse et auctor nisl. Suspendisse cursus ultrices quam non vulputate. Phasellus et pharetra neque, vel feugiat erat. Sed feugiat elit at mauris commodo consequat. Sed congue lectus id mattis hendrerit. Mauris turpis nisl, congue eget velit sed, imperdiet convallis magna. Nam accumsan urna risus, non feugiat odio vehicula eget.

## Font Size
If one of your tables is huge and you want to use a smaller font size for that specific table, you can use the `font_size` option. 
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(font_size = 7)
```

# Column / Row Specification
## Column spec
When you have a table with lots of explanatory texts, you may want to specified the column width for different column, since the auto adjust in HTML may not work in its best way while basic LaTeX table is really bad at handling text wrapping. Also, sometimes, you may want to highlight a column (e.g. a "Total" column) by making it bold. In these scenario, you can use `column_spec()`. You can find an example below. 
```{r}
text_tbl <- data.frame(
  Items = c("Item 1", "Item 2", "Item 3"),
  Features = c(
    "Lorem ipsum dolor sit amet, consectetur adipiscing elit. Proin vehicula tempor ex. Morbi malesuada sagittis turpis, at venenatis nisl luctus a. ",
    "In eu urna at magna luctus rhoncus quis in nisl. Fusce in velit varius, posuere risus et, cursus augue. Duis eleifend aliquam ante, a aliquet ex tincidunt in. ", 
    "Vivamus venenatis egestas eros ut tempus. Vivamus id est nisi. Aliquam molestie erat et sollicitudin venenatis. In ac lacus at velit scelerisque mattis. "
  )
)

kable(text_tbl, format = "latex", booktabs = T) %>%
  kable_styling(full_width = F) %>%
  column_spec(1, bold = T, color = "red") %>%
  column_spec(2, width = "30em")
```

## Row spec
Similar with `column_spec`, you can define specifications for rows. Currently, you can either bold or italiciz an entire row. Note that, similar with other row-related functions in `kableExtra`, for the position of the target row, you don't need to count in header rows or the group labelling rows.

```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling("striped", full_width = F) %>%
  column_spec(7, border_left = T, bold = T) %>%
  row_spec(3:5, bold = T, color = "white", background = "black")
```

## Header Rows
One special case of `row_spec` is that you can specify the format of the header row via `row_spec(row = 0, ...)`. 
```{r}
kable(dt, format = "latex", booktabs = T, align = "c") %>%
  kable_styling(latex_options = "striped", full_width = F) %>%
  row_spec(0, angle = 45)
```



# Cell/Text Specification
Function `cell_spec` is introduced in version 0.6.0 of `kableExtra`. Unlike `column_spec` and `row_spec`, **this function is designed to be used before the data.frame gets into the `kable` function**. Comparing with figuring out a list of 2 dimentional index for targeted cells, this design is way easier to learn and use and it fits perfectly well with `dplyr`'s `mutate` and `summarize` functions. With this design, there are two things to be noted:
* Since `cell_spec` generates raw `HTML` or `LaTeX` code, make sure you remember to put `escape = FALSE` in `kable`. At the same time, you have to escape special symbols including `%` manually by yourself
* `cell_spec` needs a way to know whether you want `html` or `latex`. You can specify it locally in function or globally via the `options(knitr.table.format = "latex")` method as suggested at the beginning. If you don't provide anything, this function will output as HTML by default. 

Currently, `cell_spec` supports features including bold, italic, monospace, text color, background color, align, font size & rotation angle. More features may be added in the future. Please see function documentations as reference. 

## Conditional logic
It is very easy to use `cell_spec` with conditional logic. Here is an example.
```{r, message=FALSE, warning=FALSE}
library(dplyr)
mtcars[1:10, 1:2] %>%
  mutate(
    car = row.names(.),
    # You don't need format = "latex" if you have ever defined options(knitr.table.format)
    mpg = cell_spec(mpg, "latex", color = ifelse(mpg > 20, "red", "blue")),
    cyl = cell_spec(cyl, "latex", color = "white", align = "c", angle = 45, 
                    background = factor(cyl, c(4, 6, 8), 
                                        c("#666666", "#999999", "#BBBBBB")))
  ) %>%
  select(car, mpg, cyl) %>%
  kable("latex", escape = F, booktabs = T, linesep = "")
```

## Visualize data with Viridis Color
This package also comes with a few helper functions, including `spec_color`, `spec_font_size` & `spec_angle`. These functions can rescale continuous variables to certain scales. For example, function `spec_color` would map a continuous variable to any [viridis color palettes](https://cran.r-project.org/web/packages/viridis/vignettes/intro-to-viridis.html). It offers a very visually impactful representation in a tabular format. 

```{r}
iris[1:10, ] %>%
  mutate_if(is.numeric, function(x) {
    cell_spec(x, "latex", bold = T, color = spec_color(x, end = 0.9),
              font_size = spec_font_size(x))
  }) %>%
  mutate(Species = cell_spec(
    Species, "latex", color = "white", bold = T,
    background = spec_color(1:10, end = 0.9, option = "A", direction = -1)
  )) %>%
  kable("latex", escape = F, booktabs = T, linesep = "", align = "c")
```

In the example above, I'm using the `mutate` functions from `dplyr`. You don't have to use it. Base R solutions like `iris$Species <- cell_spec(iris$Species, color = "red")` also works. 

## Text Specification
If you check the results of `cell_spec`, you will find that this function does nothing more than wrapping the text with appropriate HTML/LaTeX formatting syntax. The result of this function is just a vector of character strings. As a result, when you are writing a `rmarkdown` document or write some text in shiny apps, if you need extra markups other than **bold** or *italic*, you may use this function to `r text_spec("color", color = "red")`, `r text_spec("change font size ", font_size = 16)` or `r text_spec("rotate", angle = 30)` your text. 

An aliased function `text_spec` is also provided for a more literal writing experience. The only difference is that in LaTeX, unless you specify `latex_background_in_cell = FALSE` (default is `TRUE`) in `cell_spec`, it will define cell background color as `\cellcolor{}`, which doesn't work outside of a table, while for `text_spec`, the default value for `latex_background_in_cell` is `FALSE`.

```{r}
sometext <- strsplit(paste0(
  "You can even try do do some crazy things like this paragraph. Although ",
  "it seems to be a useless feature at this moment, it is so fun to play ",
  "with that I can't stop. :P"
), " ")[[1]]
text_formatted <- paste(
  text_spec(sometext, "latex", color = spec_color(1:length(sometext), end = 0.9),
            font_size = spec_font_size(1:length(sometext), begin = 5, end = 20)),
  collapse = " ")

# To display the text, type `r text_formatted` outside of the chunk
```
`r text_formatted`

# Grouped Columns / Rows
## Add header rows to group columns
Tables with multi-row headers can be very useful to demonstrate grouped data. To do that, you can pipe your kable object into `add_header_above()`. The header variable is supposed to be a named character with the names as new column names and values as column span. For your convenience, if column span equals to 1, you can ignore the `=1` part so the function below can be written as `add_header_above(c(" ", "Group 1" = 2, "Group 2" = 2, "Group 3" = 2)).
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling() %>%
  add_header_above(c(" " = 1, "Group 1" = 2, "Group 2" = 2, "Group 3" = 2))
```

In fact, if you want to add another row of header on top, please feel free to do so. Also, since kableExtra 0.3.0, you can specify `bold` & `italic` as you do in `row_spec()`.
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling(latex_options = "striped") %>%
  add_header_above(c(" ", "Group 1" = 2, "Group 2" = 2, "Group 3" = 2)) %>%
  add_header_above(c(" ", "Group 4" = 4, "Group 5" = 2)) %>%
  add_header_above(c(" ", "Group 6" = 6), bold = T, italic = T)
```

## Group rows via labeling
Sometimes we want a few rows of the table being grouped together. They might be items under the same topic (e.g., animals in one species) or just different data groups for a categorical variable (e.g., age < 40, age > 40). With the new function `group_rows()` in `kableExtra`, this kind of task can be completed in one line. Please see the example below. Note that when you count for the start/end rows of the group, you don't need to count for the header rows nor other group label rows. You only need to think about the row numbers in the "original R dataframe".
```{r}
kable(mtcars[1:10, 1:6], format = "latex", caption = "Group Rows", booktabs = T) %>%
  kable_styling() %>%
  group_rows("Group 1", 4, 7) %>%
  group_rows("Group 2", 8, 10)
```

In case some users need it, you can define your own gapping spaces between the group labeling row and previous rows. The default value is `0.5em`.
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  group_rows("Group 1", 4, 5, latex_gap_space = "2em")
```

If you prefer to build multiple groups in one step, you can use the short-hand `index` option. Basically, you can use it in the same way as you use `add_header_above`. However, since `group_row` only support one layer of grouping, you can't add multiple layers of grouping header as you can do in `add_header_above`. 
```{r, eval=FALSE}
kable(mtcars[1:10, 1:6], format = "latex", caption = "Group Rows", booktabs = T) %>%
  kable_styling() %>%
  group_rows(index=c(" " = 3, "Group 1" = 4, "Group 2" = 3))
# Not evaluated. The code above should have the same result as the first example in this section.
```

## Row indentation
Unlike `group_rows()`, which will insert a labeling row, sometimes we want to list a few sub groups under a total one. In that case, `add_indent()` is probably more apporiate. 
For advanced users, you can even define your own css for the group labeling.
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  add_indent(c(1, 3, 5))
```

## Group rows via multi-row cell
Function `group_rows` is great for showing simple structural information on rows but sometimes people may need to show structural information with multiple layers. When it happens, you may consider to use `collapse_rows` instead, which will put repeating cells in columns into multi-row cells. If you even need to specify column/row format, use `column_spec` & `row_spec` before you pipe it into `collapse_rows`.

```{r}
collapse_rows_dt <- data.frame(C1 = c(rep("a", 10), rep("b", 5)),
                 C2 = c(rep("c", 7), rep("d", 3), rep("c", 2), rep("d", 3)),
                 C3 = 1:15,
                 C4 = sample(c(0,1), 15, replace = TRUE))
kable(collapse_rows_dt, format = "latex", booktabs = T, align = "c") %>%
  column_spec(1, bold=T) %>%
  collapse_rows(columns = 1:2)
```

```{r}
kable(collapse_rows_dt, format = "latex", align = "c") %>%
  column_spec(1, bold = T, width = "5em") %>%
  collapse_rows(1:2)
```

# Table Footnote
## Notation system
You can also use `add_footnote()` function from this package. You will need to supply a character vector with each element as one footnote. You may select from `number`, `alphabet` and `symbol` for different types of notations. Example are listed below.

### Alphabet
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling() %>%
  add_footnote(c("Footnote 1", "Have a good day."), notation = "alphabet")
```

### Number
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling() %>%
  add_footnote(c("Footnote 1", "Have a good day."), notation = "number")
```

### Symbol
```{r}
kable(dt, format = "latex", booktabs = T) %>%
  kable_styling() %>%
  add_footnote(c("Footnote 1", "Footnote 2", "Footnote 3"), notation = "symbol")
```

## In-table markers
By design, `add_footnote()` will transform any `[note]` to in-table footnote markers. 
```{r}
kable(dt, format = "latex", caption = "Demo Table[note]", booktabs = T) %>%
  kable_styling(latex_options = "hold_position") %>%
  add_header_above(c(" ", "Group 1[note]" = 3, "Group 2[note]" = 3)) %>%
  add_footnote(c("This table is from mtcars", 
                 "Group 1 contains mpg, cyl and disp", 
                 "Group 2 contains hp, drat and wt"), 
               notation = "symbol")
```

# LaTeX Only Features
## Table on a Landscape Page
Sometimes when we have a wide table, we want it to sit on a designated landscape page. The new function `landscape()` can help you on that. Unlike other functions, this little function only serves LaTeX and doesn't have a HTML side.
```{r}
kable(dt, format = "latex", caption = "Demo Table (Landscape)[note]", booktabs = T) %>%
  kable_styling(latex_options = c("hold_position")) %>%
  add_header_above(c(" ", "Group 1[note]" = 3, "Group 2[note]" = 3)) %>%
  add_footnote(c("This table is from mtcars", 
                 "Group 1 contains mpg, cyl and disp", 
                 "Group 2 contains hp, drat and wt"), 
               notation = "symbol") %>%
  group_rows("Group 1", 4, 5) %>%
  landscape()
```

## Use LaTeX table in HTML or Word
If you want to include a LaTeX rendered table in your HTML or Word document, or if you just want to save table as an image, you may consider to use `kable_as_image()`. Note that this feature requires you to have [magick](https://github.com/ropensci/magick) installed (`install.packages("magick")`). Also, if you are planning to use it on Windows, you need to install [Ghostscript](https://www.ghostscript.com/). 

```{r, eval = F}
# Not evaluated. 

# The code below will automatically include the image in the rmarkdown document
kable(dt, "latex", booktabs = T) %>%
  column_spec(1, bold = T) %>%
  kable_as_image()

# If you want to save the image locally, just provide a name
kable(dt, "latex", booktabs = T) %>%
  column_spec(1, bold = T) %>%
  kable_as_image("my_latex_table")
```