Skip to contents

To more easily insert graphics into body cells, we can use the fmt_image() function. This allows for one or more images to be placed in the targeted cells. The cells need to contain some reference to an image file, either: (1) complete http/https or local paths to the files; (2) the file names, where a common path can be provided via path; or (3) a fragment of the file name, where the file_pattern helps to compose the entire file name and path provides the path information. This should be expressly used on columns that contain only references to image files (i.e., no image references as part of a larger block of text). Multiple images can be included per cell by separating image references by commas. The sep argument allows for a common separator to be applied between images.


  columns = everything(),
  rows = everything(),
  height = "2em",
  sep = " ",
  path = NULL,
  file_pattern = "{x}",
  encode = TRUE



The gt table data object

obj:<gt_tbl> --- required

This is the gt table object that is commonly created through use of the gt() function.


Columns to target

<column-targeting expression> --- default: everything()

Can either be a series of column names provided in c(), a vector of column indices, or a select helper function. Examples of select helper functions include starts_with(), ends_with(), contains(), matches(), one_of(), num_range(), and everything().


Rows to target

<row-targeting expression> --- default: everything()

In conjunction with columns, we can specify which of their rows should undergo formatting. The default everything() results in all rows in columns being formatted. Alternatively, we can supply a vector of row captions within c(), a vector of row indices, or a select helper function. Examples of select helper functions include starts_with(), ends_with(), contains(), matches(), one_of(), num_range(), and everything(). We can also use expressions to filter down to the rows we need (e.g., [colname_1] > 100 & [colname_2] < 50).


Height of image

scalar<character> --- default: "1em"

The absolute height of the image in the table cell. By default, this is set to "1em".


Separator between images

scalar<character> --- default: " "

In the output of images within a body cell, sep provides the separator between each image.


Path to image files

scalar<character> --- default: NULL (optional)

An optional path to local image files (this is combined with all filenames).


File pattern specification

scalar<character> --- default: "{x}"

The pattern to use for mapping input values in the body cells to the names of the graphics files. The string supplied should use "{x}" in the pattern to map filename fragments to input strings.


Use Base64 encoding

scalar<logical> --- default: TRUE

The option to always use Base64 encoding for image paths that are determined to be local. By default, this is TRUE.


An object of class gt_tbl.

Targeting cells with columns and rows

Targeting of values is done through columns and additionally by rows (if nothing is provided for rows then entire columns are selected). The columns argument allows us to target a subset of cells contained in the resolved columns. We say resolved because aside from declaring column names in c() (with bare column names or names in quotes) we can use tidyselect-style expressions. This can be as basic as supplying a select helper like starts_with(), or, providing a more complex incantation like

where(~ is.numeric(.x) && max(.x, na.rm = TRUE) > 1E6)

which targets numeric columns that have a maximum value greater than 1,000,000 (excluding any NAs from consideration).

By default all columns and rows are selected (with the everything() defaults). Cell values that are incompatible with a given formatting function will be skipped over, like character values and numeric fmt_*() functions. So it's safe to select all columns with a particular formatting function (only those values that can be formatted will be formatted), but, you may not want that. One strategy is to format the bulk of cell values with one formatting function and then constrain the columns for later passes with other types of formatting (the last formatting done to a cell is what you get in the final output).

Once the columns are targeted, we may also target the rows within those columns. This can be done in a variety of ways. If a stub is present, then we potentially have row identifiers. Those can be used much like column names in the columns-targeting scenario. We can use simpler tidyselect-style expressions (the select helpers should work well here) and we can use quoted row identifiers in c(). It's also possible to use row indices (e.g., c(3, 5, 6)) though these index values must correspond to the row numbers of the input data (the indices won't necessarily match those of rearranged rows if row groups are present). One more type of expression is possible, an expression that takes column values (can involve any of the available columns in the table) and returns a logical vector. This is nice if you want to base formatting on values in the column or another column, or, you'd like to use a more complex predicate expression.


Using a small portion of metro dataset, let's create a gt table. We will only include a few columns and rows from that table. The lines and connect_rer columns have comma-separated listings of numbers/letters (corresponding to lines served at each station). We have a directory SVG graphics for all of these lines in the package (the path for the image directory can be accessed via system.file("metro_svg", package = "gt")), and the filenames roughly correspond to the data in those two columns. The fmt_image() function can be used with these inputs since the path and file_pattern arguments allow us to compose complete and valid file locations. What you get from this are sequences of images in the table cells, taken from the referenced graphics files on disk.

metro |>
  dplyr::select(name, caption, lines, connect_rer) |>
  dplyr::slice_head(n = 10) |>
  gt() |>
    columns = c(name, caption),
    pattern = "{1}<< ({2})>>"
  ) |>
    locations = cells_body(columns = name),
    pattern = "\\((.*?)\\)",
    replacement = "<br>(<em>\\1</em>)"
  ) |>
  sub_missing(columns = connect_rer, missing_text = "") |>
    columns = lines,
    path = system.file("metro_svg", package = "gt"),
    file_pattern = "metro_{x}.svg"
  ) |>
    columns = connect_rer,
    path = system.file("metro_svg", package = "gt"),
    file_pattern = "rer_{x}.svg"
  ) |>
    name = "Station",
    lines = "Lines",
    connect_rer = "RER"
  ) |>
  cols_align(align = "left") |>
    style = cell_borders(
      sides = c("left", "right"),
      weight = px(1),
      color = "gray85"
    locations = cells_body(columns = lines)
  ) |>
  opt_stylize(style = 6, color = "blue") |>
  opt_all_caps() |>
  opt_horizontal_padding(scale = 1.75)

This image of a table was generated from the first code example in the `fmt_image()` help file.

Function ID


Function Introduced

v0.9.0 (Mar 31, 2023)