Franklin Demos

Click here to see the source

This website is meant to be a quick way to show how to do stuff that people ask (or that I thought would be a nice demo), it will complement the official documentation.

It's not meant to be beautiful, rather just show how to get specific stuff done. If one block answers one of your question, make sure to check the source to see how it was done. The ordering is reverse chronological but just use the table of contents to guide you to whatever you might want to explore.

Note: an important philosophy here is that if you can write a Julia function that would produce the HTML you want, then write that function and let Franklin call it.

Note 2: the numbering in georgian script in the table of content is on purpose (though for no particularly good reason other than that it looks nice... 🇬🇪)

  1. (019) From Dataframe to HTML table
  2. (018) collapsible block
  3. (017) making cells work in their own path
  4. (016) using WGLMakie + JSServe
  5. (015) Using Weave
  6. (014) Using MathJax
  7. (013) Inserting Markdown in Markdown
  8. (012) Dates
  9. (011) showing type information
  10. (010) clipboard button for code blocks
  11. (009) custom environment for TikzCD
  12. (008) (custom) environments and commands
    1. Customise with Julia code
  13. (007) delayed hfun
  14. (006) code highlighting
  15. (005) pagination
  16. (004) use Latexify.jl
  17. (003) styling of code output blocks
  18. (002) code block scope
  19. (001) how to load data from file and loop over rows
    1. Approach 1, with a hfun
    2. Approach 2, with a page variable and a for loop

(019) From Dataframe to HTML table

If you have some data that you would like to manipulate and render nicely in Franklin, you can use the following snippet relying on DataFrames.jl and PrettyTables.jl.

The following Dataframe:

val = [1,2,3,4]
tag = ['a','b','c','d']
math = ["\$x\$", "\$y^2\$", "\$\\sqrt{z}\$", "\$\\Omega\$"]
DataFrame(; val, tag, math)

will be rendered as:

val tag math
1 a \(x\)
2 b \(y^2\)
3 c \(\sqrt{z}\)
4 d \(\Omega\)

This done via a hfun_render_table which can be found in utils.jl.

(018) collapsible block

How to make a section expand when clicked, so that content is initially hidden? (Based on this html guide.)

~~~<button type="button" class="collapsible">~~~ #1 ~~~</button><div class="collapsiblecontent">~~~ #2 ~~~</div>~~~

/* Style the button that is used to open and close the collapsible content */
 .collapsible {
  background-color: #eee;
  color: #444;
  cursor: pointer;
  padding: 18px;
  width: 100%;
  border: none;
  text-align: left;
  outline: none;
  font-size: inherit;

/* Add a background color to the button if it is clicked on (add the .active class with JS), and when you move the mouse over it (hover) */
.active, .collapsible:hover {
  background-color: #ccc;

/* Style the collapsible content. Note: hidden by default */
.collapsiblecontent {
  padding: 0 18px;
  display: none;
  overflow: hidden;
  background-color: #f1f1f1;

  var coll = document.getElementsByClassName("collapsible");
  var i;

  for (i = 0; i < coll.length; i++) {
    coll[i].addEventListener("click", function() {
      var content = this.nextElementSibling;
      if ( === "block") { = "none";
      } else { = "block";

With these definitions, the expandible code sections could be added!

In the content part you can have latex: \(x^2\),


And all other stuff processed by Franklin!

(017) making cells work in their own path

Currently if you're saving a figure in a code block, you need to specify where to place that figure, if you don't it will go in the current directory which is the main site directory, typically you don't want that, so one trick is to use the @OUTPUT macro like so:

using PyPlot
plot(rand(5), rand(5))
savefig(joinpath(@OUTPUT, "ex_outpath_1.svg"))

and that directory is among the ones that are automatically checked when you use the \fig command (e.g. \fig{ex_outpath_1.svg}):

if you're writing tutorials and have lots of such figures and you don't want your readers to see these weird @OUTPUT or you can't be bothered to add #hide everywhere, you can set a variable auto_code_path to true (either locally on a page or globally in your config), what this will do is that each time a cell is executed, Julia will first cd to the output path. So the above bit of code now reads:

using PyPlot
plot(rand(5), rand(5), color="red")

and like before just use \fig{ex_outpath_2.svg}:

Note: since this was meant to be a non-breaking change, auto_code_path is set to false by default, you must set it yourself to true in your if you want this to apply everywhere.

(016) using WGLMakie + JSServe

This page shows an example using WGLMakie + JSServe. It assumes you're familiar with these two libraries and that you have the latest version of each.

Note that it requires WebGL to work which might not be enabled on all browsers.

(015) Using Weave

Here's a page where the content is generated from a Weave.jl.

(014) Using MathJax

If you prefer MathJax over KaTeX for maths rendering, you can use that. For now this is not fully supported and so taking this path may lead to issues, please report them and help fixing those is welcome.

Head to this page for a demo and setup instructions.

(013) Inserting Markdown in Markdown

Let's say you have a file and you would like to include it in another page as if it had been written there in the first place. This is pretty easy to do. You could for instance use the following function:

function hfun_insertmd(params)
  rpath = params[1]
  fullpath = joinpath(Franklin.path(:folder), rpath)
  isfile(fullpath) || return ""
  return read(fullpath, String)

One thing to note is that all .md files in your folder will be considered as potential pages to turn into HTML, so if a .md file is meant to exclusively be used "inserted", you should remove it from Franklin's reach by adding it to the ignore global variable putting something like this in your

@def ignore = ["path/to/"]

Here's an example with the insertion of the content of a file foo/; the result of {{insertmd foo/}} is:

This is some content in a file, it can contain whatever Franklin-compatible markdown as you want \(\alpha\beta\):

\[ \exp(i\pi) + 1 = 0 \]

or code for instance

for i in 1:5

You can look at utils.jl for the definition of the hfun (same as above), at to see how it's called and at foo/ for the content file. Finally you can also check out the file to see how the content page is ignored.

(012) Dates

The date of last modification on the page is kept in the fd_mtime_raw internal page variable, there is also a pre-formatted fd_mtime.

Last modified: 2024-04-16 or April 16, 2024

(011) showing type information

This is a short demo following a discussion on Slack, it shows three things:

s = "hello"
struct T; v::Int; end
    Dict(:a => T(1)),
    Dict(:b => T(2)),
2-element Vector{Dict{Symbol, T}}:
 Dict(:a => T(1))
 Dict(:b => T(2))

Here's another cell


Suppressed output:

X = randn(2, 3);

(010) clipboard button for code blocks

It's fairly easy to add a "copy" button to your code blocks using a tool like clipboard.js. In fact on this demo page, as you can see, there is a copy button on all code blocks. The steps to reproduce this are:

<script src="/libs/clipboard.min.js"></script>

and that's it 🏁.

(009) custom environment for TikzCD

Following up on #008, here's a custom environment for Tikz diagrams using the TikzPictures.jl package.

Let's first see what you get for your effort:

Cool! Modulo a div class that shrinks the image a bit, the code that was used here is very nearly a copy-paste from an example in the tikz-cd docs, the only difference is one additional bracket with the file name (here tcd1):

A \arrow[r, "\phi"] \arrow[d, red]
  & B \arrow[d, "\psi" red] \\
  C \arrow[r, red, "\eta" blue]
  & D

The corresponding env_tikzcd function is in the utils.jl file and is quite simple.

(008) (custom) environments and commands

You can define new commands and new environments using essentially the same syntax as in LaTeX:


The first one allows to define a command that you can call as \command{...} and the second one an environment that you can call as


In both cases you can have a number of arguments (or zero) and the output is reprocessed by Franklin (so treated as Franklin-markdown). Here are a few simple examples:

Result: \hello.

Result: Hello!.

\newcommand{\red}[1]{\html{<span style="color:red">#1</span>}}
Result: \red{hello!}.

Result: hello!.

  \html{<div style="text-align:center">}
Result: \begin{center}This bit of text is in a centered div.\end{center}


This bit of text is centered.
Result: \begin{figure}{A koala eating a leaf.}![](/assets/koala.jpg)\end{figure}


A koala eating a leaf.

Customise with Julia code

Much like hfun_*, you can have commands and environments be effectively defined via Julia code. The main difference is that the output will be treated as Franklin-markdown and so will be reprocessed by Franklin (where for a hfun, the output is plugged in directly as HTML).

In both cases, a single option bracket is expected, no more no less. It can be empty but it has to be there. See also the docs for more information.

Here are two simple examples (see in utils.jl too):

function lx_capa(com, _)
  # this first line extracts the content of the brace
  content = Franklin.content(com.braces[1])
  output = replace(content, "a" => "A")
  return "**$output**"
Result: \capa{Baba Yaga}.

Result: BAbA YAgA.

function env_cap(com, _)
  option = Franklin.content(com.braces[1])
  content = Franklin.content(com)
  output = replace(content, option => uppercase(option))
  return "~~~<b>~~~$output~~~</b>~~~"
Result: \begin{cap}{ba}Baba Yaga with a baseball bat\end{cap}


BaBA Yaga with a BAseBAll BAt

Of course these are toy examples and you could have arrived to the same effect some other way. With a bit of practice, you might develop a preference towards using one out of the three options: hfun_*, lx_* or env_* depending on your context.

(007) delayed hfun

When you call serve(), Franklin first does a full pass which builds all your pages and then waits for changes to happen in a given page before updating that page. If you have a page A and a page B and that the page A calls a function which would need something that will only be defined once B is built, you have two cases:

  1. A common one is to have the function used by A require a local page variable defined on B; in that case just use pagevar(...). When called, it will itself build B so that it can access the page variable defined on B. This is usually all you need.

  2. A less common one is to have the function used by A require a global page variable such as, for instance, the list of all tags, which is only complete once all pages have been built. In that case the function to be used by A should be marked with @delay hfun_...(...) so that Franklin knows it has to wait for the full pass to be completed before re-building A now being sure that it will have access to the proper scope.

The page foo has a tag foo and a page variable var; let's show both use cases; see utils.jl to see the definition of the relevant functions.

Case 1 (local page variable access, var = 5 on page foo):

var read from foo is 5

Case 2 (wait for full build, there's a tag index here and a tag foo on page foo):

tags: { index foo }

(006) code highlighting

Inserting highlighted code is very easy in Franklin; just use triple backquotes and the name of the language:

struct Point{T}

The highlighting is done via highlight.js, it's important to understand there are two modes:

The first one only has support for selected languages (by default: css, C/AL, C++, yaml, bash, ini, TOML, markdown, html, xml, r, julia, julia-repl, plaintext, python with the minified github theme), you can change this by going to the highlightjs and making your selection.

The recommendation is to:

Here are some more examples with languages that are in the default list:

# some R
a <- 10
while (a > 4) {
  cat(a, "...", sep = "")
  a <- a - 1
# some Python
def foo():
  print("Hello world!")
// some c++
#include <iostream>

int main() {
  std::cout << "Hello World!";
  return 0;
/* some CSS */
p {
  border-style: solid;
  border-right-color: #ff0000;

(005) pagination

Pagination works with {{paginate list num}} where list is a page variable with elements that you want to paginate (either a list or tuple), and num is the number of elements you want per page. There are many ways you can use this, one possible way is to wrap it in a command if you want to insert this in an actual list which is what is demoed here:

Now observe that

(004) use Latexify.jl

Latexify produces a LaTeX string which should basically be passed to KaTeX. To do that you need to recuperate the output, extract the string and pass it into a maths block.

Here there's a bug with \begin{equation} in Franklin (issue #584) which is why I'm replacing those with $$ but it should be fixed in the near future so that you wouldn't have to use these two "replace" lines:

using Latexify
empty_ary = Array{Float32, 2}(undef, 2, 2)
ls = latexify(empty_ary) # this is an L string
\[ \left[ \begin{array}{cc} -1.1548718e-33 & -1.1549056e-33 \\ 4.5884e-41 & 4.5884e-41 \\ \end{array} \right] \]

(003) styling of code output blocks

At the moment (August 2020) no particular class is added on an output (see #531); you can still do something similar by adding a @@code-output (or whatever appropriate name) around the command that extracts the output and specify this in your css (see extras.css):

x = 7

If you find yourself writing that a lot, you should probably define a command like

\newcommand{\prettyshow}[1]{@@code-output \show{#1} @@}

and put it in your file so that it's globally available.


(002) code block scope

On a single page all code blocks share their environment so

x = 5


y = x+2

(001) how to load data from file and loop over rows

This was asked on Slack with the hope it could mimick the Data Files functionality of Jekyll where you would have a file like

Eric Mill,konklone
Parker Moore,parkr
Liu Fengyun,liufengyun

and you'd want to loop over that and do something with it.

Relevant pieces:

Approach 1, with a hfun

NameGitHub alias
Eric Millkonklone
Parker Mooreparkr
Liu Fengyunliufengyun

Approach 2, with a page variable and a for loop

Writing the following

{{for (name, alias) in members_from_csv}}
    <a href="{{alias}}">{{name}}</a>