To get started, the easiest is to use the
newsite function to generate a website folder which you can then modify to your heart's content. The command takes one mandatory argument: the name of the folder (which can be
"." if you want to set things up in your current directory). You can optionally specify a template:
julia> newsite("TestWebsite"; template="vela") ✓ Website folder generated at "TestWebsite" (now the current directory). → Use serve() from Franklin to see the website in your browser.
There are a number of simple templates you can choose from and tweak.
Once you have created a new website folder, you can start the live rendering of your website with
julia> serve() → Initial full pass... → Starting the server... ✓ LiveServer listening on http://localhost:8000/ ... (use CTRL+C to shut down)
and navigate in a browser to the corresponding address to preview the website.
The initial call to
newsite generates a folder with the following structure:
. ├── _assets/ ├── _layout/ ├── _libs/ ├── config.md └── index.md
serve the first time, an additional folder is generated:
__site which will contain your full generated website. Among these folders:
the files in the top folder such as
index.mdare the source files for the generated pages, you must have an
index.htmlat the top level but can then use whatever folder structure you want (see further),
you should not modify the content of
__siteas it's generated and any changes you do in there may be silently over-written whenever you modify files elsewhere,
_csscontain auxiliary files supporting your site:
_assets/will contain images, code snippets, etc.,
_css/will contain the style sheets,
_layout/will contain bits of HTML scaffolding for the generated pages,
In this folder,
index.mdwill generate the site's landing page,
pages/page1.mdwould correspond to pages on your website (you can have whatever subfolder structure you want in here),
config.mdallows specifying variables that help steer the page generation, you can also use it to declare global variables or definitions that can then be used on all pages.
index.htmlfile instead of generating it via the
index.md. You will still need to put it in the exact same place and let Franklin copy the files appropriately.
Note that Franklin generates a folder structure in
__site which allows having URLs like
[website]/page1/. The following rules are applied:
the filename is
[path/]index.html, it will be copied over "as is" to
the filename is
[path/]somepage.html, it will be copied to
So for instance, if we ignore auxiliary files and you have
. ├── index.md ├── folder │ └── subpage.md └── page.md
it will lead to
__site ├── index.html ├── folder │ └── subpage │ └── index.html └── page └── index.html
which allows having the following URLs:
To avoid name clashes, refrain from using the following paths where
/ indicates the top directory (website folder):
Also bear in mind that Franklin will ignore
All pages with a given tag can be found (by default) at
[website]/tag/<tag_name>. The tag page location can be changed by setting the
tag_page_path variable in
config.md. For example, setting
@def tag_page_path = "my_tags" would cause tag pages to appear at
serve function can be used to launch a server which will track and render modifications. There are a few useful options you can use beyond the barebone
?serve in your REPL for all options, we list a few noteworthy ones below:
clear=false, whether to erase
__siteand start from a blank slate,
single=false, whether to do a single build pass generating all pages and not start the server.
prerender=false, whether to prerender code blocks and maths (see the optimisation step)
verb=false, whether to show information about which page is being processed etc,
silent=false, whether to suppress any informative messages that could otherwise appear in your console when editing your site, this goes one step further than
verb=falseas it also applies for code evaluation,
eval_all=false, whether to re-evaluate all code blocks on all pages.
Before deploying you may want to verify that links on your website lead somewhere, to do so use the
verify_links(). It will take a few seconds to verify all links on every generated page but can be quite helpful to identify dead links or links with typos:
julia> verify_links() Verifying links... [you seem online ✓] - internal link issue on page index.md: /menu3/.
then after fixing and re-generating pages:
julia> verify_links() All internal and external links verified ✓.
optimize function can
minify all generated HTML and CSS.
?optimize for options.
Those two steps may lead to faster loading pages. Note that in order to run them, you will need a couple of external dependencies as mentioned in the installation section.
The minifier script (an external python lib) used is far from perfect and can break your website, generally speaking, if things work locally but not when you deploy, try disabling it with
Also note that if you use GitHub or GitLab pages, the minification is not needed as these platforms compress the pages themselves.
publish, you can just push your changes and let the relevant action do the rest on the platform. See the section on deployment.
Once you have synched your local folder with a remote repository (see deployment instructions), the
publish function can be called to deploy your website; it essentially:
applies an optional optimisation step (see the previous point),
git add -A; git commit -am "franklin-update"; git push.
?publish for more information.
In any case, before deploying, if you're working on a project website i.e. a website whose root URL will look like
username.gitlab.io/project/ then you should add the following line in your
+++ prepath = "project" +++
publish function will then ensure that all links are fixed before deploying your website.
Note also that the
publish function accepts a
final= keyword to which you can pass any function
() -> nothing to do some final post-processing before pushing updates online. For instance, you can use
lunr is a function exported by Franklin which generates a Lunr search index (see this tutorial for more details).