hugo/docs/content/tutorials/github_pages_blog.md

---
author: Spencer Lyon
date: 2014-03-21
linktitle: Hosting on GitHub
menu:
  main:
    parent: tutorials
next: /tutorials/mathjax
prev: /community/contributing
title: Hosting on GitHub Pages
weight: 10
---

## Intro

Many Hugo users have expressed interest in seeing a tutorial for how to set up a blog that generated by Hugo and hosted on GitHub Pages. This tutorial will do just that. We only require that the reader has Hugo installed correctly and is comfortable with git and GitHub.

During this tutorial, I will walk you through the main steps I took to create an example blog available at [http://spencerlyon2.github.io/hugo_gh_blog](http://spencerlyon2.github.io/hugo_gh_blog). The source code for this blog is on [GitHub](https://github.com/spencerlyon2/hugo_gh_blog). Readers are encouraged to download the example repository and follow along.

### Find a Home for Your Files

As our goal is to host a website using GitHub Pages, it is natural for us to host the content of the page in a GitHub repository. Thus, the first step is to either create a new repository on GitHub or create a new directory within an existing repository where the content of the website will live. To do this I created the repository [spencerlyon2/hugo_gh_blog](https://github.com/spencerlyon2/hugo_gh_blog).

## Create the Blog

### Write a `config.yaml` File

The very first step in creating a new Hugo site is to [write the config file](/overview/configuration). This config file is important for at least two reasons: (1) this is where site-wide settings (like the websites `baseurl`) go and (2) the config file dictates to some extent how Hugo will generate the website. For the example website I created a file `config.yaml` with the following contents

    ---
    contentdir: "content"
    layoutdir: "layouts"
    publishdir: "public"
    indexes:
      category: "categories"
    baseurl: "http://spencerlyon2.github.io/hugo_gh_blog"
    title: "Hugo Blog Template for GitHub Pages"
    ...

### Define Structure of Website

Hugo assumes that you organize the content of your site in a meaningful way and uses the same structure to render the website. Notice that we have the line `contentdir: "content"` in our configuration file. This means that all the actual content of the website should be placed somewhere within a folder named `content`. Hugo treats all directories in `content` as sections. For our example we only need one section: a place to hold our blog posts. So we created two new folders:

```
▾ <root>/
    ▾ content/
        ▾ posts/
```

### Create html Templates

The next step is to define the look and feel of your new website. Because Hugo will generate the site using html templates written by the user (you), this step is very subjective. I will merely present one possible theme that could be used to generate a blog. I decided to base the example project on a Jekyll theme called [lanyon](http://lanyon.getpoole.com). The lanyon theme is pure css and a slightly modified version of the css is in the `/static/css` directory of the example repository. If you are following along, you should grab the `static` folder from the example repository and put it alongside the `content` folder you just created.

Because there are so many files needed to fully compose a complete website, I will not be able to go trough each of them here. I will, however, show what the directory structure should look like when all is said and done:

```
▾ <root>/
    ▾ content/
        ▾ posts/
            <blog posts>.md
    ▾ static/
        ▾ css/
            lanyon.css
            poole.css
    ▾ layouts/
        ▾ partials/
            <templates to be used in other files>.html
        ▾ posts/
            li.html
            single.html
            summary.html
        ▾ indexes/
            category.html
            indexes.html
            posts.html
        index.html
    README.md
```

Each of the files in the example repository is well commented with a description of what the file as a whole does as well as an explanation of all major components in the file. If you are new to web development and/or Hugo I encourage you to search through these files to get a feel for how Hugo templates work and how the site is stitched together.

### Add Some Content

The final step in creating the blog is to add some actual blog posts. To do this simply create one markdown file (with extension .md) for each new blog post. At the top of each file you should include a metadata section that tells Hugo some things about the post (see [docs](/content/front-matter)). For example, consider the yaml metadata section from the top of the file `/content/posts/newest.md` from the example repository

    ---
    title: "Just another sample post"
    date: "2014-03-29"
    description: "This should be a more useful description"
    categories: 
        - "hugo"
        - "fun"
        - "test"
    ---

The keys set in this section are the mandatory `title` and `date` as well as the optional `description` and `categories`. Each of these items is used throughout the templates found in the `/layouts` directory and gives Hugo information about the post from other pages in the website.

## Configure `git` Workflow

Once the site is set up and working properly, we need to push it to the correct branch of a GitHub repository so the website can be served through GitHub Pages. There are many ways to do this. Here I will show the workflow I currently use to manage my websites that are hosted through GitHub Pages.

GitHub Pages will serve up a website for any repository that has a branch called `gh-pages` with a valid `index.html` file at that branch's root. A typical workflow might be to keep the content of a website on the `master` branch of a repository and the generated website on the `gh-pages` branch. This provides nice separation between input and output, but can be very tedious to work with. As a workaround we will use the `git subtree` family of commands to have the `public` directory (or whatever `publishdir` is set to in your `config.yaml`) mirror the root of the `gh-pages` branch of the repository. This will allow us to do all our work on the `master` branch, run Hugo have have the site output into the `public` directory, and then push that directory directly to the correct place for GitHub Pages to serve our site.

To get this properly set up we will execute a series of commands at the terminal. I will include all of them in one place here for easy copy and paste, and will explain what each line does via comments. Note that this is to be run from the `<root>` directory (wherever the `content` and `layout` folders of your Hugo project live). Also note that you will need to change the commands that have the example repository GitHub address so that they point to your repo.

    # Create a new orphand branch (no commit history) named gh-pages
    git checkout --orphan gh-pages

    # Unstage all files
    git rm --cached $(git ls-files)

    # Grab one file from the master branch so we can make a commit
    git checkout master README.md

    # Add and commit that file
    git add .
    git commit -m "INIT: initial commit on gh-pages branch"

    # Push to remote gh-pages branch 
    git push origin gh-pages

    # Return to master branch
    git checkout master

    # Remove the public folder to make room for the gh-pages subtree
    rm -rf public

    # Add the gh-pages branch of the repository. It will look like a folder named public
    git subtree add --prefix public git@github.com:spencerlyon2/hugo_gh_blog.git gh-pages --squash

    # Pull down the file we just committed. This helps avoid merge conflicts
    git subtree pull --prefix=public

    # Run hugo. Generated site will be placed in public directory (or omit -t ThemeName if you're not using a theme)
    hugo -t ThemeName


    # Add everything
    git add -A

    # Commit and push to master
    git commit -m "Updating site" && git push origin master

    # Push the public subtree to the gh-pages branch
    git subtree push --prefix=public git@github.com:spencerlyon2/hugo_gh_blog.git gh-pages

After executing these commands and waiting for the GitHub servers to update, the website we just created was live at [http://spencerlyon2.github.io/hugo_gh_blog](http://spencerlyon2.github.io/hugo_gh_blog). 

### `deploy.sh`

Now, as you add new posts to your blog, you will follow steps that look something like the following:
 
* Create the markdown source for the new post within the `content/posts` directory
* Preview your work by running Hugo in server mode with `hugo server --watch`
* Run Hugo not in server mode so that the generated urls will be correct for the website
* Add and commit the new post in `master` branch
* Push the `master` branch
* Push the public subtree to the remote `gh-pages` branch

The first two items in the previous list are simply a way to conveniently preview your content as you write. This is a dynamic and fairly streamlined process. All the remaining items, however, are the same every time you want to add new content to the website. To make this repetitive process easier, I have adapted a script from the source repository for the [Chimer Arta & Maker Space](https://github.com/chimera/chimeraarts.org) website that is highlighted in the [Hugo Showcase](/showcase). The script lives in a file called `deploy.sh` and has the following contents:

**Note:**

The first command `hugo` assumes you are running with all the default settings. 

To use a theme, make sure to specify it with `-t ThemeName` instead (or include the theme in the config file).

    hugo -t ThemeName

To build all draft posts *(If you only have drafts, no site will be generated)*

    hugo --buildDrafts

**Deploy.sh:**

    #!/bin/bash

    echo -e "\033[0;32mDeploying updates to GitHub...\033[0m"

    # Build the project. 
    hugo
    
    # Add changes to git.
    git add -A

    # Commit changes.
    msg="rebuilding site `date`"
    if [ $# -eq 1 ]
      then msg="$1"
    fi
    git commit -m "$msg"

    # Push source and build repos.
    git push origin master
    git subtree push --prefix=public git@github.com:spencerlyon2/hugo_gh_blog.git gh-pages

Now I can replace the last four items from our workflow list with a single command `bash deploy.sh`. This script accepts as an optional argument the commit message that git should use when committing your changes. If you wish to include a custom commit message, do so by putting it quotes after calling bash on the script: `bash deploy.sh "<my commit msg>"`. If you choose not to specify the commit message, one will be generated for you using the current time.

## Conclusion

Hopefully this tutorial helped you get your website off its feet and out into the open! If you have any further questions feel free to contact the community through the [mailing lists](/community/mailing-list).