Vanity Imports with Hugo

When working on Gorram, I decided I wanted to release it via a vanity import path. After all, that’s half the reason I got in the first place (an idea blatantly stolen from Russ Cox’s

What is a vanity import path? It is explained in the go get documentation. If you’re not hosted on one of the well known hosting sites (github, bitbucket, etc), go get has to figure out how to get your code. How it does this is fairly ingenious - it performs an http GET of the import path (first https then http) and looks for specific meta elements in the page’s header. The header elements tells go get what type of VCS is being used and what address to use to get the code.

The great thing about this is that it removes the dependency of your code on any one code hosting site. If you want to move your code from github to bitbucket, you can do that without breaking anyone.

So, the first thing you need to host your own vanity imports is something that will respond to those GET requests with the right response. You could do something complicated like a special web application running on a VM in the cloud, but that costs money and needs maintenance. Since I already had a Hugo website (running for free on github pages), I wanted to see if I could use that. It’s a slightly more manual process, but the barrier of entry is a lot lower and it works on any free static hosting (like github pages).

So what I want is to have go get, actually download the code from For that, I need to serve up this meta element:

<meta name="go-import" content=" git">

or more generally:

<meta name="go-import" content="import-prefix vcs repo-root">

Where import-prefix is a string that matches a prefix of the import statement used in your code, vcs is the type of source control used, and repo-root is the root of the VCS repo where your code lives.

What’s important to note here is that these should be set this way for packages in subdirectories as well. So, for, the meta tag should still be as above, since it matches a prefix of the import path, and the root of the repo is still (We’ll get to how to handle subdirectories later.)

You need a page serving that meta tag to live at the exact same place as the import statement… that generally will mean it needs to be in the root of your domain (I know that I, personally don’t want to see go get when I could have go get

The easiest way to do this and keep your code organized is to put all your pages for code into a new directory under content called “code”. Then you just need to set the “permalink” for the code type in your site’s config file thusly:

	code = "/:filename/"

Then your content’s filename (minus extension) will be used as its url relative to your site’s base URL. Following the same example as above, I have content/code/ which will make that page now appear at

Now, for the content. I don’t actually want to have to populate this page with content… I’d rather people just get forwarded on to github, so that’s what we’ll do, by using a refresh header. So here’s our template, that’ll live under layouts/code/single.html:

<!DOCTYPE html>
  <meta http-equiv="content-type" content="text/html; charset=utf-8">
  <meta name="go-import" content="{{substr .RelPermalink 0 -1}} git {{.Params.vanity}}">
  <meta name="go-source" content="{{substr .RelPermalink 0 -1}} {{.Params.vanity}} {{.Params.vanity}}/tree/master{/dir} {{.Params.vanity}}/blob/master{/dir}/{file}#L{line}">
  <meta http-equiv="refresh" content="0; url={{.Params.vanity}}">

This will generate a page that will auto-forward anyone who hits it on to your github account. Now, there’s one more (optional but recommended) piece - the go-source meta header. This is only relevant to, and tells godoc how to link to the sourcecode for your package (so links on will go straight to github and not back to your vanity url, see more details here).

Now all you need is to put a value of vanity = in the frontmatter of the correct page, and the template does the rest. If your repo has multiple directories, you’ll need a page for each directory (such as This would be kind of a drag, making the whole directory struture with content docs in each, except there’s a trick you can do here to make that easier.

I recently landed a change in Hugo that lets you customize the rendering of alias pages. Alias pages are pages that are mainly used to redirect people from an old URL to the new URL of the same content. But in our case, they can serve up the go-import and go-source meta headers for subdirectories of the main code document. To do this, make an alias.html template in the root of your layouts directory, and make it look like this:

<!DOCTYPE html><html>
        {{if .Page.Params.vanity -}}
        <meta name="go-import" content="{{substr .Page.RelPermalink 0 -1}} git {{.Page.Params.vanity}}">
        <meta name="go-source" content="{{substr .Page.RelPermalink 0 -1}} {{.Page.Params.vanity}} {{.Page.Params.vanity}}/tree/master{/dir} {{.Page.Params.vanity}}/blob/master{/dir}/{file}#L{line}">
        {{- end}}
        <title>{{ .Permalink }}</title>
        <link rel="canonical" href="{{ .Permalink }}"/>
        <meta http-equiv="content-type" content="text/html; charset=utf-8" />
        <meta http-equiv="refresh" content="0; url={{ .Permalink }}" />

Other than the stuff in the if statement, the rest is the default alias page that Hugo creates anyway. The stuff in the if statement is basically the same as what’s in the code template, just with an extra indirection of specifying .Page first.

Note that this change to Hugo is in master but not in a release yet. It’ll be in 0.18, but for now you’ll have to build master to get it.

Now, to produce pages for subpackages, you can just specify aliases in the front matter of the original document with the alias being the import path under the domain name:

aliases = [ "gorram/run", "gorram/cli" ]

So your entire content only needs to look like this:

date = 2016-10-02T23:00:00Z
title = "Gorram"
vanity = ""
aliases = [

Any time you add a new subdirectory to the package, you’ll need to add a new alias, and regenerate the site. This is unfortunately manual, but at least it’s a trivial amount of work.

That’s it. Now go get (and will know how to get your code.