What’s New in Antora 2.0
Learn about what’s new in the 2.0 release series of Antora.
Antora 2.0.1
Fixed issues
- Issue 315
-
The URL of the most recent non-prerelease version of page is now used as the canonical URL.
- Issue 433
-
The path (i.e., relative) of an xref target can now begin with
@
. - Issue 434
-
If an xref target isn’t verified as publishable, it’s now left unresolved.
- Issue 455
-
If a repository fails to clone, the empty repository is now automatically deleted from Antora’s cache directory.
- Issue 456
-
Instead of stopping, Antora now issues a warning if it can’t resolve the start page of a site or component version.
- Issue 457
-
Resource ID path segments are now cleaned and page aliases are prevented from referencing themselves through the use of self references, parent references, and/or repeat separators.
- Issue 458
-
Spaces in URLs of rewrite rules are now properly escaped; applies to the nginx and Netlify redirect facilities.
- Issue 459
-
Component names can no longer contain
/
,.
, or..
because these characters are interpreted as path segments in the generated site. - Issue 462
-
The correct line number is now reported when an include target or tag cannot be resolved.
- Issue 463
-
A reference to a page with the same root-relative path as a navigation file is now fixed.
Thank you!
Most important of all, a huge thank you! to all the folks who helped make Antora even better.
- Amit Mendapara (@cristatus)
-
Update credential store setup command for zsh MR 319
- Jean-Baptiste (@jibecfed)
-
Correct a typo and make the runtime configuration easier to understand MR 321
- James Elliott (@Deep-Symmetry)
-
Remove _attributes.adoc from documentation MR 322
- François Violette (@fviolette)
-
Fix typos in documentation MR 329
Fix inline ID syntax example MR 340 - Daniel Mulholland (@danyill)
-
Re-order version information and mention version is used in page URL MR 334
Antora 2.0.0
Release date: 2018.12.25 | Issue label: 2.0.0
The Antora 2.0.0 release streamlines the installation process, improves platform and library compatibility, provides a simpler and pluggable authentication mechanism for private repositories, and delivers the latest Asciidoctor capabilities.
If you need a quick overview of the breaking changes, jump to the upgrade checklist. |
Hassle-free installation thanks to isomorphic-git
When Antora launched, nodegit was the only git client for JavaScript that provided the features Antora required for interfacing with git repositories (clone, fetch, resolve refs, walk trees, read blobs). Despite being a comprehensive git client, nodegit has proven to be tremendously difficult for users to install. This difficulty stems from its coupling to native system libraries for compilation, bundling of native libraries, limited portability across platforms and Node versions, misbehaving post-install scripts, reliance on an SSH agent for authentication, and lack of interest in addressing portability concerns.
Since then, a full-fledged alternative has emerged. Our friend William Hilton stepped up to create isomorphic-git, a pure JavaScript git client (fun fact, it also runs in a browser). In Antora 2.0, we’ve completely replaced nodegit with isomorphic-git.
Switching Antora to isomorphic-git not only guarantees a hassle-free installation, but drastically reduces installation time and size as well. This improvement is a result of Antora’s dependencies no longer relying on native system libraries. And it means Antora now works on all platforms and supported Node versions.
isomorphic-git also allowed us to expand our test coverage and therefore improve Antora’s stability. isomorphic-git drives all of the git tests in the test suite. And those tests now use proper remote clone and fetch operations to test remote repositories.
All-in-all, the switch from nodegit to isomorphic-git has been very positive, a huge step forward for Antora, and the main reason for the major version jump to Antora 2.0.
Goodbye SSH. Hello git credentials.
Replacing nodegit with isomorphic-git brought with it a breaking change in how Antora handles private repository authentication. Private repository authentication is now performed using HTTP Basic authentication over HTTPS instead of public/private key authentication over SSH using an SSH agent.
This change makes it easier to manage authentication because Antora no longer requires running an SSH agent. Instead, authentication credentials can be managed using the git credential store mechanism or a custom credential manager provided by the user.
If you use private repositories, you’ll need to add authentication credentials for each git host (or individual repository) to the default or a custom git credentials store before running Antora 2.0. To learn what types of credentials Antora can use for authentication and how to add your credentials to the default credentials store, refer to the Private repository authentication documentation.
Here’s a quick summary of where those credentials can be stored:
-
$HOME/.git-credentials by default
-
--git-credentials-path
CLI option — override location of git credential store -
GIT_CREDENTIALS_PATH
environment variable — override location of git credential store -
git.credentials
key in Antora playbook — override path or contents of git credential store -
GIT_CREDENTIALS
environment variable — override contents of git credential store
The documentation has examples showing how to specify a custom git credential store using each of the methods listed above. If none of these options meet your requirements, the documentation also shows how to implement your own credential manager plugin.
Finally, Antora transparently converts git SSH URLs in your playbook to HTTPS URLs so you don’t have to modify your playbook file when upgrading to Antora 2.0. If this automatic translation doesn’t work, meaning a git repository is not available at the equivalent HTTPS URL, then you’ll need to update your playbook file to use the correct HTTPS URL.
All the latest Asciidoctor features
Antora 2.0 ships with Asciidoctor.js 1.5.9 thanks to the hard work by Guillaume Grossetie. This release of Asciidoctor.js brings you the latest Asciidoctor features, updates the Asciidoctor.js API, fixes conflicts with libraries, and offers several nice enhancements to the AsciiDoc syntax. Here are some of the highlights:
-
Nest ordered and unordered lists to an arbitrary depth.
-
Autonumber callout markers using the
<.>
syntax. -
Create an xref with automatic text to a list item or table cell.
-
Attribute references in the attribute list of an include directive are replaced.
-
Section titles are scrubbed more thoroughly when auto-generating an ID.
-
The position of section anchors can be configured (before or after).
-
The xref macro now supports attributes (just as the link macro does).
-
The
docname
attribute is now set to the module-relative path of the page (minus the file extension). -
Accurate warnings if an include tag is mismatched, unclosed, or not found (additionally, Asciidoctor now emits a warning if Antora cannot locate the include target).
-
All messages are now routed through a logger (which will hook into Antora’s logger, once implemented).
-
An upgrade to Opal means Asciidoctor.js no longer goes down a rabbit hole when iterating over an object’s properties.
-
Much more of the Asciidoctor API has been mapped in Asciidoctor.js, making it far easier to write extensions.
-
The extension API provides a simple mechanism for generating an extension class (as an alternative to an extension block).
-
The AbstractNode#findBy method can be short-circuited to halt the tree walk.
See the release notes for Asciidoctor and Asciidoctor.js to find complete details.
More improvements and changes
- Next, previous, and parent pages
-
The
next
,previous
, andparent
properties are now available in the UI model so your UI can access adjacent pages in the navigation tree. See the template variables documentation to learn more. - Versioned cache folder
-
Antora uses a versioned cache folder for the cloned git content repositories. On first run, Antora will re-clone your content repositories into this folder. This was necessary to avoid conflicts with repositories that were cached using nodegit.
- Node compatibility
-
Antora now works with Node 10 (and 11).
- Antora playbook files in TOML
-
You can now write Antora playbook files in the TOML configuration language. The CSON format is no longer supported. If your Antora playbook is in CSON, you’ll need to migrate it to YAML, JSON, or TOML before using Antora 2.0.
- runtime.fetch
-
The
runtime.pull
playbook key has been replaced withruntime.fetch
and the--pull
CLI option has been replaced with--fetch
. This setting tells Antora to fetch updates to remote content repositories and grab a fresh UI bundle if marked as a snapshot. It was renamed to avoid confusion with thepull
command in git. But don’t worry, the old names still work. They’re just deprecated and will be removed in the next major version. - page.origin.private
-
If a content source requires authentication (either because credentials were defined in the URL or credentials were requested from the credentials manager), the
page.origin.private
property will be set in the UI model. - page.displayVersion
-
The display version of a component is now accessible via the UI model using the
page.displayVersion
property. The display version is defined by thedisplay_version
key in the component descriptor (i.e., antora.yml). - page.latest
-
The
page.versions.latest
property in the UI model can now be reached usingpage.latest
. - AsciiDoc doctype
-
The AsciiDoc
doctype
option can be safely set in the AsciiDoc configuration key. Navigation files will always be parsed using the article doctype. - default tags filter
-
The default tags filter for content repositories can now be set using the
content.tags
key in the playbook.
Known issues
Reference pruning
- Issue
- Summary
-
After the switch to isomorphic-git, references are no longer pruned when the
--pull
option is specified. This means that if a reference is removed from the remote repository, it will not be removed from Antora’s cached version. - Workaround
-
You can clear Antora’s cache directory or the cache of the repository in question. We’ll restore this behavior in a future enhancement.
Broken 404 pages for sites that aren’t served from root folder
- Issue
- Summary
-
If your site is not served from the root folder of a domain (such as is the case of sites running on the GitHub Pages or GitLab Pages domain), the references to UI assets and navigation links on the 404 page will not work property.
- Workaround
-
Postprocess the 404.html page to fix the paths before publishing the site.
Self-signed certificates
- Issue
- Summary
-
The HTTP client Antora uses to download the UI bundle (got) does not load the CA certificate file on the local machine. As a result, if the server hosting the UI bundle has a self-signed certificate, Antora will fail trying to download the bundle (due to the fact that Node rejects the certificate as unauthorized).
- Workaround
-
Instruct Node not to reject unauthorized certificates by setting the environment variable NODE_TLS_REJECT_UNAUTHORIZED=0 when calling Antora.
Thank you!
Most important of all, a huge thank you! to all the folks who helped make Antora even better.
- Abel Salgado Romero (@abelsromero)
-
Fix wrong gulp-ci package name replacing it with correct gulp-cli MR 294
- Antonio (@bandantonio)
-
Fix colon typo in resource id include directive MR 284
Add information about missing runtime category in the playbook configuration MR 282 - Julien Grenier (@jgrenier)
-
Fix failing build (docker-antora) MR 5
- Matthew Setter (@settermjd)
-
Correct reference to ui-bundle-url in the docs MR 280
- @Leantricity
-
Fix page-to-page-xref.adoc typo MR 278
Metrics
The raw install size of Antora drops from 154MB down to 43MB. The Docker image on DockerHub dropped from 60 MB to 29 MB. It took 11 seconds to install Antora 1.1.1 from cache. It only takes 3 seconds to install Antora 2.0.0 from cache.
Upgrade checklist
You may need to make some of the following changes to your documentation system before using Antora 2.0.
- Does Antora access private repositories in order to build your site?
-
If yes, you get to say goodbye to your SSH agent and set up your authentication credentials for each private repository URL in a git credential store. Refer to the Private repository authentication documentation for setup options and instructions.
- Does Antora access git repositories using the git: protocol or static HTTP (such as a host using cgit)?
-
Unfortunately, Antora can no longer connect to repositories that use these connection methods. You’ll need to either clone the repositories to a local folder and point Antora at those locations or you’ll need to set up a proxy server that serves the repositories over smart HTTP (as do all major git hosts).
- Is your Antora playbook file written in CSON?
-
If yes, you’ll need to migrate it to YAML, JSON, or TOML before running Antora 2.0.
If you answered No to all of the above questions, then you’re ready to upgrade to Antora 2.0 now! See Upgrade Antora for instructions.
Software Lifecycle Notices
Antora 1.1 will enter maintenance on March 25, 2019. See the release schedule for details.