Docgen generates static web pages from a DocBook 5 “book” XML. But, it
only implements the small subset of DocBook elements that we actually use, and
it has no backward compatibility guarantees.
For editing DocBook, we are using XXE,
with the xxe-addon installed, which you can find in this project.
Building
These tools must be installed:
JDK 8 (or later), was tested with Oracle 17.0.9
Apache Maven, was tested with 3.8.1
Node.js was tested with 22.15.0.
(Node.js is only used to generate static content while building Docgen itself.)
To build, ensure that npm and bpx (from Node.js) are in the path, then in the top project directory
(freemarker-docgen) issue this:
mvn install
Node.js troubleshooting
Possible node.js related problems and solutions:
“Error: ENOENT, stat “: create that directory manually,
then retry.
If the system doesn’t find npm: Open a new terminal (command window) so
that it picks up the PATH environment variable changes. Adjust it if
necessary.
If the build has once worked, but now keeps failing due to some missing
modules, or anything strange, delete the “node_modules” directory, and
issue npm install to recreate it.
Running Docgen from your IDE
If you develop/debug Docgen, it’s convenient to launch it from your IDE.
As an example, let’s generate the FreeMarker Manual. For that, create a
Run Configuration in you IDE, with main class
org.freemarker.docgen.cli.Main, and these command line arguments
(replace <FREEMARKER_PROJECT_DIR> with the actual directory):
To ease comparing the outputs of different runs, you can set a fixed value
for the last modification with a java argument like this:
-Ddocgen.generationTime=2020-07-15T17:00Z
Compiling LESS and JS
This happens automatically during build, in the generate-resources Maven phase.
The generated output is in target\resources-gulp, which will be included in
the core jar artifact.
Pagefind (search) support
Can be enabled in the docgen.cjson (a file that’s normally next to the DocBook XML) via pagefindBasedSearch: true.
Pagefind indexing is done when generating the docgen output from the DocBook XML. The output is fully static HTML, with
purely cline-side JavaScript that generates the search results. However, due to browser security restrictions, the
search functionality will only work if you visit via HTTP(S), and not via a file: URL. For local testing you can use
npx http-server for example.
When pagefindBasedSearch is true, Node.js has to be available where Docgen generates its output (the HTML-s), and
not just when building Docgen itself! That’s because Pagefind indexing depends on Node.js.
Releasing a new Docgen version
[TODO] Standard ASF release procedure (staging, voting, etc.), so that we can release
to the Maven Central. Not advertised, no announcements, no backward compatibility
promises, but makes building our dependent projects easier.
Icon Font Attribution
The icon font in this project was built using IcoMoon
and contains selected icons from:
Docgen
About
Apache FreeMarker Docgen is an internal project under the Apache FreeMarker TLP, used for generating the FreeMarker Manual, the FreeMarker homepage, and maybe some more.
Docgen generates static web pages from a DocBook 5 “book” XML. But, it only implements the small subset of DocBook elements that we actually use, and it has no backward compatibility guarantees.
Usage
For some examples see:
freemarker-siteprojectlegacy-testsin this projectFor editing DocBook, we are using XXE, with the
xxe-addoninstalled, which you can find in this project.Building
These tools must be installed:
To build, ensure that
npmandbpx(from Node.js) are in the path, then in the top project directory (freemarker-docgen) issue this:mvn installNode.js troubleshooting
Possible node.js related problems and solutions:
npm: Open a new terminal (command window) so that it picks up thePATHenvironment variable changes. Adjust it if necessary.npm installto recreate it.Running Docgen from your IDE
If you develop/debug Docgen, it’s convenient to launch it from your IDE. As an example, let’s generate the FreeMarker Manual. For that, create a Run Configuration in you IDE, with main class
org.freemarker.docgen.cli.Main, and these command line arguments (replace<FREEMARKER_PROJECT_DIR>with the actual directory):To ease comparing the outputs of different runs, you can set a fixed value for the last modification with a java argument like this:
Compiling LESS and JS
This happens automatically during build, in the
generate-resourcesMaven phase. The generated output is intarget\resources-gulp, which will be included in the core jar artifact.Pagefind (search) support
Can be enabled in the
docgen.cjson(a file that’s normally next to the DocBook XML) viapagefindBasedSearch: true.Pagefind indexing is done when generating the docgen output from the DocBook XML. The output is fully static HTML, with purely cline-side JavaScript that generates the search results. However, due to browser security restrictions, the search functionality will only work if you visit via HTTP(S), and not via a
file:URL. For local testing you can usenpx http-serverfor example.When
pagefindBasedSearchistrue, Node.js has to be available where Docgen generates its output (the HTML-s), and not just when building Docgen itself! That’s because Pagefind indexing depends on Node.js.Releasing a new Docgen version
[TODO] Standard ASF release procedure (staging, voting, etc.), so that we can release to the Maven Central. Not advertised, no announcements, no backward compatibility promises, but makes building our dependent projects easier.
Icon Font Attribution
The icon font in this project was built using IcoMoon and contains selected icons from: