(https://github.com/apache/camel-website/issues/666)
Preview at https://pr-667--camel.netlify.app <https://pr-667--camel.netlify.app/> Using an observation and some initial code from Rob Winch I’ve developed an Antora extension that can scan for files not in the expected Antora structure and map them to Antora-appropriate locations. This can replace copying and symlinking files. Right now it can’t replace all the symlinking in main camel since that has a step of scanning .adocs for included example java code and then symlinking the used files. I think it’s practical to have the Antora extension do that too, but not yet. The PR for this issue applies this to camel-spring-boot and camel-kafka-connector. I’ve also enhanced the indexer used extensively for generating tables of components to act as an Antora extension and add a template instance page for each query result found. This lets us directly generate .adoc pages from .json files without needing a pre-existing .adoc page for each .json file. I think there are quite a few subprojects that have per-component documentation that is completely generated: to start with I’ve used camel-kafka-connector as an example. Further work in this PR set: Whatever is generating the json files for camel-kafka-connector is duplicating information. Properties with enumerated values have the values both as json elements and text in the description. For consistency across the site, I’m displaying the descriptions using uniform code that shows the json element values as a list, resulting in duplication, for instance https://pr-667--camel.netlify.app/camel-kafka-connector/next/reference/connectors/camel-activemq-kafka-sink-connector.html#_connector_option_camel_sink_path_destinationType. I’d like to remove the hardcoded text rendering from the description. I discovered a bug in how one of the asciidoctor extensions is adding header attributes and put in a quick workaround: this can be fixed properly with a new release of the extensions: the html is the same either way. If this approach seems like a good idea, after merging these I can work on other subprojects. There are some such as kamelets where this approach can be used simply and directly, and some, such as camel-quarkus, where I think it will work but may be more complicated. Take a look! David Jencks
