[feature] Add ModuleFactory and IndexFactory SPI for auto-discovery of bundled modules and indexes

[duncdrum]

Summary

Introduces ServiceLoader-based SPI for both XQuery modules and index modules so that all bundled implementations are auto-discovered at startup without explicit conf.xml entries. The canonical conf.xml template is trimmed accordingly — new installations get a leaner config by default. Existing hand-maintained configs continue to work unchanged: explicit entries always take precedence over SPI discovery.

Builds on the @enabled attribute from #6550 (itself depending on #6528, #6530, #6531).

Closes #3062 — delivers both parts of the issue: the @enabled attribute (in #6550) and the ServiceLoader/CDI-style autodiscovery requested here.

What Changed

ModuleFactory SPI

• New org.exist.xquery.ModuleFactory interface: getNamespaceURI() + create(String, Map<String,List<?>>, XQueryContext) matching the existing Module factory contract
• ServiceLoader<ModuleFactory> wired into Configuration.configureModules(): SPI runs first, then the conf.xml loop; an explicit <module> entry for the same namespace URI overwrites the SPI entry, and enabled="no" suppresses it
• All 27 bundled XQuery modules register their ModuleFactory implementations in META-INF/services/org.exist.xquery.ModuleFactory
• Configuration.class.getClassLoader() used explicitly so the loader survives embedding scenarios

IndexFactory SPI

• New org.exist.indexing.IndexFactory interface: getId() + create(BrokerPool, Path, Element)
• ServiceLoader<IndexFactory> wired into IndexManager initialisation: SPI entries whose id does not appear as a live <module> entry in conf.xml are registered automatically; enabled="no" suppresses without requiring JAR removal
• All bundled index modules (Lucene, ngram, range, sort, spatial) wired; index name defaults to the config id when registered via SPI without a conf.xml element

Vector model registry

• Integrated with Configuration; @enabled on <vector-models> children so individual models can be disabled; suppressed WARN log for unavailable models (logged at DEBUG instead)

exist-distribution/src/main/config/conf.xml (canonical template)

• <builtin-modules> section trimmed — the 27 bundled modules no longer listed individually
• <modules> (index) section trimmed — the bundled index modules no longer listed individually
• Third-party / optional entries retained with comments explaining how to add custom modules

Compatibility

• Explicit <module> entries in conf.xml always take precedence over SPI discovery for the same namespace URI or index id — zero upgrade action required for hand-maintained configs
• To suppress a bundled module or index without removing its JAR, add enabled="no" to its entry
• New META-INF/services/ files are JAR-internal; no classpath changes required by consumers

Test Plan

• mvn validate -pl exist-distribution -Ddependency-check.skip=true — canonical conf.xml still validates
• mvn test -pl exist-core -Ddependency-check.skip=true -Ddocker=false — full unit suite passes
• Manual: start eXist-db with the trimmed conf.xml; all 27 built-in XQuery modules available in eXide
• Manual: start eXist-db; Lucene, ngram, and range indexes are active without explicit conf.xml entries
• Manual: add enabled="no" to a module's conf.xml entry; confirm it is not loaded even though SPI would discover it

🤖 Generated with Claude Code

[line-o]

Resolving modules to load by default over SPI received pushback on the community call on 2026-07-06
Security concerns need to be addressed before we can consider pulling this in.

@dizzzz specific concern was: "I see autoloading modules that are not mentioned in the configuration as a risk."

My concern is: if bundled modules are left out of the configuration I have to know they are there in order to disable them again.