You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Cleanly separate core ops infrastructure from ops discovery #59
For ops discovery, we want to use an industry standard mechanism (or mechanisms). Maybe Spring (#55), maybe plain ServiceLoader via JPMS module-info.java, maybe even something else. The Ops design should not conflate this with the engine itself, which has only the OpEnvironment interface, and accessor(s) for the OpInfos it has in its bag. Similarly, OpService as built on the SciJava service framework is a separate thing, which should not be part of the core ops engine at all.
To ensure this, the scijava-ops-api and scijava-ops-engine modules (or whatever they end up being named) must not depend on scijava-indexer, scijava-plugin, scijava-context, and similar layers relating to discovery, dependency injection, or application containers. It must be possible to create a "vanilla" OpEnvironment with a explicitly specified bag of ops.
While we have not yet adopted Spring, Classgraph, or any other industry-standard discovery mechanism, any/all of them should be easily integrated through the Discoverer abstraction.
Activity
gselzer commentedon May 13, 2021
Time commitment: 1 week
(NB it could take substantially longer depending on which/how many modules we want to partition, but I think this is unlikely)
gselzer commentedon Jan 21, 2022
This has been completed. I can't determine which commit did it, but it was definitely done by the time we merged scijava/incubator#41
gselzer commentedon Jan 21, 2022
While we have not yet adopted Spring, Classgraph, or any other industry-standard discovery mechanism, any/all of them should be easily integrated through the
Discovererabstraction.Merge pull request #59 from scijava/scijava/scijava-ops-engine/restor…