keksmd opened a new issue, #974: URL: https://github.com/apache/incubator-graphar/issues/974
### Describe the enhancement requested Add a dataset validator that checks a published GraphAr dataset against the format rules instead of failing later inside a reader. Given a graph URI it reports concrete, located issues: metadata that does not describe the files on storage, chunk counts that disagree with the declared chunk size, offset chunks that are not monotonic, and adjacency chunks that reference vertices outside the declared count. Component(s): Java ## Scope - Add `graphar-validator` with a `DatasetValidator` that walks metadata and storage and returns a `ValidationReport` of located `ValidationIssue` values. - Depend only on `graphar-info`, `graphar-core`, `graphar-io-api` and `graphar-storage-api`, so validation stays independent of any one format or backend. - Add tests over deliberately corrupted datasets: missing chunk file, wrong chunk count, non-monotonic offsets, out-of-range vertex id. ## Non-goals No repair or migration of a broken dataset, no reader facade, no new format backend, and no validation of user payload semantics beyond what the GraphAr format itself states. ## Compatibility and acceptance A canonical dataset from the `testing` fixtures must validate clean; each corrupted fixture must produce exactly the issue that describes it, with the offending path. Java 11 compatible. ## Dependency order Its tests build datasets through the write path, so this slice depends on `graphar-core` (#967, PR #968), a merged format backend (#964), and the write facade (#973). It should land last of the four. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
