Reduce time required to prepare for ROBOT query/verify

[balhoff]

This makes two changes:

1. In creating the Jena Dataset used to execute SPARQL queries, instead of TDBFactory.createDataset(), DatasetFactory.createGeneral() is used. This eliminates a large amount of time spent initializing the Dataset (240 seconds in the case of the GO build on my laptop).
2. I implemented a direct triple conversion from Sesame objects to Jena objects, avoiding creating a huge string representing the ontology. In the end this amazingly provides no speedup that I've been able to measure. But I expect it will reduce peak memory usage, which is a problem for us on Travis.

Behavior changes worth noting:

• I had to make a change to one test. In a SPARQL query, when specifying a named graph representing an ontology in the import chain, you need to use FROM instead of FROM NAMED. This is how I would have expected it to work in the first place, and I prefer the new behavior (in my opinion FROM NAMED is super confusing). I'm not sure if this impacts anyone. Neither was documented before.

• Method signatures with regard to checked exceptions have changed. Will this be a problem? If so I can catch and wrap those exceptions. (UPDATE—I saw that Travis failed; I wrapped the exceptions and now it passes.)

• docs/ have been added/updated

• tests have been added/updated

• mvn verify says all tests pass

• mvn site says all JavaDocs correct

• CHANGELOG.md has been updated

[cmungall]

This is great! Issue 666 ftw 😈😈😈

I assume the tests were with in memory Jena - does this affect things when the tdb options is set?

[balhoff]

@cmungall I haven't tested the disk-based TDB. It looks like a separate code path to me. @beckyjackson am I right?

[jamesaoverton]

Thanks @balhoff! I want to test this on some of my ontology projects. I'll try to do that soon.

[jamesaoverton]

I haven't yet had time for this -- I sincerely apologize.

@ontodev/robot-team Everyone is welcome to test this using this JAR: https://build.obolibrary.io/job/ontodev/job/robot/job/direct-rdf/lastSuccessfulBuild/artifact/bin/robot.jar

[beckyjackson]

Sorry for the delay in response. Yes, the --tdb option goes a separate way, never loading the ontology into memory.

I tested this on DOID and it works great - a query that previously took 20 seconds only took 9 seconds 😄

I also tested this with MRO and a query that previously took 29 seconds took 14 seconds, and the results were essentially the same. The only (tiny) difference was that a GROUP_CONCAT on synonyms returned the same values in a different order for two results (out of 941 results), but I don't think this is a problem?

[balhoff]

The only (tiny) difference was that a GROUP_CONCAT on synonyms returned the same values in a different order for two results (out of 941 results), but I don't think this is a problem?

That seems understandable for different backends. As far as I know you can't guarantee an ordering within GROUP_CONCAT.

[jamesaoverton]

Thanks @balhoff! Sorry again about the delay -- we're even busier than usual lately.

[balhoff]

No problem, thanks!!