Performance issue with deep taxonomy benchmark

[josd]

I was testing https://raw.githubusercontent.com/josd/josd.github.io/master/temp/dt.n3
and the good news is that the 30000 triples are produced but at a rather slow pace

$ uname -srmo
Linux 4.19.104-microsoft-standard x86_64 GNU/Linux
$ npm run test

> n3@1.0.0 test
> jest

 PASS  test/parse.test.ts
 PASS  test/reasoner.test.ts (1350.582 s)

Test Suites: 2 passed, 2 total
Tests:       18 passed, 18 total
Snapshots:   0 total
Time:        1350.89 s
Ran all test suites.

FYI, see also mthom/scryer-prolog#1053
esp.

This benchmark was designed by the late Harold Boley http://www.cs.unb.ca/~boley/
and is interesting to evaluate the subsumption capabilities of reasoners (using RDF and OWL)
e.g.
[[
SNOMED CT currently contains more than 300,000 medical concepts, divided into
hierarchies as diverse as body structure, clinical findings, geographic location and
pharmaceutical/biological product
]] -- https://searchhealthit.techtarget.com/definition/SNOMED-CT

[phochste]

This JavaScript code is indeed shockingly slow compared to EYE. I tried to run this test also this morning and stopped it after 10 minutes .

[josd]

Indeed when I ran it initially it went in timeout

    thrown: "Exceeded timeout of 5000 ms for a test.
    Use jest.setTimeout(newTimeout) to increase the timeout value, if this is a long-running test."

although it ran for more than those 5 sec

Test Suites: 1 failed, 1 passed, 2 total
Tests:       1 failed, 17 passed, 18 total
Snapshots:   0 total
Time:        1245.282 s
Ran all test suites.

It was able to finish after adding

jest.setTimeout(1000000000);

A while ago after a meeting with @RubenVerborgh I worked on eye.js which is doing only backward chaining, always producing a proof and using json instead of N3. Running the deep taxonomy benchmark takes

$ time node dt.js > dt-proof.json

real    0m15.476s
user    0m16.253s
sys     0m1.150s

For this test the reasoner also has to chain (backwards) through the whole taxonomy.

[phochste]

With another way of parsing N3 files I could bring the deep taxonomy benchmark down to about 110 seconds.