GH-5112 various bugs related to the ShaclSail

[hmottestad]

GitHub issue resolved: #5112 #5113 #5114 #5115

Briefly describe the changes proposed in this PR:

---

PR Author Checklist (see the contributor guidelines for more details):

• my pull request is self-contained
• I've added tests for the changes I made
• I've applied code formatting (you can use mvn process-resources to format from the command line)
• I've squashed my commits where necessary
• every commit message starts with the issue number (GH-xxxx) followed by a meaningful description of the change

[hmottestad]

Before

Benchmark                                  Mode  Cnt   Score   Error  Units
SparqlConstraintsBenchmarkEmpty.shacl      avgt    5  43.552 ± 3.038  ms/op
SparqlConstraintsBenchmarkEmpty.shaclBulk  avgt    5  22.324 ± 2.957  ms/op

Benchmark                                                          Mode  Cnt     Score     Error  Units
ComplexLargeBenchmark.disabledValidationSail                       avgt    5   358.824 ±  38.191  ms/op
ComplexLargeBenchmark.disabledValidationTransaction                avgt    5   356.420 ±  27.879  ms/op
ComplexLargeBenchmark.noPreloading                                 avgt    5   678.464 ±  98.324  ms/op
ComplexLargeBenchmark.noPreloadingBulk                             avgt    5   652.532 ±  14.735  ms/op
ComplexLargeBenchmark.noPreloadingBulkParallelCached               avgt    5   558.096 ±  54.905  ms/op
ComplexLargeBenchmark.noPreloadingNonEmpty                         avgt    5  2984.330 ±  85.934  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallel                 avgt    5  2129.350 ±  35.701  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelLmdb             avgt    5  8165.658 ± 271.752  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelNativeStore      avgt    5  6998.737 ± 377.024  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelReadCommitted    avgt    5  3838.380 ± 139.557  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelRemoved          avgt    5  2273.318 ± 202.531  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyRemoved                  avgt    5  3369.851 ± 521.790  ms/op
ComplexLargeBenchmark.noPreloadingParallel                         avgt    5   722.336 ±  65.782  ms/op
ComplexLargeBenchmark.noPreloadingParallelNoCache                  avgt    5   655.416 ± 130.262  ms/op
ComplexLargeBenchmark.noPreloadingRevalidate                       avgt    5   562.064 ±  17.714  ms/op
ComplexLargeBenchmark.noPreloadingRevalidateLowMem                 avgt    5   627.493 ±  17.177  ms/op
ComplexLargeBenchmark.noPreloadingRevalidateNativeStore            avgt    5  3335.937 ± 101.412  ms/op
ComplexLargeBenchmark.noPreloadingTransactionalValidationLimit     avgt    5   998.597 ±  91.490  ms/op
ComplexLargeBenchmark.shaclNothingToValidateTransactionsPreloaded  avgt    5     0.102 ±   0.004  ms/op

After

Benchmark                                  Mode  Cnt   Score   Error  Units
SparqlConstraintsBenchmarkEmpty.shacl      avgt    5  42.473 ± 4.833  ms/op
SparqlConstraintsBenchmarkEmpty.shaclBulk  avgt    5  20.516 ± 2.267  ms/op

Benchmark                                                          Mode  Cnt     Score      Error  Units
ComplexLargeBenchmark.disabledValidationSail                       avgt    5   332.216 ±   54.691  ms/op
ComplexLargeBenchmark.disabledValidationTransaction                avgt    5   329.656 ±   13.421  ms/op
ComplexLargeBenchmark.noPreloading                                 avgt    5   652.694 ±   61.175  ms/op
ComplexLargeBenchmark.noPreloadingBulk                             avgt    5   631.931 ±   28.237  ms/op
ComplexLargeBenchmark.noPreloadingBulkParallelCached               avgt    5   583.691 ±   68.283  ms/op
ComplexLargeBenchmark.noPreloadingNonEmpty                         avgt    5  3751.262 ±  182.784  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallel                 avgt    5  2469.678 ±  302.404  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelLmdb             avgt    5  9497.425 ± 1303.038  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelNativeStore      avgt    5  7256.163 ±  365.693  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelReadCommitted    avgt    5  4295.821 ±  431.832  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyParallelRemoved          avgt    5  2755.962 ±  538.590  ms/op
ComplexLargeBenchmark.noPreloadingNonEmptyRemoved                  avgt    5  3915.580 ±  193.414  ms/op
ComplexLargeBenchmark.noPreloadingParallel                         avgt    5   677.575 ±  129.554  ms/op
ComplexLargeBenchmark.noPreloadingParallelNoCache                  avgt    5   644.252 ±   45.313  ms/op
ComplexLargeBenchmark.noPreloadingRevalidate                       avgt    5   568.130 ±   29.726  ms/op
ComplexLargeBenchmark.noPreloadingRevalidateLowMem                 avgt    5   642.213 ±   35.515  ms/op
ComplexLargeBenchmark.noPreloadingRevalidateNativeStore            avgt    5  3466.532 ±  259.464  ms/op
ComplexLargeBenchmark.noPreloadingTransactionalValidationLimit     avgt    5   970.873 ±   66.323  ms/op
ComplexLargeBenchmark.shaclNothingToValidateTransactionsPreloaded  avgt    5     0.109 ±    0.004  ms/op

[hmottestad]

Performance degradation seems to be related to TargetClass and the removal of a call to Unique there.

It might not be the Unique, but rather that we were caching the unique and sorted data, while we now just cache the raw unsorted data.

[hmottestad]

Maybe we should try to optimize the bulk leftOuter/inner joins to not require the data be sorted. The left side can easily just be a LinkedHashMap, that way we preserve any order that there was in the data without requiring the data to be ordered.