pgjdbc/pgjdbc GitHub issues and pull requests (mirror)  

help / color / mirror / Atom feed

[pgjdbc/pgjdbc] PR #3641: feat: use PreparedStatement for DatabaseMetaData.getCrossReference, getImportedKeys, getExportedKeys
3+ messages / 2 participants
[nested] [flat]

* [pgjdbc/pgjdbc] PR #3641: feat: use PreparedStatement for DatabaseMetaData.getCrossReference, getImportedKeys, getExportedKeys
@ 2025-05-28 16:03 "vlsi (@vlsi)" <[email protected]>
  0 siblings, 0 replies; 3+ messages in thread

From: vlsi (@vlsi) @ 2025-05-28 16:03 UTC (permalink / raw)
  To: pgjdbc/pgjdbc <[email protected]>

See https://github.com/pgjdbc/pgjdbc/issues/3511

At best, we should avoid building the same query again and again, however, let's start with just using `PreparedStatements`.

Benchmarks with PG 16.3

Note: `ps.closeOnCompletion()` enables server-prepared statements even though the prepared statements are never closed explicitly (the APIs return `ResultSet`, and there's no way to close the statement other than `closeOnCompletion`).

## Before

```
Benchmark                                     Mode  Cnt   Score   Error  Units
DatabaseMetadataBenchmark.getColumns_all      avgt   15   1,156 ± 0,184  ms/op
DatabaseMetadataBenchmark.getColumns_named    avgt   15   1,085 ± 0,079  ms/op
DatabaseMetadataBenchmark.getCrossReferences  avgt   15  42,108 ± 0,536  ms/op
DatabaseMetadataBenchmark.getTables           avgt   15   0,423 ± 0,018  ms/op
```

## After: `String` -> `StringBuilder`

```
Benchmark                                     Mode  Cnt  Score   Error  Units
DatabaseMetadataBenchmark.getColumns_all      avgt   15  0,794 ± 0,041  ms/op
DatabaseMetadataBenchmark.getColumns_named    avgt   15  0,820 ± 0,056  ms/op
DatabaseMetadataBenchmark.getCrossReferences  avgt   15  4,128 ± 0,156  ms/op
DatabaseMetadataBenchmark.getTables           avgt   15  0,489 ± 0,041  ms/op
```

## After: `String` -> `StringBuilder`; `closeOnCompletion`

```
Benchmark                                     Mode  Cnt  Score   Error  Units
DatabaseMetadataBenchmark.getColumns_all      avgt   15  0,774 ± 0,020  ms/op
DatabaseMetadataBenchmark.getColumns_named    avgt   15  0,797 ± 0,034  ms/op
DatabaseMetadataBenchmark.getCrossReferences  avgt   15  0,142 ± 0,007  ms/op
DatabaseMetadataBenchmark.getTables           avgt   15  0,316 ± 0,050  ms/op
```

^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: [pgjdbc/pgjdbc] PR #3641: feat: use PreparedStatement for DatabaseMetaData.getCrossReference, getImportedKeys, getExportedKeys
@ 2025-05-28 16:34 ` "davecramer (@davecramer)" <[email protected]>
  1 sibling, 0 replies; 3+ messages in thread

From: davecramer (@davecramer) @ 2025-05-28 16:34 UTC (permalink / raw)
  To: pgjdbc/pgjdbc <[email protected]>

Fair, but I think we need to understand the root cause of the regression.

^ permalink  raw  reply  [nested|flat] 3+ messages in thread

* Re: [pgjdbc/pgjdbc] PR #3641: feat: use PreparedStatement for DatabaseMetaData.getCrossReference, getImportedKeys, getExportedKeys
@ 2025-05-30 16:54 ` "vlsi (@vlsi)" <[email protected]>
  1 sibling, 0 replies; 3+ messages in thread

From: vlsi (@vlsi) @ 2025-05-30 16:54 UTC (permalink / raw)
  To: pgjdbc/pgjdbc <[email protected]>

Any thoughts?
I think we could merge this and release yet another version to get the perf issue fixed.

The core changes here are like `String->StringBuilder`. It would be great to add `try-with-resources`, however I don't like to mix many changes into a single PR.

^ permalink  raw  reply  [nested|flat] 3+ messages in thread

end of thread, other threads:[~2025-05-30 16:54 UTC | newest]

Thread overview: 3+ messages (download: mbox mbox.gz follow: Atom feed)
-- links below jump to the message on this page --
2025-05-28 16:03 [pgjdbc/pgjdbc] PR #3641: feat: use PreparedStatement for DatabaseMetaData.getCrossReference, getImportedKeys, getExportedKeys "vlsi (@vlsi)" <[email protected]>
2025-05-28 16:34 ` "davecramer (@davecramer)" <[email protected]>
2025-05-30 16:54 ` "vlsi (@vlsi)" <[email protected]>

This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox