carterkozak opened a new pull request #555:
URL: https://github.com/apache/logging-log4j2/pull/555
Previously the entire input string was written to the output
buffer, then small sections were deleted or replaced within that
string. This meant that for each section that was abbreviated,
nearly all following characters from subsequent sections would
be migrated, resulting in O(n^2) performance.
Now we read from the existing input String, and write sections
to the output buffer piece-by-piece. For small strings this may
result in more pointer-hopping (may cost ~tens of nanoseconds) while
larger values with many abbreviated sections are substantially
less expensive. I would expect large "enterprise-grade" class
names to perform better than previously.
Note that the new approach may result in multiple StringBuilder
resize operations in the worst case, where previously writing
the original string to the output would have caused it to grow
all at once. In most cases we attempt to initialize builders
with a reasonable size, so I opted not to add an `ensureCapacity`
given the estimated size may be substantially larger than we need.
Benchmarks with this change:
```
NamePatternConverterBenchmark.benchNamePatternConverter
org.bogus.hokus.pokus.org.bogus.hokus.pokus.org.bogus.hokus.pokus.RetroEncabulatorFactorySingleton
avgt 3 202.992 ± 1.584 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
org.bogus.hokus.pokus.Clazz1
avgt 3 76.517 ± 2.979 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
com.bogus.hokus.pokus.Clazz2
avgt 3 75.369 ± 3.217 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
edu.bogus.hokus.pokus.a.Clazz3
avgt 3 97.546 ± 12.164 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
de.bogus.hokus.b.Clazz4
avgt 3 82.240 ± 14.489 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
jp.bogus.c.Clazz5
avgt 3 65.274 ± 3.938 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
cn.d.Clazz6
avgt 3 50.239 ± 0.456 ns/op
```
Previous results:
```
NamePatternConverterBenchmark.benchNamePatternConverter
org.bogus.hokus.pokus.org.bogus.hokus.pokus.org.bogus.hokus.pokus.RetroEncabulatorFactorySingleton
avgt 3 209.477 ± 62.262 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
org.bogus.hokus.pokus.Clazz1
avgt 3 76.083 ± 8.123 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
com.bogus.hokus.pokus.Clazz2
avgt 3 76.109 ± 19.293 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
edu.bogus.hokus.pokus.a.Clazz3
avgt 3 91.913 ± 46.029 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
de.bogus.hokus.b.Clazz4
avgt 3 74.703 ± 28.086 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
jp.bogus.c.Clazz5
avgt 3 59.788 ± 7.298 ns/op
NamePatternConverterBenchmark.benchNamePatternConverter
cn.d.Clazz6
avgt 3 41.092 ± 19.660 ns/op
```
--
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: notifications-unsubscr...@logging.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
