Hello, several days ago I’ve proposed some improvements into asm library: - https://gitlab.ow2.org/asm/asm/commit/ef945457bc19dbc09c02bd21b52f1950990f33... - https://gitlab.ow2.org/asm/asm/commit/9b914d524b7fe7ea694fc11ec784e67133ba38... The patches were accepted and then it turned out the code I’ve patched is similar to what we have in JDK. So I propose the same changes to be accepted into JDK 12. They improve code by using more simple constructions and replacing redundant hand-written code with JDK-provided API. As of performance, usage of String::replace instead of handwritten replacement (see Type::appendDescriptor) is much faster and memory-effective (tested with JDK 11) even on simplest class names like java.lang.String: Benchmark                                                      Mode  Cnt    Score    Error  Units CharacterReplaceBenchmark.manualReplace                        avgt  25    57,036 ±  1,132  ns/op CharacterReplaceBenchmark.stringReplace                        avgt  25    17,648 ±  0,699  ns/op CharacterReplaceBenchmark.manualReplace:·gc.alloc.rate.norm    avgt  25  112,000 ±  0,001    B/op CharacterReplaceBenchmark.stringReplace:·gc.alloc.rate.norm    avgt  25    56,000 ±  0,001    B/op Corresponding benchmark available here: https://github.com/stsypanov/benchmarks/blob/master/benchmark-runners/src/ma... https://github.com/stsypanov/benchmarks/blob/master/benchmark-source/src/mai... Usage of String::substring instead of StringBuilder.append(CharSequence, int, int) (see Type::getDescriptor) is also faster and memory-effective, results described here: http://mail.openjdk.java.net/pipermail/core-libs-dev/2019-January/057834.htm... Also, is there any workload to measure effect of attached patch? Regards, Sergey Tsypanov