Motivation: When decoding stomp frames a lot of unnecessary character arrays are created when parsing headers. For every header, an array is created to read the line into and then more when splitting the line at the colon. Modifications: Parse key and value of a header while reading the line instead of afterwards. Reuse a single AppendableCharSequence. Reduce initial size of AppendableCharSequence when reading the command as it is expected to be short. Result: Allocations when parsing stomp frames have dropped significantly. |
||
---|---|---|
.. | ||
main/java/io/netty/handler/codec/stomp | ||
test/java/io/netty/handler/codec/stomp |