tree e1d038811ba3d846a41de03f7745f8819e9a7939
parent 2e374a9c6e3d4ed71f0145de287c4b2fe43c76d6
author ulfjack <ulfjack@google.com> 1562070038 -0700
committer Copybara-Service <copybara-worker@google.com> 1562070093 -0700

Query performance: Reverse output buffering

Instead of converting String -> byte[] first and then buffering, buffer
the String data, and then convert. Bulk conversion is generally more
performant than converting each little piece individually, and also
allows us to avoid constructing intermediate String instances that are
immediately thrown away.

Writer is the intended API for writing character data, PrintStream is not.

I don't see a significant performance difference with this change by
itself, or maybe even a slight regression, but I think it's still
worthwhile, as I can see an improvement in combination with some other
(otherwise unrelated) changes.

PiperOrigin-RevId: 256145525