Narrow scope vs. efficiency

Question

narrow scope:
for (cond) {
  byte[] buf = new byte[const];
  in.read(buf);
  out.write(buf);
vs. efficiency:
byte[] buf = new byte[const];
for (cond) {
  in.read(buf);
  out.write(buf);
}What do you prefere? Do you use a profiler to make the decision?

Answer 1

But what about if buf is used throughout the procedure
(locally). It should then be declared at the class
scope for efficiency. You'll not be able to link the
buf to the method by "{}".In the choise between class or method scope I probably would declare buf in the method scope. At least if the method isn't called very often. You'll have to use common sense. On one hand you want an as narrow scope as possible but on the other hand a narrow scope cannot motivate the creation of zillions of objects in an innermost loop.