Can someone else pls have a look at http://issues.apache.org/bugzilla/show_bug.cgi?id=35904
This is holding up the 1.1 release. I like the support for all of the rounding modes that we get by using BigDecimal and - though I have not verified this - I would expect the results to be exact over a wider range of <value, scale> combinations than direct implementation would give. On the other hand, we have the string conversion and object creation overhead. I haven't done any microbenchmarks, but if someone else has the time, it might be good to look at direct impl of, say, ROUND_HALF_UP vs. the current impl. Opinions please... I am inclined to close this as FIXED (once more tests have been added). Note that this will commit us to supporting all of the rounding modes, even if we decide to replace the implemenation later. Some of these - e.g. ROUND_NOT_NECESSARY - may have funny semantics for some double / float values. Additional unit tests to document / demonstrate / validate exactly how all of these modes work will be needed in any case. Phil --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]