On Feb 29, 2008, at 8:51 AM, Rémi Forax wrote: > No, i prefer patching the current OpenJDK lib or use alternatives when > the current API doesn't fit my needs.
Yes, please, do contribute patches into the OpenJDK. On Feb 29, 2008, at 8:19 AM, Charles Oliver Nutter wrote: > > > - java.util.regex can't handle alternations over arbitrarily large > input > without blowing the stack. It also can't handle our (Ruby's) > byte[]-based strings. So we switched first to JRegex, which doesn't > recurse, and then to Joni, a bytecode-based port of Oniguruma. The OpenJDK regex is a tree-walking interpreter, and needs to be updated in a future release to a Gnu-style backtracking bytecode engine. This is not an architectural problem, but an implementation weakness in the first release. > - java.math.BigInteger has a grossly inefficient implementation of > doubleValue, which we need. So we're looking at the JScience library's > LargeInteger instead. Yes, it is grossly inefficient, but it just needs a little love and attention. If it's just an implementation problem, don't pitch it, patch it! Surely there's a reasonable way to get what you want immediately, something like: > BigInteger bigi = ...; > int scale = Math.max(0, bigi.bitLength() - 63); > long val = bigi.shiftRight(scale).longValue(); > (check for overflow by examining scale and val) > > long bits = Double.doubleToRawLongBits((double)val); > bits += scale << (64-12); // patch exponent field > double result = longBitsToDouble(bits); That could be a utility routine in the JRuby runtime, and later a patch to OpenJDK. -- John --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "JVM Languages" group. To post to this group, send email to jvm-languages@googlegroups.com To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/jvm-languages?hl=en -~----------~----~----~----~------~----~------~--~---
