Thanks for the pointer. Yeah, that's one the pieces I think we should have to do an optimal job of rewriting collection toString methods.
On Mon, Apr 15, 2013 at 1:03 PM, Steven Schlansker < stevenschlans...@gmail.com> wrote: > > On Apr 15, 2013, at 12:21 PM, Martin Buchholz <marti...@google.com> wrote: > > > On Mon, Apr 15, 2013 at 11:31 AM, Martin Buchholz <marti...@google.com > >wrote: > > > >> > >> OTOH, I'm guessing you are trying to improve the performance of > operations > >> like List.toString. > >> More efficient (single copy char[]) would be to collect all the > >> sub-CharSequences in a CharSequence[], pre-compute the final length of > the > >> char[], allocate an array of exactly the required length, and create the > >> final string directly from that using the package-private constructor > (but > >> in the unlikely event that a subsequence changed in size while > concat'ing, > >> be prepared to resize the array). > >> > > > > Proceeding further along this train of thought, I might start with > > AbstractCollection.toString() (and similar methods) and attempt to make > it > > maximally efficient. > > Maybe add a method to JavaLangAccess to make a String directly from a > > perfectly sized array (as needed elsewhere?). Maybe create a > > StringBuilder-like class that works better for typical use cases? > > For what it's worth, a patch that I contributed and Mike (and others) then > rewrote > contains this functionality already: > > > http://cr.openjdk.java.net/~mduigou/JDK-8006627/2/webrev/src/share/classes/sun/misc/JavaLangAccess.java.patch > > It's not merged yet though. > > >