On 2016-10-01 18:54, Martin Buchholz wrote:
On Sat, Oct 1, 2016 at 5:45 AM, Claes Redestad <claes.redes...@oracle.com <mailto:claes.redes...@oracle.com>> wrote: On 2016-10-01 02:41, Martin Buchholz wrote: https://bugs.openjdk.java.net/browse/JDK-8167002 <https://bugs.openjdk.java.net/browse/JDK-8167002> http://cr.openjdk.java.net/~martin/webrevs/openjdk9/xml-id-validation-by-hash/ <http://cr.openjdk.java.net/~martin/webrevs/openjdk9/xml-id-validation-by-hash/> +1, but I have to ask what the intended benefit of writing: ((fIds != null) ? fIds : (fIds = new HashSet<>())).add(name); rather than keeping the pre-existing pattern: if (fIds == null) fIds = new HashSet<>(); fIds.add(name); If this is about bytecode optimization to help with inlining or such, the latter actually generate a more compact method (14 vs 16 bytecodes). The intent was (by instinct) to not code a re-read of a field, but ... yeah .... that wasn't achieved. That could be done correctly using public void addId(String name) { HashSet<String> x; if ((x = fIds) == null) fIds = x = new HashSet<>(); x.add(name); } but I'll just revert to if (fIds == null) fIds = new HashSet<>(); fIds.add(name);
Ok, thanks for the explanation!
