31st August 2017, 2 min read

Parsing comma-separated integers in Java

2 thoughts on “Parsing comma-separated integers in Java”

1. Jay says:
   November 13, 2017 at 8:26 pm

   Your manual implementation doesn’t even work. The results don’t correlate with the actual values; partially because of some false assumptions. I could go on but it was actually easier to write a working implementation with ~4x the throughput.

   m.l.m.p.ParseInt.manualSplit thrpt 5 12624.529 � 445.839 ops/s
   m.l.m.p.ParseInt.monolithicSplit thrpt 5 41031.703 � 5054.782 ops/s

   @Benchmark
   public int[] monolithicSplit(BenchmarkState s) {
   final String text = s.myarray;
   final int length = text.length();
   final int limit = (length / 2) + 1;
   int[] array = new int[limit];
   int pos = 0, tmp = 0;
   boolean neg = false;
   for (int i = 0; i < length; i++) {
   char c = text.charAt(i);
   if (c == ',') {
   array[pos++] = neg ? 0 – tmp : tmp;
   tmp = 0;
   neg = false;
   } else if (c == '-') {
   neg = true;
   } else {
   tmp = (tmp << 3) + (tmp << 1) + (c & 0xF);
   }
   }
   array[pos++] = neg ? 0 – tmp : tmp;
   int[] result = new int[pos];
   System.arraycopy(array, 0, result, 0, pos);
   return result;
   }

   1. Daniel Lemire says:
      November 13, 2017 at 10:03 pm

      Your approach is indeed better than my manual hack and I have updated the code accordingly. Note that this was indicated in my post.